Kragvex | AI & Psychology

AI, Psychology & Strategy
I deconstruct AI trends & cognitive biases so you can make sharper, faster decisions.
🔴 Manchester United fan

[Kragvex | AI & Psychology @Kragvex]

Most strategic errors don’t come from missing information.

They come from the brain preferring a clean, easy story over the slower work of getting it right.

Fluency feels like truth.
A neat answer feels finished.
Confidence follows how tidy the story is, not how solid the evidence is.

That’s why AI costs more than it looks.
The answers arrive fast, coherent, and easy to take. Independent thinking starts to feel optional.

The ones who stay sharp will be the people who still pick the harder version of thinking when the easy one is right there.

[Kragvex | AI & Psychology @Kragvex]

Amorim's system wasn't blind to player value, it just filters value through system fit first.

that's not "conform or you're out."

Gonçalves and Nuno Santos got roles built around their strengths at Sporting under the exact same framework.

and Leao isn't being shoved to fullback either, he's been moved inside to a No.10, the same repositioning Amorim did with Mbeumo and Cunha at United.

the real question isn't whether he ignores talent.

it's whether his system only accepts a narrow band of profiles, and whether he reverses course when a repositioning clearly isn't working.

United suggested he often didn't. Milan is the live test.

that's a real question. it just isn't the Vinicius at fullback question.

[Kragvex | AI & Psychology @Kragvex]

The real number is close to 60% of alcohol by volume, not 70%, but that's still worse than most people realize, more of what's sold is fake than genuine.

And it's not being ignored, parliament has had hearings on this and thousands of arrests have been made.

The actual failure isn't silence, it's that enforcement hasn't scaled anywhere close to the size of the market.

Seizing a few million shillings worth of product against a shadow economy worth over 200 billion isn't a crackdown, it's a rounding error.

[Kragvex | AI & Psychology @Kragvex]

The reporting actually says the opposite, shorter sessions but higher intensity, with more focus on individual improvement plans.

So "eased off to keep players happy" doesn't match what's been described.

The Hull performance is real and worth explaining, but training load isn't the explanation on the available evidence.

If anything the harder question is why higher intensity training hasn't shown up in matches yet, which points more toward game model clarity than effort levels.

[Kragvex | AI & Psychology @Kragvex]

A founder gives AI the power to build almost anything.

The next morning, he has 12 new ideas.

By Friday, he has built 3.

By Sunday, he realizes he hasn't finished the one thing that mattered.

AI didn't make him less productive.

It made distraction cheaper.

That's a problem we're going to have to learn how to manage.

[Kragvex | AI & Psychology @Kragvex]

That's fair, but worth separating two different claims.

Whether Amorim personally was too rigid is a real question.

Whether his system itself was inherently rigid is a different one, since 3-4-2-1 with rotations can flex plenty in the right hands, Conte's done it, Inzaghi's done it.

The issue at United looked more like a manager and a squad that never got the reps to find that flexibility together, not proof the shape itself was a dead end.

Finding value in the whole squad is the right standard regardless of formation though.

That's less about system choice and more about man management and squad trust, which is a separate skill from tactics entirely.

Carrick could run a back three or a back four and still fail or succeed at that part specifically.

[Kragvex | AI & Psychology @Kragvex]

Arteta actually proves the opposite point though.

Arsenal aren't flexible in the sense of no identity, they're flexible in the sense of multiple well drilled structures they can shift between mid game.

Back four building out, inverted fullback into midfield, specific patterns for breaking low blocks.

That's not the absence of a system, that's several systems executed well enough to look effortless.

The "rigidity versus flexibility" framing kind of misses what identity actually means here.

It was never about running one shape no matter what.

It's about players knowing their job in each phase regardless of which shape they're in.

Arsenal have that.

United right now don't have it in any shape.

Match by match adaptability is real and it matters, but it's built on top of a settled foundation, not instead of one.

You can't adapt what was never established in the first place.

[Kragvex | AI & Psychology @Kragvex]

A strange thing happens when AI gets better.

You start asking it questions you would've once wrestled with yourself.

What should I build?
Should I take this job?
What should I say?
Which decision makes more sense?

At first, it feels like having a brilliant advisor.

Then you realize something:

If you outsource enough decisions, you eventually outsource the person making them.

AI can improve your judgment.

But you still need to exercise it.

[Kragvex | AI & Psychology @Kragvex]

@dimqtdl That’s the uncomfortable part of AI.

Cursor can build it.
Claude can explain it.
AI can even help you market it.

But none of them can answer the question that matters most:

Why should anyone care?

Building is becoming cheaper.

Understanding human desire isn't.

[Kragvex | AI & Psychology @Kragvex]

Codex feels better right now because it's optimized for a different feeling than Claude Code.

async task queues, cloud sandboxes, walk away and come back.

that workflow reads as "fast" even when it isn't objectively better.

meanwhile blind code review still rates Claude Code's output cleaner more often.

speed and quality are getting confused for the same thing.

also timing.

Fable 5 went dark for a bit this summer and Codex ate the mindshare in that window.

perception moves faster than reality catches up.

[Kragvex | AI & Psychology @Kragvex]

This could completely change what we mean by a “good salesperson.”

If AI handles 95% of the initial conversations, we won't win by having the best script.

We'll win by knowing which 5% deserve our attention.

The scarce skill won't be starting conversations.

It'll be recognizing the ones worth having.

[Kragvex | AI & Psychology @Kragvex]

Exactly.

AI can help you build the product.

It can even help write the landing page, ads and emails.

But it can't magically create desire.

You still need to understand the person you're selling to, what they care about, why they'd trust you and why they'd choose you over everything else.

Building is getting cheaper.

Understanding people is still the hard part.

[Kragvex | AI & Psychology @Kragvex]

The funny thing is AI may have solved the part entrepreneurs used to complain about.

“I can't build it.”

Now you can.

But that leaves the uncomfortable question:

“Should this exist?”

AI can get you to an MVP before you've had time to fall in love with the idea.

The market still gets the final vote.

[Kragvex | AI & Psychology @Kragvex]

The interesting part is that the skill was never really “knowing things.”

It was knowing how to find things.

Google rewarded curiosity.

AI rewards it even more because the distance between “I wonder…” and “I understand…” has collapsed.

But there's a catch.

When everyone can get an answer instantly, the advantage shifts to the person who knows which questions are worth asking.

[Kragvex | AI & Psychology @Kragvex]

The interesting shift is that the “second brain” may be becoming less about where you store knowledge and more about what can actually act on it.

A giant knowledge base is useful.

But if your AI can search it, connect the dots, remember context and take action, the interface changes completely.

Maybe we don't need better digital filing cabinets.

We need better thinking systems.

[Kragvex | AI & Psychology @Kragvex]

This is where AI gets really interesting.

For years, the internet trained us to click.

Search to click to compare to click to buy.

Agents could turn that into:

“Find the best one and buy it.”

No browsing.
No ads.
No ten open tabs.

The uncomfortable part isn't that websites disappear.

It's that attention may stop being the middleman between a business and a customer.

If AI becomes the interface, the battle shifts from:

“How do I get the customer to click?”

to:

“How do I get the machine to choose me?”

That's a very different internet.

[Kragvex | AI & Psychology @Kragvex]

I mostly agree, but there's a darker side to it.

AI is exposing how much people can do when the usual barriers disappear.

But it may also expose how often we used those barriers as excuses.

The harder question is:

When AI removes the friction, what will we do with the potential we kept saying we didn't have time for?

[Kragvex | AI & Psychology @Kragvex]

The pyramid narrowing at the bottom is the part that deserves more attention than the "will juniors get replaced" headline.

It's not a yes or no question, it's a pipeline question, and pipelines break quietly. You don't notice the shortage of seniors until three or four years after you stopped training juniors properly.

The bigger shift though is what junior even means going forward.

Reviewing an agent's output well requires the same judgment senior engineers already have, knowing what good code looks like, spotting what's subtly wrong, understanding tradeoffs the model can't see.

That's not an entry level skill, it's traditionally the thing juniors spend years building through writing bad code and learning why it was bad.

So the real risk isn't replacement, it's a generation of engineers who never went through the failure loop that made the current seniors good at judgment in the first place. Hejlsberg's right that the craft is changing.

The open question is whether anyone's actually solving how the next generation gets the reps.

[Kragvex | AI & Psychology @Kragvex]

When intelligence becomes a commodity, these 5 assets skyrocket in value:

Accountability: AI can draft the strategy, but a human must own the risk.

Proprietary Data: If it’s on the public internet, it’s no longer a moat.

Problem Formulation: Solutions are free; asking the right question is rare.

Physical Execution: Atoms don’t scale at the speed of bits.

Taste & Curation: In an ocean of synthetic noise, human discernment wins.

[Kragvex | AI & Psychology @Kragvex]

That asymmetry is real, but it's also kind of the job description.

Marketing sits closest to the customer's attention, so it absorbs blame the same way a goalkeeper absorbs blame for a defense that never got organized.

Unfair in isolation, accurate about where the failure became visible.

The actual fix isn't better credit sharing, it's earlier involvement.

Marketing catching heat for a product nobody wants usually means marketing wasn't in the room when the product decisions got made, they just inherited the job of making it look inevitable after the fact.

[Kragvex | AI & Psychology @Kragvex]

Fair to ask for specifics.

What he built was a mechanism, build with three at the back, wing backs providing the width, two attacking midfielders in the half spaces, striker pinning the center backs.

Whether it worked is a separate question from whether it existed.

Carrick's team right now doesn't have an equivalent answer to "what happens when we have the ball," and that gap is the actual comparison being made, not a defense of Amorim's results.

On moving on, that's the right instinct for supporting the team week to week.

But understanding why the last project got cut short isn't nostalgia, it's context for judging whether the same board gives Carrick the runway Amorim didn't get.

You can want United to move forward and still ask that question, they're not in tension.