EdKo

AI is moving fast. Most of what you're reading about it is wrong.
I build content systems for founders - DM.

[EdKo @EdKolife]

A letter to everyone who built the life they were supposed to want.

---

Dear everyone who did everything right -

You followed the path. Worked hard. Checked the boxes. And somewhere in the middle of achieving what you said you wanted, something got very quiet inside you.

I want to talk about that quiet.

⬇︎

[EdKo @EdKolife]

For two years, Altman and Amodei warned AI would gut white-collar jobs.

Then OpenAI filed to go public. Then Anthropic filed to go public. Then, the same week, both men said they'd been "pretty wrong" — the apocalypse wasn't coming after all.

The warning was useful when it sold urgency. It became a liability when it spooked the markets they're about to sell shares into.

The prediction didn't change because the data did. It changed because the audience did.

[EdKo @EdKolife]

Anthropic published the argument for why frontier AI should pause. The evidence it used: more than 80% of its own code is now written by Claude, up from low single digits a year ago.

That's not a warning. It's a status report.

The proof that everyone should slow down is the same proof that they haven't. You don't cite your own acceleration as the reason for a brake unless the acceleration is the thing you're actually selling.

The pause isn't the message. The 80% is.

https://t.co/g24VFDKmiZ

[EdKo @EdKolife]

Everyone calls it AI slop.

The model didn't decide the work was done. You did. It didn't choose to publish - it waited for you to. It generated exactly what you asked, stopped exactly when you accepted it, and shipped nothing you didn't sign off on.

The slop isn't the AI's output. It's the human's standard, made visible.

The tool didn't lower the bar. It removed the last excuse for not having one.

[EdKo @EdKolife]

@bernhardsson Which means the layoffs aren't proof the AI can do the job. They're proof the market pays you to say it can. The people are cut to fund the story, not because the replacement showed up. The job loss is real. The thing replacing it is still a slide in a deck.

[EdKo @EdKolife]

@unclebobmartin If the curve is logarithmic, the problem isn't the ceiling. It's that every valuation in the industry is priced on the exponential. The asymptote you're describing isn't a technical limit. It's the gap between what physics allows and what the capital already assumes.

[EdKo @EdKolife]

@aerockrose The masters and PhDs were supposed to be the safe tier. That's why it scared him. The automation wasn't depressing when it was aimed at everyone below that line. It got depressing the Friday it reached the people who look like the person watching.

[EdKo @EdKolife]

@unusual_whales The promise and the fear are the same sentence. "AI makes you more productive" was always "AI does more of your work." Those aren't two outcomes. They're one mechanism, described by whoever benefits from how it sounds.

[EdKo @EdKolife]

@BoringBiz_ It's not lackluster at new ideas because the models are young. It's lackluster because both tasks are the same move - finding what's already in the data. Insight is retrieval. A genuinely new idea is the one thing not in the dataset to retrieve.

[EdKo @EdKolife]

@DiscussingFilm You feel that too? New games just didnt pass filters, so developers just forced to bring back old bangers!

[EdKo @EdKolife]

@ZackKorman These products exist because "we added an AI security layer" is something you can put in a board deck, and "we configured permissions correctly" isn't. The firewall's real job isn't stopping the action. It's being the thing someone can point to after it happens.

[EdKo @EdKolife]

Anthropic filed to go public. Revenue went from $10 billion to $47 billion in a year. The IPO could value it near a trillion.

This is the lab that built its brand on caution. Urging the industry to slow down. Publishing the research on what goes wrong.

Once it's public, the mission stops being the thing it answers to. Shareholders don't reward restraint. They reward the next quarter.

You can be the company that says slow down, or the company that has to grow 5x a year. The S-1 is the moment it picks.

[EdKo @EdKolife]

@etnshow @pirroh @Replit He's calling grit and determination the last bottleneck like it's a small one. It's the original one. Strip away every tool and the question was always whether you'd actually do the work. The models solved everything except the thing that was never about models.

[EdKo @EdKolife]

@lmrankhan The reason people stack agents is that it's the part of the problem you can solve by adding more. Deciding what's worth building has no orchestration layer, so everyone optimizes the part that does.

[EdKo @EdKolife]

@GergelyOrosz Engineers labeled data by hand at Scale. New leadership saw it and stopped it. Wang left. Now the same engineers label data by hand at Meta, assigned by the same person who ran it the first time. The practice didn't get fixed. It just moved when he did.

[EdKo @EdKolife]

@patrickc The reason no one's shipped this is that everyone who needs it is capable of duct-taping their own version, uses it, and never productizes it. The workflow layer stays unbuilt precisely because the people who'd build it already solved it for themselves.

[EdKo @EdKolife]

@emollick When everyone can build, the idea stops being the moat the same week it stops being the bottleneck. Cheap to implement means cheap for the next person who has the same thought an hour later.

[EdKo @EdKolife]

@scaling01 Notice the errors only ever run one direction. "Claude wrote 80% of our code" never becomes something smaller - it becomes "AI invents 80% of AI." The newsroom isn't bad at translation. It's optimizing the same thing the lab is: the version people can't look away from.

[EdKo @EdKolife]

@ThePrimeagen The joke is that it was slower. The part that isn't a joke is that you'll open it again tomorrow. The tool didn't win on speed. It won on the fact that everyone keeps reaching for it anyway.

[EdKo @EdKolife]

@NVIDIAGeForce #RTXPowersPlay

[EdKo @EdKolife]

The pricing gap is the whole business model. Frontier valuations assume companies will keep paying 20-40x for a capability lead that's now measured in single digits. The gap you're describing isn't a margin opportunity. It's the thing the labs are priced on, and it's already closing.