Article 1 of 24: Should Fernanda Follow a Rule or Learn from Experience?

Understanding Rules and Machine Learning

Fernanda is an estate agent (a girl’s gotta do what a girl’s gotta do!). She knows her neighborhood, enjoys helping people find homes and has developed an admirable patience with common phrases such as “deceptively spacious” and “easily-maintained garden”.

Like most experienced professionals, Fernanda uses a mixture of direct instructions she has to follows, and lessons from her previous work. More specifically, some of Fernanda’s decisions follow instructions supplied by her agency, while others draw on patterns she has noticed through experience (for example, which property details tend to be associated with higher sale prices).

Now suppose Fernanda wants a computer to help her make an initial assessment of new properties. How should she teach it what to do?

She could give it an explicit instruction, such as “Follow this rule when assessing a property”.

Alternatively, she could give it records of previous sales and a procedure for finding a useful pattern. In this case her instruction is less explicit: “Use these examples to work out how to assess another property”.

That brings us to the question at the heart of this article:

What is the difference between giving a computer a rule to follow and giving it examples from which to learn?

Before answering, let us give Fernanda a specific problem:

1. The problem: Which homes are likely to sell for at least $300,000?

Fernanda’s agency wants to organize incoming properties into two categories. Why? Well, I am not an estate agent but I guess this may help match properties to buyers’ budgets (maybe Fernanda is a bit posh and prefers to only work on the houses that rich clients can afford?). Anyway, let’s call the 2 possible categories YES and NO (see Table 1 below):

Category	Description
YES	Includes houses which are expected to sell for at least $300,000
NO	Includes houses which are expected to sell for less than $300,000

Table 1: Categorizing Houses Based on Price

Note that at this stage, the agency does not need an exact price estimate. It simply wants an initial category to help organize its work.

Now, imagine that for our example, Fernanda receives only two details about each property (see Table 2 below):

Available Information	Possible Values in Our Example
Number of Bedrooms	2, 3, 4
Location	Standard or Prime

Table 2: Property Details Available to Fernanda

Her required answer, for each property under consideration, is simply YES (it is at least $300,000) or NO (It is less than $300,000).

Of course, actual property prices depend on many other things, such as floor area, condition, transport links, whether the “peaceful outlook” overlooks a nightclub, etc., but for our purpose here, considering only 2 features are good enough. In other words, we are deliberately simplifying the situation so that we can see what happens at every step of the decision-making process. We can make the world more complicated later. Just wait and we get there!

“What is your objective?”, I hear you yelling at me. Here it is:

Our objective:

Using the available property details to predict its price category.

2. First approach: Someone supplies a rule

Suppose the agency manager (good chap but not important enough to have a specific name for now!) gives Fernanda this instruction:

Property Price Rule:

If a property has at least three bedrooms, predict YES. Otherwise, predict NO.

The reasoning sounds plausible. Larger homes often sell for more, and bedroom count seems a convenient starting point. Beside plausibility, this rule is rather simple. It will not take much to remember and apply it to new houses so, it certainly sounds useful.

With this rule in place, Fernanda can now apply the instruction immediately (see Table 3 below):

Number of Bedrooms	Location	Apply the Rule	Prediction
2	Prime	Fewer than three bedrooms	NO
3	Standard	At least three bedrooms	YES
4	Prime	At least three bedrooms	YES

Table 3: Applying a Property Price Rule to Selected Properties

Notice from the results in Table 3 above that location makes no difference under this rule. The information about location is available, but the instruction does not use it at all.

A computer could also follow exactly the same procedure:

Receive the bedroom count,
Check whether it is at least three,
Return YES if it is, and NO otherwise.

This is a rule-based approach. A person has supplied the decision rule, and the computer applies it.

There is nothing inherently wrong with that. The instruction is certainly clear, easy to inspect and inexpensive to apply. The question (an important one) is whether it gives useful answers.

The manager’s confidence, although impressive, does not answer that question (managers rarely gives their charges confidence so, in some sense, this is not surprising😊).

3. Let us check the rule against experience

Fernanda being who she is (meticulous and not blindly trusting her manager), collects six completed sales. Because these properties have already sold, she can compare a prediction with what actually happened (see Table 4 below).

Property	Number of Bedrooms	Location	Actual Sales Price	Correct Category	Prediction: Using “At Least 3-Bedroom” Rule
A	2	Standard	$200,000	NO	NO (correct)
B	3	Standard	$250,000	NO	YES (incorrect)
C	4	Standard	$280,000	NO	YES (incorrect)
D	2	Prime	$300,000	YES	NO (incorrect)
E	3	Prime	$400,000	YES	YES (correct)
F	4	Prime	$450,000	YES	YES (correct)

Table 4: First Round of Prediction

As you can see from Table 4 above, the rule gets three out of six cases right.

That is to say that the manager’s particular rule (based on the rule-based approach) is 50% correct as seen below:

No. of Correct Predictions ÷ Total Number of Cases = 3 ÷ 6 = 0.5, or 50% correct

For now, counting mistakes is enough. We will examine more detailed ways to assess predictions later in the series.

Look closely at the three errors:

The three-bedroom Standard property (Property B) sold for less than $300,000.
So did the four-bedroom Standard property (Property C).
The two-bedroom Prime property (Property D) reached $300,000 despite having fewer bedrooms.

Fernanda begins to suspect that bedroom count alone is not a particularly useful basis for this decision.

Note something important:

The rule is being followed perfectly. It is the rule’s predictive usefulness that is in question.

That distinction matters. A computer can execute an instruction flawlessly and still produce a poor answer.

Now what?

4. What exactly does “learning from experience” mean?

As we just observed, the manager’s rule has made three mistakes across six properties. What should Fernanda do next?

She could inspect the historical sales herself, decide that location looks more useful and write a replacement rule. That would be Fernanda learning from experience. A computer given her replacement rule would simply follow this new instruction.

But we want to explore a different possibility and ask:

Can the computer use those historical examples to choose a prediction rule?

To see how this might work, let’s start with something small enough that we can follow every step.

Let’s assume that Fernanda has three suggestions:

Perhaps the manager’s original bedroom rule is still worth considering, or
Perhaps requiring at least four bedrooms would produce better results, or
Perhaps location is more useful than bedroom count.

Each suggestion proposes a possible relationship between information about a property and its price category. We can express those suggestions as three alternative prediction rules (see Table 5 below):

Candidate Rule	Rule Description
Rule 1	Predict YES if the property has at least three bedrooms, otherwise predict NO
Rule 2	Predict YES if the property has at least four bedrooms, otherwise, predict NO
Rule 3	Predict YES if the property is in a Prime location, otherwise, predict NO

Table 5: Enhanced Property Price Rules

Note that these rules are alternatives. The computer will compare them and select one. It will not apply all three in sequence to reach a single answer.

Now we give the computer the following procedure:

Apply each candidate rule to all six historical properties,
Compare each prediction with the correct category, obtained from the actual sale price,
Count how many mistakes each candidate makes,
Select the candidate with the fewest mistakes. If candidates tie for the fewest mistakes, select the one listed first (the tie instruction simply makes our procedure complete. It does not mean that an earlier position in the table makes a rule more trustworthy. This, also demonstrates that sometime there are more than one way to get to a reasonable answer😊).

But hang on. Aren’t we still using rules?

Yes. But the important change is how the prediction rule is chosen.

Previously, the manager chose a rule and instructed the computer to use it (no fuss, no muss). Its performance on the historical examples played no part in that choice.

Now, we supply several candidates and a method for comparing them. The computer uses the historical examples to determine which candidate to select. With different examples, a different candidate could win (ain’t life wonderful!).

There are therefore two separate activities to consider this time:

Learning: Using historical examples with known answers to select a prediction rule.
Predicting: Applying the selected rule to another property whose sale price is not yet known.

The selected rule is what we call the learned model in this example (it is “learned” because it has used actual examples to “learn” what works best and it is “model” because it aims to model or capture or predict what happens with new or unseen examples).

We should also be clear about the limits of what the computer is doing. We supplied the three candidates. The computer is testing the possible relationships we suggested. So, it is not inventing new ones or discovering other property characteristics for itself.

This is a deliberately limited (stylized) example of Machine Learning. Later, we will explore methods that adjust numerical settings rather than select from a short list of complete rules. For now, this small example lets us see exactly what is supplied by people and what is determined from the data (existing examples).

Learning from examples can therefore produce a rule. There is no contradiction: The rule is the result of the learning procedure.

Let’s now run that procedure and determine which of our three candidates makes the fewest mistakes on the six historical sales.

5. Let the candidates compete

We already know that Rule 1 makes three mistakes. What happens with the other two? Table 6 below summarizes the results for all 3 rules:

Property	Number of Bedrooms	Location	Actual Sales Price	Correct Category	Rule 1: At Least 3 Bedrooms	Rule 2: At Least 4 Bedrooms	Rule 3: Prime Location
A	2	Standard	$200,000	NO	NO	NO	NO
B	3	Standard	$250,000	NO	YES	NO	NO
C	4	Standard	$280,000	NO	YES	YES	NO
D	2	Prime	$300,000	YES	NO	NO	YES
E	3	Prime	$400,000	YES	YES	NO	YES
F	4	Prime	$450,000	YES	YES	YES	YES
Number of Mistakes (out of 6)					3	3	0

Table 6: Second Round of Predictions

It is clear now: Rule 3 wins😊.

So, the selected rule is:

If the property is in a Prime location, predict YES. Otherwise, predict NO.

Note that the computer did not select this rule because it understands the appeal of a neighborhood café! It selected it because, among the permitted candidates, this rule made the fewest mistakes on the supplied examples.

And this is the important transformation:

We began with examples of properties and their known outcomes,
We applied a procedure for comparing possible rules,
We ended with a selected rule that can be used on another (new) property.

CRITICAL POINT TO STRESS: If the historical outcomes had been different, another candidate could have won. That dependence on the examples is what makes this a learning process.

6. Time to introduce a few names

We have already encountered several ideas that appear throughout Machine Learning. Now we can attach their usual names (see Table 7 below).

What We Have Just Used	Equivalent Machine Learning Term
Bedrooms and location	Features: The input details available for making a prediction
The known YES or NO category for each historical property	Label (or Target): The answer used to assess a prediction
The six historical examples used to choose a rule	Training Data
The procedure that compares candidates and selects one	Learning Algorithm
The selected rule	Trained Model
Applying that selected rule to another property	Prediction (or Inference)

Table 7: Some Useful Machine Learning Terminologies

Let us pause on the difference between the learning algorithm and the learned model as this is a common point of confusion among many students:

A learning algorithm is the procedure we used to select a rule.
- Example: Try the three candidates and select the one with the fewest mistakes.
A learned model is the rule we obtained:
- Example: Predict YES for Prime locations and NO for Standard locations.

These two concepts are related, but they are not the same thing.

Also notice that the actual sale price helped us establish the correct historical category. It is not an input to our prediction. This is an important distinction to make because for a new property, we do not yet know its eventual sale price (there is no actual price yet). That is what we want to predict (output, rather than input to our model). If we supplied that price to predict its price category, our apparent brilliance would be rather less impressive than advertised😊.

OK. Here is another Machine Learning definition that I could have (maybe prematurely) added to Table 7 above:

Because we learn from examples that include known answers, this whole approach is called Supervised Learning.

The “supervision” consists of those supplied answers. It does not require Fernanda to stand beside a laptop saying, “Come on, you can do better” (knowing her, she may actually do that but that is not the standard practice in Machine Learning!).

7. Did we still program the computer or not? I am confused!

Yes. Absolutely. We did not remove human choices from the process. In particular, people decided:

What question to answer,
Which property details to provide,
Which historical examples to use,
Which candidate rules to consider,
How to compare their performance.

Our learning procedure searches only among the possibilities we allow. In this example, it cannot suddenly invent a sophisticated calculation involving floor area, interest rates and the emotional value of a south-facing balcony. After all, we did not provide those inputs or permit that kind of calculation.

The difference is which part people specify directly (see Table 8 below).

Approach	What People Specify	What the Computer Does
Supplied decision rule	The rule that produces YES or NO	Applies the rule
Learning from examples	The candidate rules, examples and selection procedure	Uses the examples to select a rule, then applies it

Table 8: Some Useful Machine Learning

Machine Learning is still implemented through programmed procedures. The defining feature here is that the resulting predictive behavior is determined through the examples.

We will come back to this during this 24-topic series we are embarking on but for now, let’s just note that in more complex methods, the procedure may adjust thousands or millions of numerical settings rather than choose among three short rules. We will reach those methods gradually. So, let’s hold our horses and simply appreciate the fact that the essential idea is already visible through this simple example, thanks to Fernanda with a little help from her, as yet, nameless boss!

8. So, a learned model can itself be a rule?

Yes, my dear and this is worth making explicit:

If we inspect Fernanda’s final model, it looks like an ordinary instruction:

If Prime, predict YES. Otherwise, predict NO.

Someone could have written exactly that instruction without running a learning procedure (I mean someone really clever). But the final wording of the rule alone would not tell us how it was obtained.

The distinction lies in its origin: Was it supplied directly, or selected through a learning procedure using examples?

That is why “rules versus learning” is a useful opening question, but not a complete division of the world. Learning can produce rules.

In fact, some methods produce rules that are easy to read. Others produce numerical relationships that are much harder to inspect. Both can be learned from examples.

9. Does “zero mistakes” mean Fernanda has solved the problem?

Hmmm. Err! It means that the selected rule made zero mistakes on the six examples used (training data) to select it. That is certainly encouraging. And yet, it is also a very specific statement. I mean, it is specific to these 6 samples only. What would happen if we can extend the training data from 6 to, say, 8?

Let’s for example, assume that Fernanda now checks two additional historical sales that she deliberately kept out of the selection process (if I forgot to tell you before, let’s settle this now: Fernanda is quite clever. Can you see why keeping some historical examples out of the training data for further verification after a rule is selected is a particularly useful idea?). Now let’s see the 2 new examples left out of the original list (see Table 9 below):

Property	Number of Bedrooms	Location	Actual Sales Price	Correct Category	Model’s Prediction
G	2	Standard	$220,000	NO	NO
H	2	Prime	$320,000	YES	YES

Table 9: Second Round of Predictions

Well, well, well! Both predictions are correct.

These examples provide a small check of how the model performs on cases it did not use for learning. They do not establish that it will perform well on every future property. Let’s face it: Two successful checks are only two successful checks, however enthusiastically the agency’s marketing department might describe them.

Later, we will examine how to separate data for learning, choosing between approaches and evaluating the final result.

For now, just remember this:

Doing well on the examples used for learning is not enough. We also need evidence from other examples.

10. And one day another property arrives…

And today is the day when Fernanda receives a special two-bedroom property in a Prime location. And as expected by now, the model predicts YES. Loud and clear.

But eventually, the property sells for $260,000. What can I say. Sometimes you are desperate to cash in and can’t wait for the price to reach or exceed the minimum value Fernanda is expecting! The world does not revolve around Fernanda, does it?

The bottom line is that in this case the prediction was wrong.

There are legitimate reasons for this: Perhaps the property needed major repairs. Perhaps its floor area was unusually small. Perhaps market conditions had changed. We would need evidence to find out, of course.

Whatever the explanation though, our model had access only to bedrooms and location. And it could not take account of details it never received.

As for Fernanda, she now has several possible responses:

Collect more examples to see whether the mistake reflects a wider pattern,
Investigate whether important information is missing,
Consider a broader set of possible models,
Decide whether this simple method is useful enough for its intended role.

Notice what she should not do: Assume that every mistake can be fixed by shouting “more AI” at the problem!

The truth is that sometimes we need a better learning method. Sometimes we need better data. Sometimes we need a clearer question. Often times we need a combination of these 3.

And sometimes a simple prediction should remain a suggestion for a person to review and reject or approve (we call this human in the loop).

One more point: Our model does not automatically change because a sale has occurred. In this example, someone must record the new outcome and run the learning procedure again. Automatic updating would be an additional design choice. It is certainly possible but it has less to do with Artificial Intelligence or Machine Learning and more to do with the Engineering aspect of predictive work.

11. When should Fernanda simply follow a supplied rule?

Not every task requires learning from examples. Sometimes, the agency has already decided what must happen, and Fernanda’s job is to follow that instruction. For example:

Only send a client’s contact details to an outside photographer if the client has given permission.

In this case Fernanda checks whether permission has been recorded. If it has, she may send the details. Otherwise, she must not. It is really that simple.

In other words, there is nothing here for the computer to learn about whether permission should be required. That requirement comes from the agency’s policy. Even if employees sometimes ignored it in the past, their behavior would not make it acceptable and should not be considered as valid training data for extracting the correct patterns regarding the application of the policy (it may be used to identify patterns of bad or wrong behavior but that is a different question altogether). A bad habit does not become a good policy just because we have plenty of examples of it!

Predicting a property’s price category is different. The agency cannot issue an instruction that makes a house sell for at least $300,000 (although I suspect the manager would welcome that option😊). The sale price is an unknown outcome, and historical examples may help us predict it.

The key is to ask what the task needs:

Applying an established requirement, or
Estimating an unknown outcome?

Even for prediction, Machine Learning is not always necessary. A simple supplied rule may work well enough. We should choose an approach because it helps solve the problem, not because it sounds more impressive.

12. Often, the sensible answer is to use both

For example, Fernanda’s agency might combine:

A learned model that suggests a property’s price category,
A supplied rule requiring certain checks before a listing is published,
A person who reviews unusual cases and remains responsible for the final action.

Each part performs a different role.

The prediction draws on patterns in examples. The checks enforce explicit requirements. The person deals with context and responsibility beyond the narrow prediction task.

As we progress through this series, we will return to this question:

Which part of the work should be learned from data, which part should be specified directly, and where should human judgment remain?

That is often more useful than asking whether a system is “AI-powered”, even though just mentioning anything AI fits nicely on a company website!

13. Question time

Before moving on, consider these questions:

Fernanda writes “If Prime, predict YES” after personally inspecting the sales table. A computer then follows it. Has the computer learned the rule?
A programmed procedure compares three candidate rules using historical examples and selects the one with the fewest mistakes. Where does learning occur?
Our selected rule gets every training example right. What should we check next?
Does the pattern in our small dataset prove that changing a property’s location would cause its price to rise?

14. What have we learned?

We saw how Fernanda began with a simple instruction about bedrooms. By comparing its predictions with historical sales, she discovered that it made several mistakes. We then followed her as she specified a procedure that used those historical sales examples to select a better rule candidate for prediction.

Along the way, we saw that:

A computer can follow a rule correctly while making poor predictions,
A learning algorithm uses examples to determine a model,
The learned model can itself be a readable rule,
People still choose the objective, data, available inputs and learning procedure,
Success on training examples does not guarantee success on new cases,
Explicit rules, learned models and human judgment can work together.

Our example is intentionally small. Real learning systems can search through far more possibilities, and they do not all learn by counting mistakes across three candidate rules.

Nor does our comparison claim that people and computers learn in the same way. Fernanda brings understanding and experience to her work. Our little procedure compares predictions and counts errors. Keeping that boundary clear helps the analogy do its job.

15. Next: Can Fernanda predict an actual price?

A YES or NO category is useful, but sooner or later a seller will ask:

That is lovely, Fernanda. But how much do you think my house will sell for?

And in this case, a simple YES is unlikely to be accepted as a price estimate.

In the next article, Fernanda will combine property details to produce a number. We will examine what each detail contributes, why a starting amount matters and how a simple calculation can be adjusted using past sales.

Question time: Suggested answers

Here are some answers to the questions posed in section 13 above:

No. In this description, Fernanda learned from the examples. The computer received and applied her rule.
Learning occurs when the procedure uses the examples to select the model. Applying the selected model afterwards is prediction.
Check its performance on examples not used to select it. Also consider whether the data represent the properties on which it will be used.
No. We have identified a predictive pattern in a small dataset. We have not established a causal effect. We will explore that distinction in a later article.

Feedback please😊.

Next article: How Does Fernanda Put a Price on a Home?
Understanding Linear Regression

Back to the AI Explained Roadmap

Weekly Technology Brief #1

From Bigger Models to Bigger Systems

8 October 2026

Dear readers,

Here we go, the very first weekly technology brief…

If there is one idea running through this week’s technology news, it is that the interesting action is moving outward. AI is no longer just a story about better models (this has, arguably, been the case for the best part of 2026 so far). The constraints, opportunities and risks are increasingly appearing in the systems around them. This includes the agents that act on our behalf, the memory chips that enable lots of the fancy work, the electricity that keeps the data centres running, the biological datasets that let the models reason about living systems, and the verification machinery needed before we trust claims about quantum computers, autonomous robots, or the incredible number of mathematical proofs thrown at the scientific community.

That shift matters because it changes the question. Instead of asking only, “Which model is smartest?”, we increasingly need to ask: What infrastructure makes the intelligence useful? What becomes the next bottleneck? And, perhaps most importantly, how do we know that the system actually works outside a carefully controlled demonstration (this last one itself deserves a deep dive soon)?

1. The Enterprise AI Race Is Becoming a Control Problem

Google Cloud used its Gemini at Work event this week to introduce what it calls a single universal Gemini agent for work. The important part is not that the agent can answer questions, create content or write code. After all, we have seen plenty of that already. The more interesting part is where Google is putting the surrounding machinery.

The agent is designed to operate across Gmail, Drive, Docs, Slides, Sheets, Chat and Calendar, while carrying business context between them. Google is also emphasizing identity, permissions, authorization, secure sandboxing, network gateways, policy management and spending controls. In other words, the product is beginning to look less like a clever chatbot and more like an employee operating inside an institutional control system. At least that is the idea. Whether or not any of the individual angles are well covered is a matter of further verification and a source of more articles!

However, this is an important transition. A model that occasionally produces a bad answer is irritating. An agent with permission to send an email, alter a spreadsheet, query confidential data, execute code or initiate a business process can create a very different class of problem. As the ability to generate actions becomes cheaper, the value of deciding which actions are permitted, and independently checking what actually happened, should rise. Also, let’s not forget that verifying, and ultimately pulling the plug on, a model generating a single wrong answer is easy. Once an agent gets to work, the potential for generating nonsense also scales and it becomes quite possible that by the time we know what is wrong the extent of the any damage cause could be truly substantial.

In any case, the transition means that the emerging stack may actually look something like this:

Model → Agent → Permissions/Control Plane → Execution → Verification.

The model companies will naturally compete on intelligence. Enterprises, however, may eventually care just as much about the layers around that intelligence, including identity, audit trails, rollback, cost control, data sovereignty and liability.

As a result of this transition, intelligence itself may become abundant, but trustworthy control probably may not necessarily follow the same trajectory.

2. AI’s Electricity Problem Is Turning Existing Nuclear Plants into Technology Infrastructure

The physical side of AI became unusually visible this week when Google announced a major agreement with Constellation Energy. The arrangement is intended to unlock 890 megawatts of additional nuclear capacity through upgrades to 11 reactor units across six existing U.S. nuclear plants. Google says its agreements involving uprates and reactor restarts have now enabled more than 1.5 gigawatts of new nuclear capacity.

The interesting detail is that this is not primarily a story about waiting for futuristic small modular reactors. Google is financing improvements to infrastructure that already exists. That is revealing. When electricity demand is rising quickly, the fastest new megawatt may be the one squeezed out of an existing asset rather than a completely new power station.

This is another example of the bottleneck migrating downstream. Better AI models create demand for more accelerators. More accelerators create demand for memory and networking. More servers create demand for data centers. Data centers create demand for power, cooling, land, grid connections, permits and financing.

Build this, and that becomes the next bottleneck!

The implication is broader than nuclear power. AI infrastructure is becoming an industrial system, and industrial systems are constrained by things that software companies historically did not have to worry about quite so much, such as turbines, transformers, transmission lines, water, construction schedules, local communities and the cost of capital. The future of AI may still be written partly in code but an increasing amount of it will be built in concrete and copper. Sounds strangely poetic, right? 😊.

3. Memory Is Still Telling Us Where the AI Bottleneck Lives

Samsung’s preliminary third-quarter results provided another reminder that the AI boom is not confined to GPU vendors. The company projected operating profit of 107.4 trillion won for the quarter, nearly nine times the level a year earlier (12.17 trillion), with strong demand and tight supply in memory playing a central role. Detailed results are still to come later this month, so the precise divisional picture deserves some patience.

The more useful lesson is structural. Modern AI accelerators can perform extraordinary numbers of calculations, but those calculations are useful only if data can be moved to and from the processors quickly enough. That is why high-bandwidth memory has become such an important part of the AI system, and why conventional DRAM and NAND conditions still matter as capacity is redirected and supply remains tight.

There is also a warning here. Extraordinary earnings do not prove that extraordinary growth continues indefinitely. Semiconductor markets have a long history of turning shortages into investment, investment into capacity and capacity into eventual oversupply. AI may have changed the scale of demand, but it has not repealed economics. In other words, the inherent cyclicality of memory products remains a real risk for investors.

So, the question is not simply whether memory demand is strong. It plainly is. The better question is whether supply can expand faster than AI workloads broaden. And where the constraint moves if memory becomes less scarce. Networking? Advanced packaging? Power? Capital? This is why following bottlenecks is often more informative than following product announcements.

We used to say (well, I used to have this as my LinkedIn moto for a long time): In data we trust. As an investor, especially in the crazy age of AI one should say: In bottlenecks we trust!

4. AI for Biology Is Discovering That Data Must Sometimes Be Manufactured

One of the most interesting developments of the week came from biology rather than the usual AI laboratories. The U.S. government, Biohub, Meta, Google DeepMind and Isomorphic Labs are participating in a roughly $1.8 billion effort to build large, standardized datasets for AI-driven biological research.

This matters because biology exposes one of the weaknesses in the familiar internet-era AI story (let me be clear: Weakness in this context is not exactly a negative word. If you get biological data wrong and then use it to make a decision that may subsequently prove wrong, people may die. I mean, literally. That is why anything to do with biology has been traditionally more regulated and, by definition, slower. And yes, that includes data availability. While there is always room for improvement everywhere, including in biology, expecting every industry to run under the start-up maxims of “break it first and see where it ends”, or “fail fast” is not exactly advisable!). Language models benefited from enormous quantities of text that already existed. In biology, the most valuable training data may not exist yet. Researchers have to create it through experiments. This is not my area of expertise but I read this includes perturbing cells, measuring what happens, standardizing the observations, then training models, generating new hypotheses and finally running more experiments.

In other words, the loop becomes:

Question → Experiment → Data → Model → Prediction → New/Refined Question → New/Refined Experiment → …

Each cycle uses the evidence generated by the previous one to refine the next question and experiment.

That is potentially much more consequential than simply attaching an AI assistant to a laboratory (there is nothing wrong with the AI assistant but, eh!). If models can help decide which experiments are most informative, and automated laboratories can execute those experiments at scale, scientific discovery itself becomes a closed learning loop. Granted, scientific discovery was always expected to be loopy (using feedback to refine questions and experiments) so, the concept may not be new. What is new is our ability (thanks to modern AI) to accelerate and increasingly automate parts of that loop, and to do it at scale.

But there is an important discipline to maintain here. A better biological model is not a medicine. Neither is a plausible mechanism the same as a clinical outcome. Drug development still has to cross the brutal sequence from biological insight to candidate, preclinical evidence, human safety, efficacy and eventually scalable treatment (I just told you about the regulations and the slower pace of change in biology, did I not?). AI may accelerate parts of that chain without magically eliminating the chain (I am, assuming, that biology will not fall into a Muskian loophole any time soon. Remember DOGE?).

5. Quantum Computing Is Entering the ‘Show Me’ Phase

Quantum computing produces no shortage of impressive announcements (on a personal note, and as a Physics graduate, I have to admit that there is a lot to be excited about in this field these days. It is so, very, grooooovy!). This week’s more interesting development was deliberately less glamorous though: DARPA moved Atom Computing, Diraq, IBM and IonQ into Stage C of its Quantum Benchmarking Initiative. They join Microsoft and PsiQuantum, which had already reached the equivalent final evaluation stage.

DARPA’s question is refreshingly concrete: Can any of these approaches produce a utility-scale quantum computer whose computational value exceeds its cost by 2033 (as I said in my separate intro before publishing this very first weekly brief, my aim is NOT to give my readers investment advice. I am simply not an expert in that. But the investor in you may want to pay attention to the clear deadline introduced here. What might credible progress, or failure to make it, mean for company valuations between now and 2033? Now, that is an interesting question.)? Stage C moves the emphasis from roadmaps toward verification and validation of the hardware and systems as they are built.

That is exactly the kind of pressure the field needs. Qubit counts, fidelity records and elegant demonstrations are useful evidence, but none alone tells us whether a machine can solve economically valuable problems reliably enough, at sufficient scale and at acceptable total cost.

The deeper pattern is worth watching beyond quantum:

Claim → Execution → Measurement → Independent Verification → Trust

As technology becomes more powerful and the financial incentives to announce breakthroughs become larger, verification itself becomes valuable infrastructure (who will do the verification? Does this smell like a possible business opportunity, I wonder!).

6. Robotics Is Beginning to Build Its Own Safety Layer

A recent, smaller announcement this week may be more important than many flashy humanoid demonstrations. Agility Robotics and FORT Robotics expanded their work on a safety architecture for Digit 5, including interfaces that connect the robot to external safety systems.

That sounds less exciting than a humanoid folding laundry or performing acrobatics (and even less exciting than a video of a humanoid folding laundry while performing acrobatics!). But I say that is good. In fact, bring more of this type of news on! I learned a long time ago that real deployment is supposed to become boring. Else, the said deployment may not be nearly as reliable as one would hope.

A robot operating beside people in a warehouse or factory needs more than an impressive policy model. It needs predictable emergency behavior, communications with external (or 3^rd party) safety infrastructure, clear operational boundaries and ways for humans or other systems to intervene. The relevant scoreboard therefore should not be “How human-like did the demo look?” but something closer to useful task completion, autonomous operating hours, intervention rate, safety incidents and cost per useful hour.

Physical AI will become interesting when robots stop being primarily demonstrations of capability and become dependable pieces of infrastructure.

The Connection: Intelligence Is Moving into the Real World

Taken separately, these stories concern enterprise software, nuclear energy, memory chips, biology, quantum computing and robotics. Taken together, they point in the same direction.

We spent the first phase of the current AI boom staring at models. The next phase is forcing us to look at systems:

Agents that need permissions and verification,
Accelerators which need memory,
Data centers that need electricity,
Biological models which need experimentally generated data,
Quantum roadmaps that need independent validation,
Robots which need safety infrastructure.

This is why I increasingly find the word “AI” too small for the technological change we are trying to understand. No, I am not going to suggest Super Intelligence. I mean, give me a break!

The model may be the intelligence engine, but the economic and social consequences emerge from the larger system built around it.

What I Am Watching Next

Three questions stand out for me. First, whether enterprise agents genuinely move from answering questions to completing useful work without creating an unacceptable control and verification burden. Second, whether the AI infrastructure bottleneck keeps migrating from chips and memory toward electricity, grid capacity and financing. Third, whether AI-for-science systems begin producing experimentally validated discoveries faster (and not merely more convincing predictions).

Those are harder tests than benchmark scores or product demonstrations. They are also, I suspect, much closer to where the real value will be created.

The question for next week’s brief remains the same: What happened in the prior 7 days that meaningfully changed our understanding of where technology is going?

Sources & Further Reading

Introducing the Weekly Technology Brief

Finding the signal in an increasingly noisy technological world

My dear readers,

Sorry for the long absence. For the past few months, I have been travelling around while spending plenty of time creating various teaching materials, primarily on AI-related topics.

You know how it goes: You have the best intentions to keep up with friends and family but something, no matter how small, always happens to delay things. It is pretty annoying really, but then again, sometimes that is the best way to describe life in general so, I have learned to take it as it comes and stop moaning about it!

Anyway, one good thing coming out of my intense “content-generation” work has been the realization that, given the crazy pace of change we are witnessing these days, keeping up with technology has become strangely difficult.

Now, as a technologist, this is great news for me as I am never short of new reading materials but even for someone of my background it has gradually become a real effort to remain informed and relevant.

Every day seems to bring another breakthrough: A more capable AI model, a humanoid robot doing something remarkable, a quantum-computing milestone, a new semiconductor bottleneck (is it a signal for a new investment opportunity?), an energy project driven by data centers, or a promising biotechnology discovery. And I did not even mention the next planet (or is it still Mars?) that Elon Musk is going to populate with, well, whatever he wants to populate it with today!

Some of these developments genuinely matter. Then again, many of them don’t. And even the important ones are often reported in isolation, when their real significance only becomes apparent once we connect them to what is happening elsewhere.

Anyway, the bottom line is that:

I have spent a lot of time during the past few months generating content, which has forced me to read plenty of content, and
I realized there is too much content to go through, some of which are critical, and
I noticed that to make sense of it all it is not enough to just consume this content (even when related to critical topics) in isolation, and I found myself thinking even harder to link various topics together.
Finally, it occurred to me that at this stage of my life and career (oh, in December 2025, I finally, and happily, left corporate life to concentrate on more interesting activities, such as research, content generation, teaching, writing, advisory roles, running, travelling, and sleeping!) I actually enjoy sharing what I know with others, especially the younger folks out there who may need some advice to make them less fearful of AI takeover and more hopeful about the future.

And all these have now given birth to my Weekly Technology Brief.

But “what is the Weekly Technology Brief?”, I hear you wondering.

Each Thursday, I will look across seven areas that I believe are particularly important to understanding the technological changes ahead:

Artificial Intelligence
Robotics
Quantum Computing
Semiconductors
Energy
Space
Biotechnology

Now, let me be clear:

This will not be a comprehensive news digest. There are already plenty of places to find technology news, and I have little interest in reproducing another list of everything that happened during the week. I would rather identify four developments that genuinely changed my understanding than twenty or thirty that merely generated headlines. And boy, do we have plenty of headlines to go through every day!

No. I am not that kind of technologist!

The questions I am interested in are different. For example:

What actually changed this week?
Why does it matter?
What evidence supports the claim and what remains speculation?
How does it connect with developments in other fields?
And what might it tell us about where technology is heading next?

Looking for the Connections

Increasingly, the most interesting stories don’t fit neatly into a single category (maybe that’s because life has become too connected and, therefore, exceedingly complex?).

Artificial Intelligence, for example, may appear to begin with algorithms and models, but its rapid expansion is creating enormous demand for advanced semiconductors. Those chips require increasingly sophisticated memory and manufacturing. Data centers require electricity, cooling, land and financing. Electricity demand affects grids, nuclear power and renewable-energy investment.

Follow the chain far enough and what initially looked like an AI story becomes a semiconductor story, an energy story, an infrastructure story and eventually an economic and political story.

The same is happening elsewhere.

AI is entering laboratories and helping scientists generate hypotheses and design molecules. Robotics can then automate experiments and generate new biological data, which in turn can train better AI models.

Quantum Computing depends not merely on increasing the number of qubits (don’t worry, keep checking my blog and I will soon put in some teaching materials to explain what on earth a qubit even is!), but on error correction, algorithms, verification and ultimately demonstrating that a quantum computer can solve something economically useful.

Robotics is not simply a race to build the most impressive humanoid (mind you, with so many dumb humans in charge of world affairs, it does not take me too long to feel impressed about any humanoid out there). The more interesting question may be which combinations of intelligence, mobility, manipulation and autonomy actually produce useful machines.

These connections are where I think some of the most important signals can be found.

Evidence before Excitement

Technology also has a hype problem. Now, I believe strongly that technological hype is, arguably, as old as time itself! Apparently, generating hype (some call it marketing) is a perfectly respectable and useful activity that has long been ingrained in our economy. But (and here is a BIG but):

A company announcing that its new AI model beats a competitor is not the same as independent testing (what does “beating a competitor” even mean and based on what criteria?).
A robot performing an impressive demonstration is not necessarily autonomous (you know a human operator may be managing this remotely, right?).
A promising biological mechanism is not a successful medicine (has anyone even told us whether this medicine has actually produced a meaningful clinical benefit in patients yet?).
A quantum company raising hundreds of millions of dollars is not evidence of useful quantum computation (does the company show their ability to solve at least part of a real problem?).
And finally, an enormous investment in AI infrastructure does not guarantee an enormous economic return (although it certainly tells us something about the current level of investor sentiment!).

So, in my Weekly Technology Brief I will try to distinguish carefully between what we know, what the evidence suggests, what remains uncertain, and what is mostly speculation.

And the combination of what we know, the actual evidence and uncertainties is key here. If there is not enough meat to suggest a piece of technological news is worthy of more attention, that is what my report will emphasize. As a result, you will be exposed to conclusions such as “Interesting, but we don’t know enough yet!”. Well, I believe that is a perfectly acceptable conclusion when reviewing so much information coming at you from so many angles. At the very least, it signals that there is no point spending too much precious time on the topic unless something drastically changes.

Following the Bottlenecks

There is another question I will return to frequently in the brief:

If this technology succeeds, what becomes the bottleneck next?

The thinking is, remarkably, straightforward:

Build more powerful AI models and compute becomes scarce,
Build more accelerators and advanced memory may become scarce,
Solve memory and networking (or electricity) may become the constraint,
Build more data centers and the bottleneck may move to the grid, permits, financing or even local acceptance.
Build this and that becomes the next bottleneck!

As some of you may already know from direct experience, technology rarely advances in a straight line. Solving one constraint often exposes another, and on and on we go.

However, following those moving bottlenecks can tell us a great deal about where innovation, investment and scientific effort may move next.

An Important Point to Stress:

Remember that I just said “…may move next”. Even with all the information coming our way, we should never “guarantee” an outcome in a world full of uncertainty. Just wanted to make sure your expectation of this brief is not sky-high😊.

What This Brief Is, and Isn’t

I hope it is clear by now that the Weekly Technology Brief is not intended to predict the future with false precision. Nor is it an investment newsletter, a collection of product announcements, or a catalogue of technological wonders.

Instead, it is an attempt to build a continuously evolving picture of how several powerful technologies are developing. And, especially, how these technologies are beginning to interact.

There will also be occasional Deep Dives on certain topics, which I will share on this blog as separate pieces. When do I publish these deep dives? When a subject deserves more than a few paragraphs. The next one (which will be the first one) could be about the AI memory bottleneck, autonomous laboratories, humanoid robotics, quantum error correction, AI governance, energy constraints, or whatever else emerges as genuinely important (hint: the first one is likely about the AI memory bottleneck😊).

The objective of the brief (and the deep dives that appear from time to time) is simple:

Less news, more signal. More relevant context, more insightful questions.

Every Thursday, I will ask one overarching question:

What happened this week that meaningfully changed our understanding of where technology is going?

That will be my job.

Your job? Follow the leader, leader, leader, follow the leader…

TESCREAL+: A White Paper on Ethics of Strategic Inefficiency

Prologue: Some Useful Background

This white paper, and a few follow-up articles, mini-articles, and social media posts, that it will inevitably lead to, are my latest attempt at making sense of the efficiency-obsessed world we all live in.

As a technology hawk (I would like to think of myself as a harmless technology optimist, but who’s to say) I arrive at this point with more than two decades of work in AI (Artificial Intelligence) across various industries. My experience and studies encompass the first wave of logic-based AI, aka the good old days of intelligence based on certainly, as well as the new age AI, or what I have previously called the age of probability-infested pseudo-intelligence. And I am certainly crazy (or romantic?) enough to hang around as much as I can to see, and hopefully be part of, the next phase of AI development (why a next phase is needed is a subject of more articles in the near future).

With this background in mind, I suppose there should not be much of a surprise to learn that I have gradually found myself drawn to the Ethics of AI. This gradual shift is partly due to the fact that getting the ethics of AI right is an inherently fascinating technical challenge to overcome and party a result of my intense belief that AI, as a mirror for HI (Human Intelligence), is the humanity’s best chance to come to terms with our deepest biases, fears, and hopes.

One thing you learn when contemplating the impact of modern AI is the increasing influence of efficiency as a driving force for making any decision and for improving any algorithm. This obsession with efficiency has not only been tolerated but, in fact, aggressively encouraged, as a normal mode of operation, by most CEOs and technology investors whose definition of progress can be summed up with a few “strategically defined” KPIs (Key Performance Indicators) that tend to simplify (oversimplify is a more accurate word here) anything and everything into a set of “measurable” numbers.

While I cannot deny that this approach has resulted in tangible progress in many areas, overuse of it has certainly contributed towards the erosion of nuance, in favor of a more easily understood “average behavior”, and ultimately, the slowing down of the rate of innovation. After all, there are only a few ways one can apply a given set of techniques to “optimize” a problem and if everyone did just that, soon all solutions to the problem converge and we will be looking at a world of uninteresting uniformity where outliers are dismissed and thinking differently is abandoned on the charge of leading to suboptimal outcome.

The technological evidence of this “efficiency driven” approach is best observed in the way the new Gen AI (Generative AI) models have swept the AI landscape. Those of us who have been working with the latest Gen AI “toys” did not take long to notice the shortcomings of such solutions in their current form, from the annoying tendency for repeating the same incomplete or wrong answers to a painful “uniformity or blandness of output” across multiple Gen AI engines, which lack both creativity and tolerance for nuance. Needless to say, such shortcomings can, and often do, lead to outcomes that are ethically questionable, to say the least.

The topic of AI Ethics touches on many areas of discourse, such as philosophy, psychology, language, etc., and it is easy for anyone to lose their bearings, especially in the beginning of their journey. Therefore, I consider myself quite fortunate to have come across the seminal paper by Torres and Gebru⁽¹⁾, which helped me greatly in looking at the problems of modern AI, not as a set of simple mathematical equations to solve, but as an extension of a broader set of philosophies that drive its progress towards its nirvana of AGI (Artificial General Intelligence).

Going through this marvelous paper a few times, allowed me to connect a few dots that have been nagging at me for many years. And on further reflection, I decided to:

Take on the challenge of extending their original bundle of related philosophies, which they call TESCREAL, and introduce TESCREAL+,
Put forward an argument for relating the above extended bundle of philosophies to an “unhealthy” search for efficiency,
Propose a manifesto for reframing this particular area of AI development in order to bring much needed humanity to its progress.

In what follows, I have tried, as much as possible, to adhere to the expected format of a formal white paper throughout (while remaining accessible to a wider audience) but as always, I appreciate constructive feedback on both form and contents from everyone who shares a common interest in the topic.

Here we go folks…

Introduction

In 2023, Torres and Gebru⁽¹⁾ introduced the concept of TESCREAL, an acronym for Transhumanism, Extropianism, Singularitarianism, Cosmism, Rationalism, Effective Altruism, and Longtermism which, they argued, collectively form the ideological foundation behind contemporary AGI development. Interestingly, they traced the genealogical roots of TESCREAL to 20th century eugenics and techno-utopian worldview, highlighting how the promise of AGI is often expressed in terms of optimization and going beyond accepted human limits (an idea knows as transcendence).

In this paper we build upon that foundation but shift the analytical approach somewhat. We propose an expanded framework, called TESCREAL+, which incorporates additional ideological movements not originally addressed, namely Accelerationism, Dataism, and Computational Immortalism. While these ideologies are clearly distinct in both their origins and terminology, we argue that they do share a fundamental core idea, namely an undying commitment to efficiency as moral foundation of progress. Whether through pursuit of algorithmic optimization, technological advancements, or human transcendence, the guiding principle of TESCREAL+ is the optimization of all human systems, be it biological, cognitive, social, or ethical, towards an imagined state of perfection.

We term this phenomenon Optimization Determinism, a belief that friction, ambiguity, and redundancy are not only inefficient but morally inferior, and must be avoided at all costs. Inspired further by Holling’s systems theory⁽²⁾, as well as works on human-centered ethics by Vallor⁽³⁾ and Benjamin⁽⁴⁾, we argue how this Optimization Determinism risks repeating historical mistakes, particularly those related to eugenic, colonial, and technocratic approaches.

In response to this worldview, we introduce Constrainism, a counter-philosophy based on what we call Strategic Inefficiency. Where search for optimization encourages elimination of friction (any form of resistance), Constrainism sees constraint as an integral source of ethical insight and moral wisdom. It offers a new design ethic for AI systems, which values and welcomes ambiguity, friction, and, above all, human judgment. In other words, Constrainism treats these characteristics not as design flaws, but as features essential to a fair and resilient system.

We conclude the paper by anticipating, and addressing, a number of potential critiques of the TESCREAL+ framework and propose a better path forward for how AI is taught, researched, used, and managed. And we argue that unless we move away from treating optimization as a virtue and efficiency as a moral guide, AI will not just reflect existing inequalities, it will, in fact, make them worse, even as its biggest supporters, investors, and evangelists claim to be solving them.

TESCREAL Revisited: A Brief Refresher on the Original Ideological Bundle

As referenced by Torres and Gebru⁽¹⁾, TESCREAL stands for Transhumanism, Extropianism, Singularitarianism, Cosmism, Rationalism, Effective Altruism, and Longtermism
Each of the above philosophical branches has its own focus, but they all share a central idea, which is that advanced technology, guided by abstract ethical reasoning, can and should radically improve the human condition. Here’s a brief summary of each⁽⁵⁾:

Transhumanism promotes using technology to transcend (enhance) human abilities, be it biological, mental, and emotional, through tools such as AI, genetic engineering, or nanotechnology, and the like.
Extropianism, which is a spin on transhumanism developed in the 1980s by thinkers such as Max More, supports a belief in the merits of endless progress and overcoming physical or biological limits.
Singularitarianism, in its more recent formulation, centers on the idea that AI will eventually surpass human intelligence, triggering a singularity or the birth of superintelligence, which renders human control unnecessary.
Cosmism combines science with spiritual ideas, imagining a future where we are capable of resurrecting the dead, colonizing space, and extending consciousness forever.
Rationalism refers to a philosophical viewpoint that sees logic, Bayesian reasoning, and utility maximization as the most trustworthy ways to make decisions.
Effective Altruism (EA) tries to apply reason and data to charity, or funding actions that aim to save (or improve) the most lives per money or resource spent. Some versions of this philosophy go further, focusing on long-term risks in the context of speculative futures.
Longtermism argues that future lives matter much more than present ones. Its main objective is the reduction of existential risks to humanity, and therefore, it claims that most of our energy and resources should be dedicated to shaping humanity’s long-term survival.

As Torres and Gebru argue, what connects these ideas isn’t just futurism or the idea of abstract morality. It’s a deeper logic that resembles early 20th century eugenics, the belief that humanity should be optimized, or engineered, towards a universal idea of progress. They link TESCREAL to an earlier wave of technocratic thinking that valued purity and efficiency over diversity and complexity.

This paper builds on their work but focuses more on what we call TESCREAL’s operational core, which is the belief that optimization is a moral good, and that ambiguity, resistance, and ethical friction should be eliminated from system design. From this angle, TESCREAL is not just a collection of disparate ideas, it’s a design blueprint that increasingly ignores human complexity, unpredictability, and personal experience.

That doesn’t mean all these ideologies are the same. For instance, transhumanism values individual enhancement, while longtermism focuses on more collective outcomes. But in practice, they often converge on a shared mindset, which favors perfect abstraction over real-life messiness, speed over reflection, and simplicity over nuance.

In the next section, we’ll expand TESCREAL bundle by introducing additional ideologies that further deepen this logic of optimization, in a new framework which we call TESCREAL+.

Expanding the Bundle: Why TESCREAL+?

The original TESCREAL framework provides a powerful foundation for analyzing AGI ideologies, but it is not complete. Since its introduction, other influential movements have emerged, most notably Accelerationism, which Torres and Gebru⁽¹⁾ noted as a variant of Effective Altruism. Alongside it, Dataism and Computational Immortalism have gained popularity and gradually become mainstream, particularly in the tech industry^(a). These additions somewhat magnify TESCREAL’s core ideas, most notably the belief that optimization, disruption, and a constant focus on growth and scale are not just tools, but unquestioned virtues or moral principles.

To capture this broader bundle of related ideas, we propose the term TESCREAL+, which, in its current form, also includes the above additional ideological viewpoints: Accelerationism, Dataism, and Computational Immortalism.

Accelerationism

The origin of Accelerationism, as a philosophical movement, can be traced back to Carl Marx and his view of free trade in 1848⁽⁶⁾. While Marx never used the word “Accelerationism,” his position fits the idea. He defended free trade even though he believed it would worsen class conflict. As he put it, free trade “…pushes the antagonism of the proletariat and the bourgeoisie to the extreme point.”. However, he still supported it based on the belief that “…free trade system hastens the social revolution”. This belief that speeding up harmful systems can force needed change is a core part of what later became known as Accelerationism.

The terms itself was introduced by Benjamin Noys in his book Malign velocities: Accelerationism and capitalism⁽⁷⁾, in which he argued that embracing speed as a political strategy risks worsening the negative effects of the very systems the strategy hopes to change.

More recently, some right-wing thinkers and modern technologists have embraced Accelerationism as a strategy for speeding up social, economic, and technological change. They argue that rapid change can help society break free from existing political systems and mental limitations. And in its current AI-manifested form, Accelerationism has taken on new urgency and encourages:

Breaking things over improving them (move fast and break things),
Treating instability as innovation, and
Taking pursuit of speed as a valid strategy towards truth.

Accelerationism often discourages ethical reasoning, portrays caution as weakness, and paints a preferred version of the future system, as both “natural” and “inevitable” next step in evolution. It fits seamlessly with longtermist and singularitarian outlooks, which similarly advocate for technological progress which outpaces human reflection. Importantly, Accelerationist rationales are often found in startup culture, crypto-finance, and venture-capital discourse that actively funds AGI labs (not to mention in the boardroom of most major enterprises in the western world).

Dataism

The term Dataism was first used by David Brooks in a 2013 New York Times column⁽⁸⁾, but it was Yuval Noah Harari who gave it broader visibility in his 2016 book Homo Deus⁽⁹⁾. Dataism is the belief that the universe and human society are best understood as a series of data flows, and that the most ethical or advanced systems are those that maximize data collection and processing.

While Rationalism trusts internal reasoning, Dataism moves the focus outward to measurements and algorithms. This shifts ultimately results in treating more data as equivalent to more truth, considering algorithmic insights as integral to better decision making, and relegating subjective experience to the rank of statistical noise.

Though not often framed as a formal ideology (at least not so far), Dataism has quietly become the default worldview in many branches of the AI industry. It underlies everything from recommendation engines to predictive analytics. And in its more aggressive interpretation, it risks eliminating subjectivity as an ethical category altogether.

Computational Immortalism

Over the past two decades, the growing tendency to center human progress around technological breakthroughs has given rise to a set of worldviews that blend advanced computing with spiritual or metaphysical beliefs. While spiritualism itself is broad, ranging from religious traditions to metaphysical speculation, what concerns us here is a more specific thread, exemplified by influential thinkers such as Ray Kurzweil in his 1999 book The Age of Spiritual Machines⁽¹⁰⁾.

We use the term Computational Immortalism to describe this emerging worldview, which unites several closely linked ideas, including:

The belief that consciousness can be uploaded or preserved digitally,
The view that AGI is a stepping-stone towards godlike intelligence, and
The idea that death is an engineering problem to be solved.

These ideas, while once considered fringe, now surface in AI labs, investor circles, policy documents, and the rhetoric of technopreneurs. This belief system often borrows religious tones, for example, by referencing transcendence, eternal life, or destiny, with the slight deviation that it replaces the divine with software (god as code). In doing so, it offers moral meaning (and justification) for high-risk or speculative technological paths and postponing ethical responsibility with promises of a radically improved future, which is a pattern we describe as moral deferral.

Computational Immortalism differs from (its most closely related) TESCREAL components in key ways:

Unlike Transhumanism, which focuses on enhancing the human body or mind, this ideology seeks to transcend them entirely,
Unlike Cosmism, which imagines the resurrection of the dead and expansion across the cosmos, it centers digital continuity over physical extension, and
Unlike Longtermism, which justifies present sacrifice in service of future generations, it justifies sacrifice in pursuit of personal digital salvation.

The impact of this idea on AI discourse is already visible, be it in projects that aim to simulate our loved ones, preserve digital legacies, or train models to mimic our consciousness. As such, Computational Immortalism is not just an extension of TESCREAL; It is a critical update for understanding where AI ideology may be heading next.

A Unified Logic Driving TESCREAL+

The inclusion of Accelerationism, Dataism, and Computational Immortalism into the TESCREAL+ bundle is not merely additive, for the sake of including new variants of the same core ideas. It reveals a deeper convergence of worldviews. Across all these ideologies, we find:

A moral elevation of speed, scale, and abstract thinking above all else,
A consistent treatment of resistance to a preferred (and highly abstract) vision of progress as irrational or regressive, and
A tendency to sideline human experience in favor of universal optimization logics.

TESCREAL+ is thus more than a set of philosophies. In fact, one may call it a composite operating system for imagining the future of intelligence, morality, and governance. It informs both design and development, reinterpreting ethics itself through the lens of calculability, economic utility (i.e., maximizing outcomes or efficiency), and risk-reward calculus.

In the next section, we identify the shared foundation uniting TESCREAL and TESCREAL+ namely, the obsession with efficiency, and its ethical, political, and knowledge-related consequences.

Efficiency as Ideological Core: Optimization as Moral Principle

Across the ideologies grouped under TESCREAL and its extended formulation, TESCREAL+, one theme appears with undeniable regularity: The treatment of efficiency as a moral principle. Whether it is framed through biological enhancement, existential risk reduction, increased data throughput, or long-term planning of civilization, optimization is not just a technical objective; It is a statement about what should be done.

This belief isn’t always made explicit. For example, in Effective Altruism, it shows up in cost-effectiveness calculations that prioritize interventions based on measurable impact. In Accelerationism and Singularitarianism, it is tied to indicators of exponential growth, technological inevitability, and recursive self-improvement. Across the board, efficiency appears as a moral compass. It justifies simplification (perhaps, one should say oversimplification), allows trade-offs to pass unquestioned (as long as mathematically justified), and turns complex ethical issues into seemingly neutral equations that can be solved (interestingly enough, often with inaccurate/estimated solutions!).

But this reliance on efficiency can be dangerous. When systems are optimized without space for uncertainty, contradiction, or judgment, they risk erasing the very things that make ethical decision-making possible and necessary. And as a consequence, bypassing moral considerations can be promoted as progress.

Optimization Determinism

We term this phenomenon Optimization Determinism. This denotes the belief that technological systems should, and inevitably will, be optimized towards some ideal end state, whether cognitive, moral, or economic. This belief rests on several interrelated assumptions:

That systems can and should be made frictionless,
That calculation outperforms deliberation and reflection,
That any form of slowness signals inefficiency, and therefore should be considered as failure, and
That constraints are flaws, not features.

This worldview, while often expressed in progressive language, reduces moral complexity to solvable problems (an optimization challenge) and frames resistance or objection as ignorance or obstruction. In doing so, it fosters overconfidence about what we can truly know, dismisses diverse perspectives, and limits the range of acceptable ethical thinking.

From Eugenics to Efficiency

This logic closely mirrors the thinking behind early 20th century eugenics, which emphasized rational planning, large-scale control, and the belief that society could (and should) be engineered from the top down. The problem with this approach, as seen in past social planning efforts, was not necessarily a lack of intelligence, but too much confidence in our collective ability to simplify and control complex realities.

Today’s systems may not sterilize or segregate, as was the case with eugenics, but they still rank, filter, and exclude, using algorithms that reward things such as productivity, intelligence scores, or credit ratings. In this way, the language of optimization hides the fact that many of the same harmful patterns continue, just in new forms. In other words, despite our undoubted progress and intelligence, we keep repeating the same mistakes over and over again.

Ethical Friction and the Loss of Insight

In many AI systems inspired by TESCREAL+ logic, speed is treated as synonymous with intelligence. Human-in-the-loop processes are removed in favor of real-time prediction. Edge cases are framed as statistical noise rather than indicators of potential structural bias or system limitations.

But friction is not inherently negative. It often signals the presence of moral complexity, and edge cases (outliers) often mark the thresholds of insight and birth of innovation. Historically, many scientific revolutions have emerged from unexpected anomalies. The Michelson–Morley experiment preceded the theory of special relativity, quantum mechanics was born from deviations in blackbody radiation models. In literature and art, too, breakthroughs often emerge from disruption, ambiguity, or scrutinizing the established forms. When systems are trained to suppress the exception, they may unwittingly block the conditions of innovation itself.

The irony is quite amusing: Many of the ideologies within TESCREAL+ are promoted by the very figures who publicly champion disruption and originality. Yet the knowledge habits they promote, including over-optimization, standardization, and the elimination of uncertainty, tend towards sameness and a loss of intellectual depth. The risk, then, is not just ethical or political, but also about how we understand the world. And the path they offer is likely to lead to a future that is efficient, smooth, and utterly incapable of surprise. Indeed, death by boredom seems to be the order of the day in the future!

Optimization and the Lessons of the Past

Concern about the risks of treating all human challenges as optimization problems is not new. Many researchers and thinkers have been warning for decades that this mindset can lead to fragile and harmful systems. Their insights reminds us that the drive for efficiency, when left unchecked, often repeats mistakes of the past, be it related to earlier eugenic movement, colonial philosophies, or technocratic projects.

For example, Holling’s work⁽²⁾ on complex ecosystems shows that attempts to over-control natural environments, such as industrial forest management, often lead to collapse. Systems lose resilience whenever diversity and flexibility are stripped away in the search for increased efficiency. Vallor⁽³⁾ reminds us that true ethical progress comes from cultivating virtues, such as patience and humility, rather than focusing only on measurable outcomes. And Benjamin⁽⁴⁾ shows how algorithmic tools used in policing and credit scoring can deepen racial and social inequalities, especially when optimization replaces human judgment. Together, these lessons warn that focusing only on speed, efficiency, or outcome risks repeating the same patterns of harm observed in the past, as in earlier efforts to reshape society through control or exclusion.

Efficiency and the Delegitimization of Constraint

Constraint, in many optimization approaches, is treated as a technical limitation to be overcome, as seen in technological challenges involving compute power, latency, noise, and variability. Yet in human systems, constraint is often a moral safeguard. Constraints preserve space for reflection and resistance, allow for error correction, and prevent systems from spiraling out of control.

When these are stripped away, whether in infrastructure, education, or AI governance, what remains is a high performing system with no ethical safety locks. This leads to systems that are optimized for something, but are accountable to no one.

In the following section, we introduce Constrainism. This is a counter-philosophy that does not reject efficiency wholesale, but repositions it within a broader ethical architecture. Constrainism proposes that some inefficiencies, and indeed, some frictions, are not bugs but ethical features.

In doing so, it seeks to reinforce the value of slowness, ambiguity, redundancy, and resistance within the practice of technological design.

Strategic Inefficiency and the Case for Constrainism

If TESCREAL+ is driven by a belief in relentless optimization, then any serious challenge to its influence must address its ideas, design logic, and real-world consequences. To this end, we propose Constrainism, an alternative design ethic and philosophical stance, which treats Strategic Inefficiency as a necessary and principled response to the moral and knowledge-related failures of over-optimization.

Defining Strategic Inefficiency

Strategic inefficiency is not a rejection of performance or a call for technological stagnation. Rather, it is the deliberate embedding of friction, redundancy, and constraint within technological solutions, particularly those engaged in decision-making, classification, or resource distribution. It operates on the assumption that some forms of slowness, resistance, and ambiguity are not only tolerable but ethically indispensable.

Strategic inefficiency is evident in long-standing human systems:

In democratic governance, where checks and balances intentionally slow executive action,
In judicial systems, where appeals and adversarial reasoning delay resolution but preserve fairness, and
In education, where tests and revisions are integral to effective learning.

These systems treat constraint as an essential part of their designs, and consider inefficiency as the cost of reflection, pluralism, and safety.

What Is Constrainism?

Constrainism is the core idea behind strategic inefficiency. It offers an alternative to the accelerationist, rationalist, and utopian ideologies that dominate the current AI landscape. While TESCREAL+ envisions the future as a perfectly optimized and engineered nirvana, Constrainism argues that ethical systems must stay human-centric, use case specific, and open to interruption and reflection. Its core ideas include:

Friction is a feature: Ethics should resist the urge to flatten ambiguity into certainty,
Redundancy promotes resilience: Layered systems can help avoid catastrophic failures,
Human agency (judgment) can’t be removed or outsourced: Some decisions must remain deeply human,
Transparency matters: Some parts of a system, such as how decisions are made or what data is used, should be open and easy to verify, and
Integrity matters too: Other parts, including emotions, intuition, or personal experiences, should be protected from being turned into tools (instrumentalization) or tracked (measured) as hard cold numbers (KPIs).

Constrainism is not anti-AI. But it does reject the idea that performance scores alone determine ethical value in AI solutions. It encourages a deeper commitment about how knowledge is understood and applied. And it portrays a view of intelligence and ethics as processes which are paused, disrupted, and reflected on.

From Theory to Design: Constrainism in Practice

The principles of Constrainism can be translated into practical design constraints. These include:

Human-in-the-loop architectures, where oversight is not optional but structurally required,
Decision latency layers, where algorithmic recommendations must wait for human confirmation under conditions of uncertainty,
Plural objective functions, which allow systems to weigh competing ethical criteria rather than optimize a single scalar value,
Error-promoting simulations, where edge-case behavior is stress-tested and embraced as a learning vector, and
Schema enforcement for data systems, where logic is closely tied to use cases, errors are discouraged by design, and deviations trigger human oversight.

These are not merely technical suggestions. They reflect deeper commitments about how knowledge is understood and applied. They reflect a view of intelligence and ethics as processes interrupted and shaped by disagreement, disruption, and delay.

Sidebar: Chilán as a Constrainist Language

One instantiation of Constrainism in technical practice is the design of Chilán^(b), a graph-native functional programming language aimed at enforcing schema discipline, constraint propagation, and interpretability in large-scale, complex data environments. By leveraging functional programming’s emphasis on correctness, referential transparency, and immutability, Chilán avoids the lack of clarity and hyper-flexibility often found in contemporary machine learning pipelines and languages.

Unlike many present-day systems that tolerate or even encourage schema-less flexibility, Chilán is built on the principle that constraints and schemas are essential safeguards. They ensure clarity of purpose, enforce domain-specific logic, and reduce dependence (over-reliance) on probabilistic guesswork. By embracing well-defined structures rather than avoiding them, Chilán implements the Constrainist belief that safety and intelligence begin with intentional design boundaries.

Anticipating and Addressing Criticism

Any approach that questions dominant ideas, especially ones painted as “rational,” “effective,” or “progressive”, must be ready to face resistance, particularly when those ideas are deeply rooted, institutionalized, and hold considerable cultural influence. In this section, we look at the most likely critiques of both the expanded TESCREAL+ analysis and the Constrainist alternative, and present our responses as part of the wider debates around AI ethics, knowledge, and system design.

Bundling Too Many Distinct Ideologies

Critique: The TESCREAL+ framework combines diverse ideological projects with different goals and origins. Transhumanism is not the same as Effective Altruism, Rationalism is not Longtermism, and Computational Immortalism is not Dataism.

Response: We acknowledge that these ideologies are internally diverse. Our claim is not that they are identical, but that they share a similar structure, with a common commitment to optimization, abstraction, and technological transformation as paths to moral progress. While they differ in face value, they converge on a design logic that favors speed, scale, and KPIs. The TESCREAL+ bundle is useful not as a classification, but as a diagnostic tool for tracing shared assumptions across philosophical ideologies that shape today’s technological progress, specifically within the realm of AI (and AGI).

Attacking Something that Saves Lives

Critique: Optimization is not inherently harmful. In fact, it has saved lives, for example, through better logistics in disaster relief, more accurate diagnostics in medicine, and higher efficiency in energy systems. Why critique it so broadly?

Response: Our argument is not against optimization per se, but against its unquestioned elevation to a moral principle. We distinguish between situational optimization (bounded, contextual, and deliberate) and optimization determinism, which holds that systems should always aim to eliminate friction, redundancy, and ambiguity. Many of the harms associated with algorithmic injustice come not from optimization alone, but from applying it without constraint, reflection, or considering diverse perspectives.

Luddism or Technophobia Revisited

Critique: The proposal to embed inefficiency into AI systems sounds like a technophobic resistance to progress. Isn’t this simply a modern version of anti-technology sentiment?

Response: Constrainism is not against technology, it’s against the idea that progress has only one fixed objective, namely, some form of optimization. We don’t reject AI systems. Instead, we argue they should be built to protect, and to promote, human oversight, moral complexity, and humility about what we know, while remaining cautious in situations where we do not know! Constrainism relies on careful systems thinking and good design. What it rejects is the transformation of moral values to narrow and impersonal metrics.

Operationalizing a Vague Concept

Critique: Terms like “friction” and “strategic inefficiency” are philosophically rich but technically vague. How would this actually be implemented in code, process, or policy?

Response: Constrainism is already visible in real systems, including human-in-the-loop pipelines, adversarial training in machine learning, techniques that protect individual data privacy, democratic oversight bodies, and regulatory tools like algorithmic impact assessments (formal reviews of how automated systems might affect people and society). What we offer is not an entirely new approach, but a reconsideration of design decisions as ethical choices. In addition, tools like Chilán (discussed earlier) demonstrate that constraint-first design is both technically viable and productive in how it supports meaningful knowledge work.

AI Ethics by Another Name

Critique: Isn’t Constrainism just a repackaging of responsible AI or human-centered design principles?

Response: Constrainism builds on, but goes beyond, existing AI ethics frameworks. While many responsible AI initiatives emphasize explainability, fairness, and safety, they often treat ethics as something to be added after the system has been optimized. Constrainism challenges this logic. It puts constraint at the core of the design process, not as an afterthought, nor as a separate filter, but as the starting point. This is not ethics-as-regulation. Rather, it is ethics as the foundation of the system itself (ethics-as-architecture).

Short-Term Thinking: Why Ignoring the Past Undermines the Future

One of the most dominant characteristics of TESCREAL+ ideologies is how unevenly they treat time. Longtermism, Accelerationism, Singularitarianism, and other forms of futurism all place the distant future at the center of moral consideration. They treat the present as an obstacle (a mathematical challenge, if you will) to overcome and the past as little more than an amusing footnote. But this approach is not just ethically shaky, it is also logically flawed.

Speculation Without Foundation

Many TESCREAL+ ideologies appeal to rationalism, evidence, and formal reasoning. And yet, they frequently rely on speculative futures that cannot be tested, updated, or falsified. Projections about posthuman intelligence, intergalactic civilizations, or moral value across astronomical timeframes are often presented with confidence and utmost mathematical precision. But precision is not the same as certainty or clarity. These speculative claims are intellectually fragile as they contradict their own assumptions by operating without reasonable verification or moral accountability.

The Erasure of the Past

By prioritizing an imagined future, TESCREAL+ advocates often treat the past as irrelevant. Consequently, ethical lessons from history, deep cultural knowledge, and intergenerational memory are dismissed, oversimplified, and eventually bypassed altogether. But this obsession with the current vision of the future, while dismissing history, fails to recognize that every future eventually becomes a past in its own right. Designing for the future while erasing the legacy of the past harm and mistakes only serves to deepen injustice. A system that ignores its own historical journey cannot learn, cannot adapt, and cannot be just.

The Insignificance of the Present

In many TESCREAL+ ideologies, the present is treated as morally insignificant. A single suffering person today is seen as less valuable than trillions of hypothetical lives in the future. Cultural diversity, systemic injustice, and ecological collapse become footnotes to the primary objective of safeguarding an imagined future populated by imagined beings. But such reasoning ignores the fact that the present is the only place where tangible action is possible. Disregarding the now in favor of speculative futures leads to a collapse of morality, where human pain becomes an acceptable cost of optimization on the road to a, seemingly, better future.

Towards a More Coherent View of Time

Constrainism takes a different view. It views time as layered, cyclical, and morally complex. The past is a source of wisdom and reflection, the present is a site of responsibility and action, and the future is a domain of possibility and progress. But none can be ignored in favor of the others. Systems built on Constrainist principles honor memory, preserve ambiguity, and embrace the complexity of ethical consideration across time. In doing so, they promote neither misplaced nostalgia nor vague techno-utopian fantasies. They are not paralyzed by the past, but guided by it.

The inability of TESCREAL+ ideologies to honor both the past and the present makes them logically incoherent. Constrainism, by contrast, insists that ethical design must work across time, not outside it.

Sidebar: A Technologist’s Ethical Dilemma: Am I What I Am Critiquing?

Constrainism reflects something rather personal for me. As a technologist, I was trained to value elegance, precision, and optimization. I built systems meant to reduce complexity, uncover structure, and eliminate waste. My early career was shaped by rationalist logic, measurable progress, clean models, and formal systems. I admired thinkers who promised a better world through intelligence, and I believed that technology, if designed rigorously enough, could save us from ourselves.

But, somehow along the way, something shifted for me.

The more I worked on AI, the more I realized that the problems were not just technical. They were philosophical. The tools I used were powerful, but the assumptions behind them were questionable. I started noticing how easily ideas, which at first, felt like rational thinking turned into clever justification (in what you may call a shift from rationalism to rationalization). I saw how useful tools and approaches, such as optimization techniques, were used to mask harmful oversimplifications. The systems we were designing reflected not just our intelligence, but also our blind spots, and perhaps even our worst emotional instincts and biases.

So, I started to entertain the following question: Am I what I am critiquing?

In many ways, yes. I have certainly benefited from the very ideas I now interrogate. I still believe in the power of science and technology. I still write code and admire clarity. But I no longer believe that clarity is always good, or that efficiency is always right, or that abstraction is always harmless. I no longer believe that the best systems are the ones that disappear into uniformity. I believe in systems that interrupt, invite pause, slow us down, and, above all, leave space for judgment, contradiction, and reflection on context.

This paper, and the concept of Constrainism, is my attempt to reconcile those internal tensions. To build a future not free from friction, but informed by it. To show that constraint is not an obstacle to intelligence, but a condition, a necessary condition, for wisdom. And to remind myself, and anyone listening (reading), that the most dangerous systems are not the ones we mistrust. They are the ones we think are inevitable and beyond criticism.

Sidebar: What AI Could Be: A Constrainist Vision of Progress

Despite the sharp critiques laid out in this paper, Constrainism is not a rejection of AI. It is a refusal to build it without reflection. It is a refusal to frame AI as destiny. And it is an invitation to design it deliberately.

In the TESCREAL+ worldview, AI is often imagined as a god, a prophecy, or a governor. Something that is destined to arrive, and is assumed to be both polished and optimized. In contrast, Constrainism sees AI as a mirror: it reflects our fears, our values, and our contradictions. Its role is not to dictate the future, but to open a continuous dialogue and deepen our understanding of ourselves.

A Constrainist AI is interruptible. It is not a black box or an oracle, but a tool that slows itself down when the stakes are high. It welcomes ambiguity rather than rushing decisions based on forced clarity. It augments human thought, but never replaces it. It is accountable, and not simply autonomous. It learns from the past. It accounts for the present. And it is designed for moral friction, not moral shortcuts.

What AI could be is not the endpoint of intelligence. It could be the beginning of humility. If we build with constraint, we do not lose power. Instead, we gain clarity about when not to use it.

Conclusion: Embracing Constraint, Reclaiming Ethics

The future of AI (and AGI) is not being written purely in code. Instead, it is being guided by a number of diverse, and mostly invisible, set of ideologies, TESCREAL, and its expanded offshoots, which promote the pursuit of speed, scale, and oversimplification as the only morally correct path for the future of humanity. These ideologies do not merely guide technical advances. They tend to redefine the very essence of ethical reasoning.

In this paper, we’ve argued that what ties these ideologies together is not just a shared vision of the future. It is a much deeper belief in what we call Optimization Determinism. In response, we introduced Constrainism, as a counter-approach that sees Strategic Inefficiency not as a flaw, but as a core design and ethical value. Constrainism argues that constraint, often dismissed as technical debt or unwelcome friction, is essential for reflection, resilience, and fairness. And more than that, constraint is a way to inject much-needed human intelligence into AI. While automation removes nuance and promotes speed as an ultimate objective, well-placed constraints encourage human reasoning, ethical considerations, and contextual thinking.

In doing so, constraints do not become limits to intelligence, but act as extensions of it, highlighting that learning is not just about computation, but also conflict, balance, and context.

The implications are broad: For system designers, it means treating constraint not as a limitation, but as a source of guidance and control. For researchers, it means developing new methodologies which preserve and utilize ambiguity rather than simply eliminate it. For educators, it means resisting the temptation to reduce AI ethics to predefined checklists, secondary filters, or compliance dashboards and abstract KPIs.

It is an uncomfortable fact that we are living in a world where ethics is sold as a service, or an external collection of requirements, designed to simply enhance an already built system. Constrainism rejects this “ethics-as-a-service” worldview and instead, promotes ethics as an essential component of responsible system design and architecture.

To embrace constraint is, therefore, to bring ethics back to real life. It means accepting that uncertainties, interruptions, and different points of view are not problems to fix, but realities we must consciously design for.

The choice before us is not between progress and stagnation. It is between over-optimization without understanding and accountability and bounded systems that preserve space for reflection, discovery, objection and, ultimately, responsibility. Constrainism does not reject the future. It simply insists that we arrive there thoughtfully, and with intent.

Appendix A: Teaching AI Within Constraint: Towards a Constrainist Approach to Education

As the TESCREAL+ ideologies continue to shape not only the design of AI systems but also the training of those who build, use, and promote them, it becomes increasingly important to examine how AI education treats their underlying logics. Technical curricula often internalize optimization determinism implicitly. For example, performance metrics are emphasized, deployment speed is glorified, and ambiguity is framed as a problem to solve, not a condition to explore and factor in.

If system design is a consequence of ideology, then teaching must be considered as instrumental to design. In other words, the approach we take in classrooms, and the mindset we promote, will influence the systems we create. That’s why Constrainism must be treated as more than a design philosophy. It should also guide how we teach.

From Optimization Mindsets to Ethical Architectures

A Constrainist worldview challenges educators to change the framing of AI education, from efficiency to restraint, from blind solutionism (the belief that all problems, including complex social issues, can be solved, often with the aid of technology) to deeper exploration, and from automation by default to guided intervention.

It invites instructors to explore how ethical friction, ambiguity, and redundancy can be integrated more broadly into learning, rather than introduced as disparate or standalone “ethics modules” or “compliance lectures”.

Curriculum Design Principles

A Constrainist curriculum might include:

Ethical Friction Labs: Exercises that deliberately embed ambiguity, conflicting objectives, or incomplete information to promote careful judgment over blind pursuit of automation at all costs,
Failure as Feature: Assignments in which identifying and analyzing modes of failure are valued just as highly, or over, arriving at a pre-defined “correct” result,
Diverse Logics: Exposure to non-Western ideologies, indigenous knowledge systems, and alternative ways of managing data as legitimate inputs to AI design,
Strategic Interruptibility: Projects that enforce human-in-the-loop mechanisms and test systems’ ability to pause, reflect, or defer, as part of its “normal” behavior, and
Constraint Injection: Creative design tasks where constraints are intentionally, and randomly, introduced (e.g., no floating-point math, no cloud access, limited time horizon, etc.) to encourage creative thinking and fresh perspectives.

The Manifesto as a Teaching Tool

The Constrainist Manifesto (see Appendix B below), can serve as a useful and flexible guide offering:

A conversation starter in design ethics lectures, seminars, and the like,
A framework for debate in multidisciplinary workshops on design, and
A challenge to stimulate student reflection, inviting critiques, idea generation, or alternative manifestos.

By inviting students to take the manifesto as a starting point for discussion, instructors can foster a more active, reflective, and consequential engagement with AI design.

A Call to Educators

Teaching Constrainism is not a rejection of technology or technical excellence. it is an invitation to deepen our understanding of technology. It insists that ethical reasoning is not separate from code, architecture, or systems design, but should be embedded in them. It challenges us to produce technologists who are not only builders, but interpreters, stewards, and critics.

In a time when “ethics-as-a-service” threatens to turn morality into a one-dimensional compliance checklist, Constrainism reminds us that education must resist that logic. And it offers a simple principle that reads: To teach constraint is to teach sound judgment.

Appendix B: The Constrainist Manifesto

Constraint is not a failure of design. It is its foundation.

Constraint is not a lack of imagination. It is an invitation to imagine differently.

We live in a time when optimization is treated as morality, any form of slowness is translated as inefficiency, redundancy is dismissed as waste, and ambiguity is considered failure. But these assumptions are not purely technical or neutral. They are, in fact, ideological. They reflect a worldview in which speed, scale, and simplicity are valued above justice, resilience, and meaning.

Constrainism rejects this worldview.

We believe that:

Slowness leads to wisdom: Systems that move too fast cannot be interrupted, questioned, or redirected,
Redundancy offers protection: A single elegant solution is more fragile than a network of imperfect backups,
Ambiguity promotes insight: Uncertainty and contradictions are invitations to reflect, not errors to suppress,
Constraint brings clarity: The most reliable systems are those that define limits, expose clear boundaries, and enable deeper understanding, and
Human judgment is irreducible: No optimization function can perfectly capture moral reasoning, cultural complexity, or historical depth.

Constrainism claims that ethics must be built into system architecture, not treated as an afterthought or a separate interface. It is a commitment to design systems that protect human judgment, not eliminate it.

We do not reject technology. We reject blind solutionism.

We do not reject progress. We reject mindless acceleration without thoughtful sense of direction.

We do not reject intelligence. We reject intelligence without care or accountability.

And above all:

Good Artificial Intelligence Must Be a Mirror to Human Intelligence.

***Notes

a. To the best of our knowledge, Computational Immortalism is a term introduced for the first time in this paper. When we state that “…Computational Immortalism have gained popularity and gradually become mainstream…” we mean that many of the ideas that make up this concept are becoming more visible in the current technology-related discussions.

b. Contact the writer at: info@parvizforoughian.com for more details about Chilán programming language.

***References

Torres, É. P., & Gebru, T. (2023). The TESCREAL bundle: Eugenics and the promise of utopia through artificial general intelligence. First Monday: https://firstmonday.org/ojs/index.php/fm/article/view/13636
Holling, C. S. (2001). Understanding the Complexity of Economic, Ecological, and Social Systems. Springer Nature Link: https://link.springer.com/article/10.1007/s10021-001-0101-5
Vallor, S. (2016). Technology and the virtues: A philosophical guide to a future worth wanting. Oxford University Press: https://www.research.ed.ac.uk/en/publications/technology-and-the-virtues-a-philosophical-guide-to-a-future-wort
Benjamin, R. (2019). Race after technology: Abolitionist tools for the new Jim Code. (Polity 2019): https://www.ruhabenjamin.com/race-after-technology
For more background on these ideologies, see Torres & Gebru (2023, as per reference 1 above), and related entries on Wikipedia.
Marx. K (1848). On the Question of Free Trade. Democratic Association of Brussels, 9 January: https://cooperative-individualism.org/marx-karl_on-the-question-of-free-trade-1848.htm
Noys, B. (2014). Malign velocities: Accelerationism and capitalism. Zero Books: https://www.amazon.sg/dp/1782793003
Brooks, D (2013), The Philosophy of Data, New York Times: https://www.nytimes.com/2013/02/05/opinion/brooks-the-philosophy-of-data.html
Harari, Y N, 2016, Homo Deus: A Brief History of Tomorrow: https://tinyurl.com/mppnfjsp
Kurzweil, R. (2000). The Age of Spiritual Machines: https://tinyurl.com/4h9u5f6c

Subscribe to this blog and follow me on LinkedIn to get early access.

Related posts:

AI Ethics

The Missing Structure in a Fragmented World

Why Graphs Matter Now

We’ve built AI systems that can predict, write, and generate.
We’ve built data pipelines that move petabytes across clouds.

But ask most systems:
“How are these things connected?”
And you’ll often get silent response or a sketchy one.

That’s where graph thinking comes in, not just as a data model, but as a way of seeing the world.

Graphs let us:

Represent relationships, not just records,
Understand context, not just content,
Build reasoning systems where logic is traceable, explainable, and human-aligned.

And yet, most organizations still think in rows and columns, even as they claim to be AI-driven.

What We Will Explore

In this category, we’ll cover:

The difference between graph models and traditional databases,
Why graph-native reasoning is essential for explainable and ethical AI,
How graph thinking can unify data strategy, AI strategy, and even Quantum models (oh, I love this one and this topic alone is worth you clicking the subscribe button and following me on LinkedIn for latest developments),
Use cases in healthcare (to start with), and more.

Graphs aren’t a niche tool. They’re an opportunity to rebuild meaning, one connection at a time.

I am currently designing a new graph-native, explainable, Quantum aware programming language to called Chilán. Subscribe to this blog and follow me on LinkedIn to get early exposure to it.

Related posts:
Data Strategy as Foundation | AI Strategy | The Hidden Thread

Navigating Probability-Infested Times

The Illusion of Intelligence

We are living in probability-infested times.

Text models predict the next word. Image models fill in noise. Code models autocomplete syntax. And all of them do so without true understanding.

This doesn’t just lead to hallucinated facts; It leads to hallucinated decisions, with very real consequences.

What We Will Explore

Generative AI is the new kid in town and everyday, we hear more people claiming years and years of experience in it (as well as tons of experience in Agentic AI and god knows what else). Anyway, it is not possible to ignore the beast so, that alone means it deserves its own category. Having said that, let me be clear that, in the right hands (or as those in the hood call it, with the right prompt) it can be a wonderful tool that can massively boost our productivity.

I am actually a big fan, as you will see throughout the coming posts, but, I refuse to ignore its deficiencies or sell perceived capabilities that are simply not there yet (and some of those capabilities are unlikely to ever be there as those are tightly coupled with the initial assumptions and the overall approach). Think of it this way, if real AI (or AGI as some will call it) is compared to Relativity, then the current Generative version of AI is equivalent to Newtonian Mechanics (I love Newton, but there are problems that his wonderful laws of Mechanics are just not equipped to answer, hence the birth of Relativity and the emergence of a certain Albert Einstein).

Now, enough of this. For this introductory post, let’s say that under Generative AI category, we will cover many topics including the following:

Why relying exclusively on probabilistic logic is both risky and regressive,
What AI Agents vs. Agentic AI really means.
Silly but illuminating examples of probabilistic failure in daily life that can help in our overall understanding of Generative AI, its strengths and its limitations,
How we might use Quantum logic, structured data, and human values to rebuild trust.

This space is crowded with hype. Our aim here is to teach clearly, challenge assumptions, and reclaim agency.

The above tagline was originally written as “Our aim here is to cut through the hype and some monumental BS so that we can better understand our assumptions and see more clearly.” The end result, however, is a revision by ChatGPT:-). I think we can all agree that ChatGPT’s output is more succinct and certainly more polite so, hats off to ChatGPT here. Like I said, if you know how to use these toys, they find a way to please you.

Related posts:
Quantum Computing | AI Ethics | The Hidden Thread

A Chance to Reclaim Logic in an AI World: Rationality at Scale!

The Paradox

Ironically, the most probabilistic field in science, Quantum Physics, might help us restore structure and logic to the AI ecosystem!

Where classical AI is drifting toward black-box opacity, Quantum gives us an opportunity to:

Represent complexity with nuance,
Model relationships more organically,
Reintroduce logic-based computation into systems currently driven by surface-level patterns.

Let’s not let Quantum go the way of Generative AI: all hype, no ethics.
I am optimistic on this and believe that we still have time to learn from AI hype, and in this series, we’ll explore how to use Quantum wisely.

What We Will Explore

In this category we will explore a number of topics including:

What Quantum can and can’t do (let’s demystify this thing once and for all and not let Quantum Computing become another Blockchain that nobody understands!),
How Quantum Computing could complement Agentic AI and ethical reasoning,
Why probabilistic computation might paradoxically help counterbalance the dark side of probabilistic AI.

Related posts:
AI Strategy | AI Ethics

Not Just Oversight: It’s About Upholding Human Dignity

Beyond Bias and Audits

AI Ethics is not just about minimizing harm.
It’s about defending what matters, dignity, autonomy, and accountability, in the face of powerful and ever-expanding automation.

If our AI system optimizes engagement at the cost of truth, or automates policy decisions without oversight, then it’s not an accident that happens, but it’s a failure of design that is showing its ugly face.

What We Will Explore

Under the category of AI Ethics, we will explore all things human, as it should, given the I in AI is supposed to refer to Human Intelligence. A sample of what we cover will include:

Why AI systems must have clear, value-aligned objectives,
What human-in-the-loop really means in practice,
Why ethical oversight is not just about compliance, but responsibility.

A white paper on AI Ethics is coming soon.
Subscribe to this blog and follow me on LinkedIn to get early access.

Related posts:
AI Strategy | Data Strategy | The Hidden Thread

Why Human Values Must Anchor Every System We Build

Strategy ≠ Tools

AI strategy isn’t about which model to use. It’s about whether we are building something that reflects our values (more accurately, our human values) or just optimizing for what’s easy to measure.

Too many systems are being built with unclear assumptions, opaque outputs, and no explainability.
This isn’t just a technical flaw. It’s a failure of leadership. and it’s lack of holistic view of life in

What We Will Explore

The articles published in this category (AI Strategy), will explore a number of interesting topics including

The difference between having a model and having a strategy,
Why auditability and explainability must be systemic, not superficial,
The risks of unchecked probabilistic models, and how to build AI systems that reflect human dignity (I bet most of you did not even think of dignity in the context of AI, right? Buckle up! I have a lot to say about that in AI Strategy, AI Ethics and Rant categories!).

You’ll also see how our contents here connect to Agentic Systems, Quantum Logic, and the ability to reason at scale.

Related posts:
Data Strategy as Foundation | The Hidden Thread

Why Data Strategy Is the Foundation of AI, Quantum, and Everything Else

The Problem Nobody Talks About

Many organizations today are eager to adopt AI, deploy Generative models, or explore Quantum Computing. And yet, most still haven’t solved the basics.

Data strategy remains the weakest link.
Disparate systems, undefined ownership, unclear lineage, and some more, are not just annoying inefficiencies (and boy, are they annoying and inefficient!). They’re risks, sometimes big risks. And they prevent everything else from working responsibly.

You cannot build trustworthy, explainable, or ethical AI on a foundation of scattered, misaligned data.

If this resonates, my book Data Unplugged unpacks this in detail, no fluff, no buzzwords (I did not want to make this or any of my posts a sales pitch but, objectively speaking, this is a good book:-)).

What We Will Explore

I wrote these lines quickly as a starting point so that you get the idea what is coming in the next few weeks and after. I have a lot of stuff to go through and share with you and have to balance it between a number of categories so, if you have a specific topic in mind you cannot wait for, send me a comment. And now that we are talking about topics my readers would want me to talk about next, I promised my dear friend Willem to write something juicy on Data Strategy next so, once I have put in a few lines under each category in the blog, I will jump on Data Strategy again.

Anyway, in the Data Strategy category, I will discuss the following and then some more:

Why data strategy is not just technical architecture, but organizational architecture,
The link between governance, ethics, and business value,
Why most enterprises are still operating on data sandcastles.

Related posts:
25 Not-To-Dos of Data | The Hidden Thread