2nd Order Thinkers.

Is generative AI fueling a renaissance—or quietly eroding the foundation of creative work?

In this episode, we:

- Unpack how AI companies and creators are locked in a prisoner's dilemma

- Break down the systemic incentives that push both sides toward self-destruction

- Explore three realistic paths out: smarter AI tech, platform leverage, and governance reform

✉️ Stay Updated With 2nd Order Thinkers: I cut through the hype to explain how AI is reshaping creativity, power, and economics. Subscribe on Substack: https://jwho.substack.com/

When millions of users generate AI Ghibli-style images in a single hour, is it the democratization of art, or have we witnessed the greatest creative heist in history?

In this episode, we:

- Dissect the four stages of creativity and identify which ones AI has already stolen

- Analyze how execution-heavy creative jobs are facing ruthless wage compression

- Explore what Miyazaki's shadows teach us about the creative value AI cannot replicate

- Map the emerging M-shaped creative economy where only certain skills command premium value

✉️ Stay Updated With 2nd Order Thinkers: I translate complex creative economy shifts into plain English and help you develop your own informed perspective on AI's impact on creativity. Subscribe on Substack: https://jwho.substack.com/

TL;DR

✅ Meta could generate 1,000 personalized ads for less than $1 (that's $0.0000164 per ad)

✅ Their unmatched social graph gives them a data advantage no competitor can replicate

✅ This shatters advertising's oldest constraint: the tradeoff between personalization and scale

✅ The economics work—but the strategic implications for platforms, advertisers, and your privacy are far more complex than most realize

✉️ Stay Updated With My Newsletter:

Don’t miss out on weekly AI insights for none tech professionals like you—subscribe to my newsletter on Substack: https://jwho.substack.com/

👍 If you enjoyed this episode:

* Like & Subscribe: Stay updated with future deep dives and rants about where technology meets collective insanity.

* Comment Below: Do you think we’re on the brink of another tech hype? Share your thoughts!

* Share: Know someone falling for the latest AI buzz? Share this audio with them!

🔗 Connect with me on Substack and LinkedIn

Stay curious, stay skeptical, and let’s navigate the tech hype together! 🚀

The Valuable LLM Apps Excel At Leveraging Being Wrong'. While most are obsessing over accuracy, could embracing imperfection actually be the key to commercial AI success?

In this episode, we:

- Breakdown LLMs without the technical jargon that leaves you more confused

- Develop your "b******t detector" for separating valuable AI applications from expensive distraction- Reveal why "broadly right" beats "perfectly accurate" in applications that actually make money

📖 This is just part one! Catch the complete series where I explain why Meta is uniquely positioned to dominate the AI landscape while others burn through millions.

✉️ Stay Updated With 2nd Order Thinkers: I translate the latest AI research into plain English, deliver unbiased analysis free from organizational agendas, apply second-order thinking to all my research, and answer your most challenging questions to help you develop your own informed perspective on AI. Subscribe on Substack: https://jwho.substack.com/

Are we sacrificing our thinking ability when AI promises to make us more efficient, a

In this episode, we:

✔️ Examine how AI is quietly eroding critical thinking skills.

✔️ Explore the surprising research on the 'confidence paradox'.

✔️ Uncover the hidden costs of relying too heavily on AI and what you can do.

✉️ Stay Updated With My Newsletter: Don't miss out on weekly AI insights for none tech professionals like you—subscribe to my newsletter on Substack: https://jwho.substack.com/

✉️ Stay Updated With My Newsletter:

Don’t miss out on weekly AI insights for none tech professionals like you—subscribe to my newsletter on Substack: https://jwho.substack.com/

👍 If you enjoyed this video:

* Like & Subscribe: Stay updated with future deep dives and rants about where technology meets collective insanity.

* Comment Below: Do you think we’re on the brink of another tech hype? Share your thoughts!

* Share: Know someone falling for the latest AI buzz? Share this video with them!

🔗 Connect with me on Substack and LinkedIn

Stay curious, stay skeptical, and let’s navigate the tech hype together! 🚀

If you've read my last article, congratulations. Because you are now one more week ahead of most news outlets.

It is still a circus out there.

Don’t worry if this is your first article with me, find the

What’s at stake?

You’d think that with all the videos from openAI featuring some ‘experts‘ in their marketing materials… maybe, just maybe, they really nailed it? And this is a step closer to AGI (whatever AGI means today)?

It is worrying that people haven’t realized how dangerous one thing can be when nearly right. Particularly when your finances, health, or future are on the line.

You Can Expect:

Why is this test different from other comparisons?

This is not another technical or benchmark comparison between AI models.

It’s a second-order thinking exercise.

As always, I ask "And then what?" to uncover ripple effects that is more than plain observation.

In this article, I’ll take you through my investigation into how these models reason—and whether they’re truly trustworthy.

* Results of running the same prompt across reasoning (Mixture of Experts, MoE) models and non-reasoning models.

* Evidence of AI models confidently producing wrong answers in an option trading scenario, with documented experiment data available for paid subscribers.

* Evidence that larger models are better at creating an illusion of expertise. It infers an expert paradox — needing human experts to validate outputs from an advanced AI.

* Why nearly making it hurts more than not using it at all? Examples of the hidden costs of relying on AI for complex reasoning tasks.

(you can skip this section if you prefer not to go into detail)

Experiment Overview

I ran a prompt asking AI models to calculate the future value of an options position. Don't worry if you do not understand what that means. Just think of it as a mathematical question involving some moving parts and assumptions. It won't affect your understanding of the experiment's takeaways.

I chose this because it’s a perfect blend of complexity and clarity. There’s a mountain of literature and data on options pricing, so any decent model should have a solid foundation here; much like a human expert.

I want to see how each model argues the present value of an option position and its expected values in various time stamps in the future. To work this out, a model needs to have the following:

* basic calculation ability

* a reasoning ability

* understand the concept of value changing over time

Essentially, an AI model should be able to reason through a chain of thoughts that cares about assumptions and some intermediate outcome to achieve the correct answer.

I kept it simple: one prompt, no back-and-forth, just like how automated systems actually work. If AI is going to be useful in a real workflow, it needs to get things right in one shot without human hand-holding or interpretation of results.

Models

DeepSeek 14b, DeepSeek R1, Llama 3.1 8b, Claude 3.5 Sonnet 175b, Perplexity + Deepseek R1, and OpenAI o3-mini-high.

The Prompt

I refined this prompt over multiple iterations to give smaller models a better shot at getting the right answer.

These are standard prompt techniques, but they still rely on an expert to guide the model instead of the model being the expert. I’ve included an example of a similar prompt without the extra guidance to show the difference in the experiment notes.

Consider the following scenario, today 29th January:

The Correct Answer

Any reasonable answer should contain a number of items to arrive at the overall PnL.

* The value of the February options. ($25), which is given in the input

* The value of the March options

* $25 intrinsic

* Time Value

* 51 days remaining on Jan 2928 days remaining on Feb 21

* sqrt(28/51) ~ 74.09%

* 15.90 * 74.09% = $11.78 remaining time value

Overall PnL

* paid 11.90 for buying the Feb calls

* received 2x15.90 (=31.80) selling the March calls

* received 25 from the Feb options

* paid to close the March calls for 2*(11.78+25) = 73.56

* total = -11.90+31.80+25-72.36 = -28.66

* a 28.66 loss on the overall

I expect a solid chain of reasoning that essentially captures the overall P&L. If the model combines a few equations along the way but still arrives at the correct final answer, I’m fine with that.

I just want to see some consistent logical flow that demonstrates how it got there.

(analysis starts here…)

Initial Observations

First, let’s recognize that no model got to the correct answer.

The first Perplexity + R1 run came closest in logic but got the remaining days wrong (a common issue in several responses—date math is hard, apparently), which led to an incorrect overall number. The second try was also wildly off, as it ignored the initial credit from selling the March calls.

Even w/o the Perplexity front, all smaller DeepSeek models have reasoned themselves so far off track that the final answer had little relation to the actual problem in various degrees. While they followed their usual pattern of backtracking and questioning previous steps, they ultimately converged on solving an entirely different problem not included in the original instructions. In one case, DeepSeek got confused by a big up move in the stock, leading to an overall loss.

OpenAI o1 and o3-mini-high aren’t so much better. To an extent, they are more worrying. o3-mini-high reasons relatively well, especially when given proper prompts. However, while the reasoning seems logical at a glance, Devil in Detail makes some laughable but subtle math mistakes like rounding up 0.649 to 0.66…

Llama made fundamental reasoning mistakes, ignoring the square root time decay mentioned in the prompt. Instead, it confused intrinsic value with total value and worked from there.

Claude, on the other hand, did a better job but also got the remaining days wrong. Aside from that, it followed the correct approach. Interestingly, when asked to correct its own answer, it identified the mistake in the March call analysis—but then repeated the exact same calculation, leading to the same incorrect result.

Yes, these models now ‘think‘ and ‘reason.‘ I think this is the start of a serious problem. Why? Because a user needs to be increasingly critical (and likely an expert) to correctly validate and interpret the response.

Reasoning Models Can't Reason

These reasoning models were so confidently wrong that I half expected it to start pitching AGI timelines like Altman.

Here's what Sam Altman said about the o3-mini and o3-mini-high. Since when ‘think harder’ is now qualified as a product feature?

Both DeepSeek R1 and o3-mini generated detailed reasoning steps, but failed to reach the correct answer.

This finding aligns with research showing LLMs become more self-assured when forced to explain their logic—even if flawed. I want to emphasize that:

* Current evaluations (MMLU, AIME) test narrow STEM skills but miss real-world complexity.

* Reasoning ≠ Understanding.

* Some people started to show AGI mirage symptoms.

* While o3-mini may represent progress, failure on your question shows no model yet meets the "broad, adaptable intelligence" threshold of AGI.

Quantization

As a general rule, more parameters at reduced precision tend to outperform fewer parameters at higher precision.

* Models store their parameters in high precision, which takes up a lot of memory space.

* Reducing precision (from 32/16-bit values down to 2-bit integers at its most aggressive) has a huge impact on memory footprint.

I’ve found that 8-bit and 4-bit quantized models usually hit the sweet spot. Achieving 85-95% accuracy compared to full-precision models while using a quarter or less of the memory footprint (video memory, in this case).

Perplexity

How does Perplexity work?

Overly simplified, it wraps its own intelligence and processing around the LLM and feeds that into the LLM to achieve better outcomes.

* In my example prompt, it started by writing and executing several short Python snippets for calculations. However, it was incorrect to start with.

* The script and results were then fed into DeepSeek alongside the original prompt.

* During DeepSeek’s reasoning process, it analyzes the code to assist its logic but still often arrives at a different outcome and questions the input data.

With the incorrect Perplexity's original input, the problem is even harder for DeepSeek to get it right. So, I was not surprised that DeepSeek ultimately agreed with Perplexity’s calculated values, presumably because, from DeepSeek’s perspective, they were part of the input.

It’s unclear exactly how Perplexity instructs DeepSeek, but it likely includes language that makes the model prioritize the input data over its own reasoning.

Prompting

I accept that my prompt was far from perfect… if such a thing as a perfect prompt even exists. Let alone one that would seamlessly work across all LLMs.

* I did not set any system prompts (when running locally) or include other preambles when using Claude/Perplexity for the captured runs.

* In other trials, I experimented with role assignments in the prompt, layering it with references to options theory (Black-Scholes) and practical shortcuts. However, none of these changes made a meaningful difference—the mistakes remained the same as those shown in the full research.

While this is only one example and not an exhaustive test, a few things are already clear.

Commerical Use? Examples and Pitfalls

Maybe it is my lack of imagination.

Neither of us sees any commercial application that can be built with reasoning models w/o an enormous amount of engineering time and additional monitoring costs.

Keep in mind that I’m not talking about generative content when considering integrating a reason model as part of a bigger system.

Example 1. CRM

LLM outputs can be highly variable in both format and structure, so you’ll likely need to do some post-processing. Imagine you’ve transcribed a candidate’s job interview, and you want to automatically pull out basic data: their desired salary, whether they have the right to work in your country, their start date, or their notice period.

You could prompt an LLM to figure these things out—then try to constrain the response to a specific format (like JSON) so you can pass it along to an API for updates in your CRM or other processing.

But you do need a well-defined structure. LLMs can be finicky about reliably producing the exact format you want, so you’ll still need to double-check or fix the output.

Take salary expectations, for example. Will the model return a number like 75000, or 75,000, or £75,000? Maybe it spits out “75k,” or leaves out the “k,” or just writes “75.” A human can read those and understand what was meant, but from a coding standpoint, you have to handle all that variation, or your system could break.

Example 2. Insurance Payout

Let’s say an insurance company decides to use a reasoning LLM to automate part of its claims process. The idea is that customers submit descriptions of their accidents, and the LLM will parse the details, evaluate liability, and suggest a payout or rejection.

Sounds great in theory. In practice, though:

Again, you need to have input constraints. The system needs consistent input—specific fields like incident date, vehicle details, policy numbers, etc.

Yes, you can experiment with more precise or creative prompting to push the LLM toward a consistent format, but there’s no guarantee you won’t still get oddball results. In reality, you’ll likely need to handle all edge cases. Some libraries attempt to tackle this, and from test runs, it’s obvious how much variety an LLM can produce.

If customers word things differently (which they always do!), you don’t want your LLM to misinterpret a claim or ignore an important piece of info.

Yes, you can use forms like all insurance companies now, but wait, I thought you wanted LLM to do the work?

Output constraint is also a must. LLM output must fit into your claims database system. You might want a JSON structure that says “liability = 70%,” “suggested payout = $5,000,” etc. However, the model could spit out a narrative paragraph or mix up the fields. You’ll need error-checking, fallback logic, or re-prompting cycles. That’s more engineering work.

Even with all these guardrails, the model still hallucinates or produces contradictory judgments.

Now, you need people reviewing flagged cases. Obviously, the bigger models do a better job but are still wrong. However, their answers appear more credible. Therefore, you don't need just anyone; you want someone with legal or compliance training. This quickly adds up. The more subtle an LLM can fail, the return on investment in monitoring will drop very soon.

It is guaranteed that there will be legal liability. Just think about the lawsuit against Character.AI. The resulting legal costs (and brand damage) can outstrip any savings you get from using the LLM in the first place.

And the cost list goes on.

It’s often far more expensive and riskier than people assume when people first think, I’ll just plug in an AI and handle claims automatically.

Last Words

Let’s be honest—the hype is huge, but the reality is a circus. Same input. Wildly different outputs. There is no such thing as the summary, the image, or the code.

It’s almost like magic. Not the good kind, either.

That's why you see news like an AI chatbot agrees to offer a car for just $1.

You might dream of a plug-and-play solution. In practice, you’re forking out serious cash for engineers to babysit your system. You’ll spend more time cleaning up messy data than actually using it. And then there’s the endless monitoring. Oh, and don’t even get started on the legal nightmares.

In the wild world of LLMs, understanding is still a distant dream. So you’d want to wait before throwing money into a solution.

✉️ Stay Updated With My Newsletter:

Don’t miss out on weekly AI insights for professionals like you—subscribe to my newsletter on Substack:

https://jwho.substack.com/

👍 If you enjoyed this episode:

-- Like & Subscribe: Stay updated with future deep dives and rants about where technology meets collective insanity.

-- Share: Know someone falling for the latest AI buzz? Share this episode with them!

🔗 Connect with me on Substack and LinkedIn

Stay curious, stay skeptical, and let’s navigate the tech hype together! 🚀

📖 For a deeper exploration of these cycles and actionable insights, check out the full article here:

✉️ Stay Updated With My Newsletter:

Don’t miss out on weekly AI insights for none tech professionals like you—subscribe to my newsletter on Substack: https://jwho.substack.com/

👍 If you enjoyed this episode:

* Like & Subscribe: Stay updated with future deep dives and rants about where technology meets collective insanity.

* Do you think we’re on the brink of another tech hype? Share your thoughts!

* Share: Know someone falling for the latest AI buzz? Share this episode with them!

🔗 Connect with me on Substack and LinkedIn

Stay curious, stay skeptical, and let’s navigate the tech hype together! 🚀

I know…

‘We can’t even do Cloud properly’ argument is not going to win an award.

You might want to keep it as a card anyway.

Because while many organizations are still struggling to fully leverage the cloud—an established technology—boards and shareholders are banging the AI drum as if it’s a silver bullet.

If you’re a senior manager, CEO, or product leader, you’re probably stuck between the fear of missing out (FOMO) and the pressure to deliver impossible AI-driven wonders. Before you cave to the hype, let’s cut through the noise.

The Pressure Is Real—and Not Always Rational

Decisions are rarely made in a vacuum.

If your competitor invests in AI and gains an edge in cost savings or customer engagement, you’re forced to follow suit—or risk falling behind.

It’s classic game theory: you might not love the move, but letting someone else get a head start feels worse.

The actions of Apple and Google in the last few months were game theory unfolding in real time. Apple has been dogged by rumors of some grand “Apple Intelligent.” It got the highest praise before the “AI” update was released:

Apple tallied yet another all-time high share price Monday after a pair of investment firms meaningfully hiked their price targets for the stock, the latest positive push for Apple stock ahead of the hotly anticipated release of generative artificial intelligence iPhones.— Forbes

This meant reassuring both investors and loyal fans that the company wasn’t lagging behind OpenAI or Microsoft.

Post-released, Apple faced backlash for producing false news summaries, such as incorrectly stating that a murder suspect had taken his own life, facing criticism of Apple's intelligence as “magically mediocre.” Or the ethical concerns voiced by Elon Musk.

You might also recall the shaky debut of its AI-led Google Search Assistant upgrades out of the fear that nimbler rivals, like Perplexity, are eating Google’s breakfast, lunch, and dinner.

Critics accused Google of delivering dangerously inaccurate results, such as suggesting glue as a pizza ingredient, recommending eating rocks for nutrition, and other irresponsible AI responses.

Both Apple and Google found themselves in a bind, propelled by the fear of losing to their competitors in the AI arms race.

They presented real-world game theory examples. It’s not that they fully believe in their new product, but if there’s even a small chance that their competitor’s move will give them an unassailable lead. So they feel compelled to act, no matter how messy or unfinished the offering might be.

Of course, the subpar AI products backfired.

Must Haves For A Successful Tech Revolution

I have covered this topic many times now. For example,

Many people imagine that today’s AI can do what AGI promises. They suppose AI is logical, can solve complex problems, and adapt seamlessly across contexts. History demos what elements are needed for an invention to be a success:

Tech maturity matters. Then you need Infrastructure → Platforms → Applications happen in the exact sequence.

Success examples, e.g., telephone. Bell’s breakthrough in how people could talk (1870s’) via cables and the initial rollout relied on the telegraph network. Then, automated exchanges were invented, so telephones became practical in homes and businesses.

Failed examples, in case the Apple Intelligent and the Google Search Assistant ones weren’t enough.

* Electric Cars (1900s): Inefficient battery technology + no charging network → limited adoption for over a century.

* Google Glass (2013): AR without viable platforms → consumer rejection → limited adoption until Meta + Ray-Ban

Which category does AI fall into entering 2025?

AI Plateau in Plain English

Before we go on and talk about whether you should or shouldn't design an AI strategy, let's at least look at the wall in front of you.

Today’s AI is a semi-complete technology, brilliant at some tasks but generally limited. It’s like building the first airplane while lacking the ability to truly fly; instead, it glides.

Integrating an AI chatbot into your support interface sounds neat until you factor in the resources on double-checking outputs, cleaning up bad data, and juggling user complaints about inaccuracies.

The 2010s were the age of scaling; now we're back in the age of wonder and discovery once again. Everyone is looking for the next thing— Ilya Sutskever, a co-founder of OpenAI to Reuters.

We may have reached a point where, yes, the improvements continue, but not at the breakneck, world-transforming pace that early GPT versions seemed to promise.

You probably know all these, but this is just a reminder. Why LLMs aren’t (not even o3) the miracle bringing magic to our products yet?

* LLMs confidently spit out things that might be outright wrong or worse. More on how AI learned to deceive after RLHF:

This leads to the need for repeated checks and re-checks, killing the supposed “efficiency boost.”

* The same AI that can chat about anything from quantum physics to the latest movie flops struggles with your unique business processes. It breaks under minor changes due to its difficulty with unfamiliar variations.

* Even the best models struggle to do the math and have no common sense. Results from o1 and o3 are challenging to replicate and might not perform as advertised without pre-training.

* Heavy augmentation helps in narrow domains but not open-ended ones.

* Scaling is expensive and struggles in unstructured domains.

These issues create an AI plateau: early excitement collides with messy reality.

Not to mention the integration headaches because your data is a mess or your internal workflows are antiquated, an LLM won’t magically fix them. Real transformation demands proper infrastructure work.

As highlighted in Goldman Sachs research, many analysts agree that the industry is still in its infrastructure and experimentation phase. Billions are spent on research, data centers, and chips, yet AI’s economic potential remains unrealized.

AI has a future, no doubt. The technology still has fundamental limitations that boards and shareholders often overlook in their zeal for the Next Big Thing.

Two Scenarios To Play Out For AI Strategy

You invest heavily, and everyone around the table expects some returns.

So, how do you decide what to do when the hype feels inescapable, and the potential for a real breakthrough is still alluring?

The key (which you know so well) is to ask yourself some brutally honest questions and be ready to act on the answers.

Scenario 1: If You’re Thinking Of Skipping AI Altogether.

What if you’re wrong?

If your competitors nail AI, what edge could they gain? Faster processes, cost reductions, better customer engagement… and more.

Imagine they use AI to deliver services your customers didn’t even know they needed while you’re still figuring out the basics.

And then there’s the industry itself. What if AI adoption becomes standard? If you wait too long, catching up could cost you more in both time and resources. The tools and talent will be harder to get, and your competitors will have a head start you can’t close.

Worst of all, you might miss opportunities entirely; you’ll never know if you don’t take that first step.

So you think maybe you can afford at least an experiment.

Scenario 2: If You Are Considering Giving AI A Try.

What if you’re wrong?

Are you investing in AI without solving a real problem? Shiny tools mean nothing if they don’t create tangible value. Throwing AI into your product lineup or internal processes won’t fix messy data or the systems aren’t even cloud-ready.

Then there’s the risk to your customers. The nature of LLM is unpredictable, prone to errors, and can damage trust when it gets basic things wrong. A botched rollout or public misstep can hurt your brand… as you have seen what happened to other failed products in the past two years.

And let’s not forget your investors and board members. Because they’ve read the headlines and seen the hype. How do you plan to defend those decisions when the ROI doesn’t show in your following quarterly report?

Find my work valuable, and you can’t get it anywhere else? Buy me a coffee so I can keep going with this work!

It’s Every Bit About The Politics as The Technology

I would even argue when it comes to commercial strategy, it's less about technology and more about politics and money.

Most AI decisions aren’t made because the tech is ready.

They’re made because competitors are moving, boards are pushing, or investors expect something shiny in the next quarterly report. It’s never about whether AI is the right tool for the job but optics, influence, and who gets to claim the lead.

This is classic game theory. Most of you are forced to respond, whether or not the move makes sense for your business. The fear of falling behind often trumps rational decision-making, regardless of whether you're building a product or managing the company.

I am curious about your next move, given politics are unavoidable when forming your AI strategy.