Credibility is the new diploma

Credibility is the new diploma

---

When I say AI in this post, I mean the current wave: LLMs, agents, coding assistants.

Learning has never been this fast. University took years to hand you knowledge, the internet sped it up, and now, with an infinite tutor in your pocket, you learn anything at a speed that didn’t exist before.

So speed stopped being the problem. And right there the new one shows up: if anyone learns fast, how do you prove you actually know? How do you earn credibility for what you learned?

I’ve been chewing on this for a while, reading and learning, and this is my proposal for that problem.

The problem is no longer learning, it’s being believed

Today two groups coexist that look the same from outside. One oversells what they know: talks a good game, and there’s little underneath. The other is learning more and faster than ever, for real. The problem is you can’t tell them apart at a glance.

And on top of that there’s the stigma. You do something good and someone shows up: “haha, you learned that with ChatGPT, it’s copied, you have no credibility.” As if using the best tool available took away your merit.

Put the two together and you have today’s real bottleneck: it’s not accessing knowledge, it’s someone trusting that you actually have it. That’s a validation problem. It’s not solved by knowing more, it’s solved by proving what you already know.

Platforms shouldn’t compete on speed

This is where a lot of learning platforms are getting it wrong. Some want to be a chat that explains to you faster and prettier. Others use AI to pump out courses at full speed, but they’re the same courses as always, with the same structure and the same methodology. In both cases they’re speeding up the old thing, not adapting to the new game.

And against an infinite, free tutor you don’t compete on speed. That chat is already in Claude, Gemini, and ChatGPT, with far more resources behind it. If your product is that, you’re a shell on top of a model that isn’t yours, and it’s only a matter of time.

The play isn’t to run faster. It’s to do two things the speed race ignores: form judgment and vision, and validate that you actually have it.

Forming judgment and vision

Knowledge got democratized. Vision, knowing what to look for, is getting democratized too, because the same model that explains things to you can show you what you didn’t know existed. And that matters, because nobody learns something they don’t know is there.

What you can’t get by just asking the model is judgment. Knowing, among the hundred options the model opens up, which one actually works. And that’s earned in one way only: by building, failing, and understanding why it failed. Experience with consequences is what trains the eye.

Here the platform has a real job. Not to give you more content, but to put you to decide, to defend your decisions, to compare the good with the bad and see why. And something that used to be very expensive: accelerating your experience. Feeding you lots of real failures and decisions, yours and other people’s, well explained, so you gain in months the eye that used to take years. A prerecorded video can’t do that. The video hands you information in one direction, the same for everyone. It doesn’t react to your decision, it doesn’t press you when you choose wrong, it doesn’t put your own mistakes in front of you. It shows you the answer already clean, not the road of trial and blows that made it good. And judgment lives right there, on that road.

Where the model falls short

And there’s a point that to me is the most important, where profiles really separate.

It’s when you push complexity so far that the model no longer gives you a clear answer. It starts throwing several options at you, and you’re the one who has to choose. Or it tells you “uff, this is complicated, let me see.” Or you throw it a challenge and not even it has a clean way out.

Whoever has already fought with hard problems knows it: there are no perfect solutions, only tradeoffs. Every path gains in one thing and loses in another. And the question that really matters, the one the model doesn’t answer for you, is which one you choose.

That’s where things change. Where you pick a path and weeks later realize it was a bad decision. Where you get stuck and neither the model nor you know what to do, and you have to go learn from other people or open a book. That frontier, where the model falls short and you have to bring your part, is the new point of complexity. And that’s exactly where a platform should take you, not before.

Validate and certify

The other half is credibility. And here even what a certificate means changes.

The old diploma said “watched forty hours of video.” That’s worth nothing now. The new one has to say something else: “this person really understands, and can defend what they did.” It doesn’t matter that your project works, any model puts that together. The question is whether you understand how it’s built, why this way and not another, what you discarded and why.

That’s what separates the one who learned from the one who copied. And it’s what gives credibility back to the one who actually earned it. A platform that certifies real understanding, and not hours watched, is selling the only thing scarce today: trust.

The idea: a mini company of your own

If I put it all together, I imagine the new education less like a course and more like a mini company of your own.

There you build, and the platform doesn’t watch you from outside: it guides you, it keeps handing you harder and harder problems, it raises the bar. To solve them you have to create things, and along the way you get errors you didn’t expect. That’s where you really learn, solving at a level you hadn’t touched before. It’s just like a company, but it’s yours, and the risk is pretend while you learn. With the tutor beside you pushing you and accelerating your experience.

And this isn’t theory: in ethical hacking that’s already how you train. On HackTheBox or TryHackMe you move through harder and harder machines, each one a challenge you have to break into. The model works, it just rarely leaves its niche.

Now, let’s be honest: those machines aren’t fully real. Passing a challenge isn’t the same as preventing a hundred thousand dollar loss in production, and those platforms get criticized for exactly that. That’s why the mini company: the real challenge is getting as close as possible to facing real problems, with real consequences, not staying in the lab.

Practice and theory at once. Theory comes in when you need it to build, not before and in a vacuum.

And since it’s your company, you don’t learn one thing in isolation. You end up seeing how the areas cross, which today work closer together than ever. Sales with product and engineering. Finance with fraud and systems. Marketing with product and data. Operations with data and automation. And so on with almost any combination you can think of. Real problems get solved with those areas side by side, not each in its silo.

That’s why I see it less like a single-specialty track and more like a mini MBA, but practical. Modules you pass, each one crossing a couple of areas, until you build a vision of how a whole company works and understand how it all fits.

And when you close each area you certify, not for having finished, but for having proven that you solve. This isn’t even new: the coding challenge sites have always worked like this. You keep solving, you level up, and your profile shows what you did to get there, and anyone can check it. What would change isn’t the mechanism, it’s what’s behind it: not isolated exercises, but real problems you understood and can defend.

Publishing your work I’d leave optional, but recommended. In the end you’re after credibility, and showing what you did is the most honest way to have it.

I imagine something as simple as a subdomain of your own, validated by the platform, where your profile and what you built live. An address anyone can open and check. Sounds hard to scale. But that’s exactly the idea: today what used to seem impossible is becoming normal.

And that way everyone wins: you with real credibility, the platform running at market speed, and the company hiring with a signal it can finally trust. That said, you have to validate all the time, because where there’s certification there’s going to be someone trying to cheat. And since you’re teaching people to move fast, you have to be fast too. You can’t be the slow one teaching speed.

The problem doesn’t change, the tool does

And there’s something else I like about building it this way, around building and solving: it gets you out of depending on whatever AI is trendy.

Think about it. If your goal is to sell more, it doesn’t matter which model you use to do it. The goal is to sell more. If you’re going to cut an operation from twelve days to one hour, it doesn’t matter which AI you do it with. What matters is that it drops.

  Your focus
    │
    ▼
  THE PROBLEM          (doesn't change)
  sell more · operate better · build
    │
    │  is solved with
    ▼
  THE TOOL             (changes on its own)
  the model of the moment: Claude, Gemini...

When you teach people to solve that, and not to use the new version of some model, the content adapts on its own. Whatever comes out next week, the problem stays the same, and the tool underneath is the only thing that changes. That’s the difference with the course that’s born old: the one that teaches the tool expires with every version, the one that teaches how to solve doesn’t.

And on top of that, solving real problems is what expands your vision and forms your judgment. It all points the same way.

And won’t the general model do that?

The obvious question: if the LLM already explains and can even give you a test, what’s the platform for?

Because there are things the generalist, by design, won’t build. It barely knows you, with a general, loose memory, it doesn’t carry your history or your gaps over time.

It’s made to moderate you and please you, not to push you and form you. And it doesn’t build you judgment, experience, or vision either. That doesn’t come from an answer, it comes from building, failing, and correcting, and it’s not made for that.

And above all, it won’t issue trustworthy credibility to a third party. For a certificate to be worth something, whoever’s hiring has to trust who signs it, and that calls for a neutral validator, with a name and a reputation.

Certifying people isn’t OpenAI’s or Google’s business, and it doesn’t interest them. That gap stays open, and there a platform has something the generalist won’t take from it.

The same core

In the end, this connects with the other post: the scarce thing is no longer knowledge or access, it’s vision and judgment.

None of this is a formula or a closed truth. It’s just how I imagine education in this new era, after turning it over a lot. I’m surely missing things, and there are surely better ideas. I share it in case it adds something, and in case it gives a starting point to someone who wants to build in education.

And a question for the next one: if anyone can build a project, how much quality will it have, or will it be generic? Because producing got cheap. The good stuff didn’t.