preommr

> The same applies to software: you'll describe a problem and a network of agents will write a solution faster than you can Google it, download it, install it, and start using it. It'll work with your accessibility requirements out of the box; it's UI will be in your language; the data will be stored where ever you like; and so on.

This exists now.

You can have gemini AI - the that's built into google search - just make little applets. I just made one for a color picker, and generating palettes, including one that applies aces tone mapping. It showed made little adjsutable sliders and everything.

For a while, chatgpt web didnt count against usage, and you could use sol, it was isnane, you could upload a screenshot and just ask it to build out a little editor with threejs, and whatever else.

There's grok bot, dots, and other 'bots' that are just doing independent actions in the background. That can do digital actions like organize your documents, send emails for rebates, etc.

If you accept the shame, you can literally have bite sized story telling that are like mini soap operas, completely cheesy, predictable plots but that are very good guilty pleasures.

Vibecoded photoshop will never work.

If you're a casual user, just use AI directly - upload image and tell model to remove background, or make it cinematic, etc.

If you're a professional, you'll want the best.

Google Cloud is also the scariest because of how much damage it can do and how bad their payment system can be.

I recently loaded up on prepaid api credits for gemini and it somehow triggered some billing shenanigans in my linked accounts where it said I had a negative balance (from the credits), and they were going to discontinue my services. I had to reset some settings to sort it out, mainly using their chat ai and mine (because theirs gave me right status info, but wrong conclusions).

It's pretty messy across like aistudio.google.com, and their typical console, and google workspace business account. I'd be so fucked if they froze my account, I'd rather just pay openrouter to access credits in the future.

> show costs are piling up for these AI companies

OpenAI cancelling the $200/20x and then introducing the $500/25x was basically the canary in the coal mine.

I am looking forward to the 2026 numbers from Anthropic in preperation for their IPO, because that's where the really big numbers are going to be.

Hasn't stuff like this happened multiple times already?

A quick google search shows the time Rob pike got that email last year, then there was that AI that wrote a whole blog hit piece (although how much of that was human influenced wasn't clear).

AI sending an email doesn't seem that novel in late 2026.

yes.

A simple, "this entity is a sponsor, and therefore there is a conflict of interest and we will not comment on recent controversy" is enough.

If it's big enough, refuse to take further contributions.

I know it's not entertaining, but that's why we have video games.

> just replace the glass panels and tabs with something that actually fits the rest of the aesthetic.

Where is this glass?

Is it the lighter green cards on the home page? That's not really glass, that's just a different a lighter color scheme. The tabs also look fine.

like what?

The website is simple enough that any design decicion could just as well ahve been done by a real human.

The em-dashes, and the gradients were always egregious because they're strong opinionated decisions. This is just a straighforward, clean, tasteful website (really what most sites should aim for).

Except that it would be priced based on pass-through prices rather than existing value perception. It used to cost hundreds to send a letter on horseback, it costs pennies to send an email, consumers aren't going to tolerate spending hundreds when they know it take pennies to send.

Models are getting cheaper, and more efficient, but I don't see how they cross the threshold of full autonomous intelligence in the way that an employee needs to be fully accountable for their actions.

Incredibly obvious to anyone paying attention.

Altman shared a post yesterday that basically (I am ovrsimplifying) covered how the best coding, fastest, smartest models is less relevant than building generalist models because that's what builds a platform. Lots of reasons why, like how there's no stickiness for models which is a problem for monetization. They're also using these generalist models to then distill down to make other variants for specialized purposes.

So everything is about getting that huge collection of data and generalization.

What I wanted to say yesterday is that we hit 25M active users (Aug 30) - tibo

https://x.com/thsottiaux/status/2094252447271366730

---

"Happy Tuesday. Codex has hit 4M active users, adding over 1M users in less than two weeks. To celebrate we will reset the rate limits again in a few hours. Enjoy!" 10:52 AM · Apr 21, 2026

https://x.com/thsottiaux/status/2046602907077038501

Because it's not 10%, it's ~25mn (tibo kept putting out resets at major milestones). It was like 4mn a few months ago.

The api margins are like 70%-90% (depending on who you ask, and not taking into account r&d), even lower with luna and their recent cuts. The subscription is even worse because they often use it as a loss leader for building hype at releases.

Well now we know how/why opus 5.5 usage is so generous

I am getting increasingly worried that coding is not solved, and that AI won't lead to some kind of coding singularity where we never have to read the code any time soon.

In which case we've royally fucked ourselves that the level of engineering we've reached is... prompts. Because there is a deadline where we have to show productivity to justify all the investment spending.

People need to build with tools in a reliable, constructive way. Not vodoo magic based off vibes. We need better structured output, better transparency on what these models can do, better controls overla, maybe new ideas on loops graphs, and ways to use the models. Like, at least people were trying new things with jev.

since we're posting pointless gotchas:

Do we pay more to a lawyer if they handwrite a legal document using expert penmanship?

someone else mentioned how their blood pressure spikes from using microsoft products...

The sheer anger that I feel over how fuckingly stupid it is that they named everything copilot that basic conversation is difficult... I tried going to the article and got swarmed by paywall, finally checked the archive link and I think you're right.

Copilot was fantastic, it's usage was insane because it was by request not tokens for the longest time. You could make a long markdown file with lots of planned edits, and just let it rip.

It also had some really good models for basically free while it was in preview instead of being tied to a single provider.

And finally, it's vscode integration was really good. I use codex now, and changes to a thread are in this awful diff view that shows all changes in a single tab. And even that wasn't working for a while there. No clue if cursor is better never tried it because it takes a lot for me to swtich tools because of all the small things like shorcuts and configurations that I've set over a long time.

They are not having a good week.

I am not even excited about the expected reset because the model quality has been less than stellar.

> I may have to resort to polishing up the presentation. The fact that I'm trying to give something with no strings does not obviate the need to make it look palatable.

People say shit like this when they're the producer. But then turn around and nitpick when they're the consumer.

It's not selling something, it's about it being a proxy for if you understand the product well enough to communicate it's key points.

To wrongly quote Twain: "I would have written a shorter letter, but did not have the time"

> $10 OpenCode Go sub

I suppose this is as good a plan as any.

I had to migrate away from Github Copilot in June, and I am not looking forward to doing the whole moving thing again.

These aren't even that good in comparison to the stuff being posted on Reddit and twitter/X.

Some of the stuff I've seen is mindblowingly impressive in how they've captured quality creative decisions, and how the models can reason about visually appealing designs.

It understands things like continuity, themes, facial expressions, cultural memes/references, symbolism, etc. And it generates things using a lot of multi-modal behavior by using 3d, images, etc.

Some really incredibly impressive stuff.

From the article (that I am sure you read):

> In Mirvahedy v Henley, the claimant was injured after the car he was driving was hit by the defendant’s horse. The horse had pushed over a wooden fence and an electric wire fence and bolted from its field after being frightened. This test was met here because, when a horse is sufficiently alarmed or panicked – ie in those particular circumstances – it is a known characteristic of a horse to bolt and cover long distances.

So, it's not just 'lions', but regular animals that people know can behave in a particular way.

I would say AI companies know the cybersecurity dangers since they literally advertise it as part of their marketing.

---

And as a tangent, it isn't simple if we're talking broadly - e.g. I don't think Anthropic should be liable for agents that are directed by a user to write code that takes part in a crime. Nor should they be held liable if their internal security measures are lax in a way that an employee steals data that gets distilled into another model. But we're not talking about those edge cases.

> "Not necessarily"

Why did you even link that article when it doesn't help your point?

I supports the idea that the person responsible or an animal is held liable - it just makes clarificaitons on common sense caveates like when a professional is moving the animal. It even doubles down on making it clear that expected behavior of an animal is taken into account even if unlikely, like how ai swarms have a reasonable potential to just go awry and commit cyber crimes.

Yea, I think that there's pretty much a ceiling with day-to-day models that have already been hit months ago. Maybe you need SOTA for reviews, high-level planning, or research, but long running tasks like writing out a feature, testing, getting feedback and making refactors can be done for low-end models (like luna). And the margins on those models are basically evaporating.

> AI is not a bubble

When people say "AI is a bubble", they mean economically as a whole, which includes data centers.

Perhaps we need better terminology for "product useful; numbers nonsensical"

The AI hype cycle is always looking for the next big thing. It really doesn't take much for enthusiasts to get very excited and push something into the stratosphere. Just not having a vibe coded website, and someone that made ChatGPT is enough to set them apart. Hitting pain points like pricing and speed and also implicitly mentioning llms (even if it's to mention it can't generate text in contrast to llms) make it seem like a major step-up.

Well I think there's an expectation that it's similar to the probability score, something that's outputted by the model itself, and so there's some level of "intelligence" (e.g. being able to recognize if the subject matter is relevant to the data it's been trained on, or as an accumulation of errors e.g. it couldn't figure out the question).

> its confidence scores

important to note that the "confidence" score is... maybe not what people think it is - kind of useless, and just a convenience step from the probabilities.

from the docs: "confidence is a statistic computed from the probability distribution the answer already gives you." [0] I actually encourage people to visit the docs because it has a specific page on this with a little applet to really make this clear.

[0] https://docs.typesafe.ai/confidence

And yet there's none of this uncertainty or caution with his posts that then go on to have massive ripple effects because of his position at Anthropic and the marketting related to claude.

I personally have a lot of anger and frustration with many people in the ai hypesphere that are just mindlessly frolicking around without a care in the world, happy to speak into the megaphone offered by masses that are in a rat-race to avoid some AI dystopian hellscape that keep getting painted by these thought leaders... and then going "oopsies... I am just human guys... y so mad?!"

This will be insane for tool usage, and probably where the major economics for day-to-day usage will be.

The goal is going to be to use llms to distill operations down to some dsl, and pass it into something like Jev.