notatoad

seriously - if netlify can do 15ms warm start + 10ms cold start, wtf is lambda doing for the other 1500ms?

i don't know if you could say they're buying it "happily". more like begrudgingly.

the deal has a 30-day cancellation policy, and they raised a bunch of debt around the same time to fund their own datacenter expansion.

I suspect the $500 will sell well. Maybe not for the staff developer, but I can definitely see the exec who thinks he can replace all his developers by vibe coding things himself inbetween meetings seeing a 6x speed up button and buying without a second thought.

My side by side evaluation this week was to build a tool for mounting my app’s UI components in a headless chrome and feeding mock data into them, for the purpose of taking screenshots for help docs. Not super complicated, but a real task I needed done.

I gave the task to codex first, sol 6 xhigh. it took a couple back and forth prompts to define the project and then it worked for a bit and to took a couple more prompts before I decided it was good enough - not perfect, but close. It re-implemented some wrapper components in a simplified way that lost some of the UI, but it would work.

Opus 5.5 high took the same prompt with no back and forth, it just went off and one-shotted a tool that takes pixel-perfect screenshots of exactly what my app looks like.

>You can put it in a sandbox, it doesn't make a difference, for it to be useful it inherently needs wide, unattended access.

only as long as you're trying to replace a human's job. because human jobs are structured to do a wide variety of things.

a useful agent needs a wide variety of inputs, and one single restricted action it can take. it doesn't need permission to do everything, it need permission to do the tiniest possible useful thing it can do, and nothing else.

they've reversed this decision https://www.yahoo.com/entertainment/movies/articles/youtube-...

these are just openrouter stats.

they don't have 64% marketshare. they have 64% of openrouter usage, because people mostly don't use openrouter to access claude or chatGPT.

The generalization I would use is that governments should tax any scarce resource based on the amount of revenue it should produce if used effectively, rather than the revenue a given consumer can actually produces. Whether that is land, radio spectrum, or clean air.

“Undesirable behavior” is overly broad to the point of being actively misleading, and gets you into sin taxes. smut or drugs are not the scarce resources anybody is talking about here.

that still essentially works. ask it to write a plan doc to a file. then when you're ready to implement, start a new session with a prompt to review the plan doc and then start building.

well then you aren't using this project

That was not my understanding. Unless this installation was supposed to be producing fuel on site somehow, fuel cells are a mechanism for turning fuel into electricity.

https://www.bloomenergy.com/technology/

>But what if the problem is that none of the alignment techniques that are applied to models today actually work?

if that were true, the millions of people who use these models that have had the alignment training applied would notice that. the reason we all believe that the models doing the hacking are models that haven't been told not to hack is because the models that are told not to hack don't do this.

after a few hours use, my impression is that Opus 5.5 writes like gemini. drastic improvement over all opus 5.1's claudeisms.

and i think i was generally more tolerant of 5.1's writing style than most people.

yeah, Astra burned through 70% of my weekly usage in ~5hrs on a $100 plan. even fable doesn't run out that quickly for me. it's great, but it's on the same tier as fable for me - use it sparingly, only when really necessary.

>and potentially about Anthropic future profitability too

have they ever shared anything about their revenue mix between consumer plans vs per-token billing? this is a revenue cut on their API billing, but they're not saying anything about increased limits on the plans. so all the plan revenue just got more profitable.

typically, when any AI company says a model performs as well as fable, all they're really telling us is that the benchmarks that exist for measuring AI capabilities aren't very good.

i don't think that's the point anybody is making. "good lawyers" is not the profession at risk here. a top tier courtroom lawyer won't be replaced by AI.

but how much of the revenue of the average law firm comes from that, vs the day-to-day "we need this relatively routine contract reviewed" sort of work?

canada already has CETA for EU trade and CEFTA for trade with the EFTA. it seems like "associate membership" must mean something that goes beyond just a free trade agreement, because that already exists.

maybe wait until their plan B succeeds before calling it genius. all we know so far is that plan A has failed.

This seems like a useful write up of the concepts involved here, but I just hate reading Claude’s writing style so much.

This is content marketing, at least give it a review pass before you hit publish… if your marketing is unpolished AI slop, I have to assume your product is too.

I think the abstraction is only part of the problem. The other part is that all these companies offering ai products are deeply untrustworthy, and I don’t want to let them any further into my stack than I have to. Claude code and codex are great because they are lightweight, and operate on top of the rest of my tools with little to no change needed, so they can be eliminated or migrated away from with zero cost. They’re not a dependency of anything. And that’s as much as I’m willing to trust OpenAI or Claude.

a month of opencode go is cheaper than lunch, so trusting them with this but not your lunch seems reasonable.

I had the exact same experience. Charging an Apple Watch is a routine thing, charging a garmin is an infrequent annoyance.

And like the other commenter says, I could just fix my habits. But no, that’s not how my brain works

Doing it live isn’t going to bring Steve back. He was a natural showman, the people doing it now aren’t. They’re engineers and executives. Putting them on a live stage is going to make it even tougher on them.

It’s been a while since I watched an Apple presentation, but I caught a bit of this one. And I spent the whole time trying to figure out who it was for - the presenters didn’t seem to express any genuine excitement about the products, and it didn’t convey any useful information that couldn’t be better shared by just posting the slides.

The only thing I can see is that it generates some clips for the presenters to share on their LinkedIn pages. I assume there is prestige in being selected as a presenter, and the rest of the show serves its purpose simply by existing.

Seems to be lots of people in here worried about “if i had”, and nobody who actually has done this.

Anybody actually using deepseek in a production system affected by this want to share their experience?

this is what regulatory failure looks like. There's no downside for google to do this, because nobody is going to hold them accountable. Nobody is telling them they have to explain their actions, and it's easier for them not to, so why would they?

from the thread last week, it seems to be installed by default on every mac install, possibly not on windows installs.

https://news.ycombinator.com/item?id=49528079

i think that's a perfectly coherent and reasonable way for both these apps to work.

people don't want AI features leaking into their non-ai apps. I'm not an AI hater - I want my AI app to be able to read and write word docs, but when i decide to open up libreoffice i'm doing that because i want to write something myself, not use AI. i don't need my word processor to bundle an inferior version of what the ChatGPT can do for me.

does google actually use their own models in consumer products? i'm not an android user, but the weather widget in the google search results for my area always seems to be attributed to weather.com.