gentlewater

Limp-to, maybe.

It _was_ fantastic. Then it transitioned to abysmal over night when they switched to usage based billing. Now, if you put the same 50 USD towards GitHub copilot compared to Codex, I’m pretty sure Codex gives you 20x the usage or something. Really sucks for us who are stuck with GitHub copilot at work for compliance reasons.

Which has really only been accepted because 99% of people don’t understand how it works. Hard to be outraged by something you can’t see or understand. In my book, the practice is akin to malware. I recently opened a website for an AI service in incognito mode, because I didn’t care to have it in my search results. Despite avoiding third party cookies here, the site still fired off a tracker to Meta, who then correlated my home IP address with my Facebook account, and filled my feed with ads for this AI service. When I was tracked like this despite a somewhat informed defense, which defense do normal people have against this? None.

I use Kotlin as my main language and I’m very satisfied with it.

Can I trust code I’ve written myself with no guarantees from the language? Maybe. Can I trust code written by dozens of other developers (and/or agents) working on the same project over multiple years? Definitely not.

It’s still missing null safety, right? Which means it’s still a hard no for me.

Oh, I’ve tried that too. It will promise to keep it in English from here on out, then switch back to Chinese after two or three exchanges. When ever it needs to do a web search, it seems to load so much Chinese text that it forgets any language instructions. Just thought my experience yesterday was more to the point. Right now the chat is absolutely hopeless.

I finally uninstalled the app yesterday after giving it plenty of chances over several months. Yesterday, I asked it whether «DeepSeek has fixed the issue where it erroneously answers in Chinese?» and it answered in Chinese.

Not really comparable IMO. Astra and Fable are not the every day workhorse you reach for to do basic tasks (unless your company has fuck you-money), they’re the tool you break out when you need the absolute strongest performance. There are plenty of tasks where finding and fixing one or two extra edge cases saves the business a lot of money, even if the cost is high. The best example would be scanning for vulnerabilities, if these models weren’t kneecapped in that area.

ARC-AGI this and ARC-AGI that. All I care about is ARC-pelican.

Darn it, model probably escaped again

Ah. Sweet.

What’s the code for http unicorn? /s

I’ve been refreshing hacker news constantly for a week now waiting for v4 pro, after they stated it would follow «soon». I have learnt «soon» is a matter of definition.