It's such a joy to work in a domain where you can sometimes be amazed by what an AI can do and yet still often have to guide and correct it in substantial ways.

If we're going to do this, I would like to have some control over the stylistic direction because I would want the added details to match my head-canon expectations. Also wasted opportunity not showing some old classics but "remastered"

Stefan Reinalter i'm clearly not as deeply technical as you, but i like to think of myself as some who cares about their craft. what i'm seeing is an industry wide change (and hype) that i want to understand as deeply as i understand other areas of my craft. only then can i know what actually

🚨 Shocking: Frontier LLMs score 85-95% on standard coding benchmarks. We gave them equivalent problems in languages they couldn't have memorized. They collapsed to 0-11%. Presenting EsoLang-Bench. Accepted to the Logical Reasoning and ICBINB workshops at ICLR 2026 🧵

This is more evidence that current frontier models remain completely reliant on content-level memorization, as opposed to higher-level generalizable knowledge (such as metalearning knowledge, problem-solving strategies...)

this guy has 29 models on huggingface at page 2 ranking. no lab behind him. no sponsorship. $2,000 from his own pocket on GPU rentals. he compressed GLM-4.7 to run on a MacBook and quantized Nemotron Super the week it dropped. all public. all free. nvidia is a trillion dollar

This week's drama in ai

Please dont let it be something like cursor that kills the golden goose...

While i do think some users overexploit ai subscriptions (e.g. constantly creating new kimi accounts for $1/month sub) it's a bit sad to see the current arms race on locking out 3rd party harnesses. It's just in the spirit of devs to fiddle with and overengineer their tools!

Was feeling smug about having locked in a good VPS subscription before the RAM price hikes until...

We don’t talk about this enough. Opus scored 20% higher in Cursor than in Claude Code.

There's been no better time to be either barely literate or an erudite

We asked and Claude itself answered.

It's been about a week in and I'm still in disbelief that I found a job that's this cool in Singapore.

All the frontier labs are falling into such predictable, singular, and pathologically distinct brand personas. Almost like they are doing vibe pr lol

This but the opposite.

It is the year 2026 and charts have become an ancient technology lost to time. No one knows how to make them anymore. Cool stuff tho.

I'm about as pro ai and perhaps technocratic as they come. But anthropic, please. Why are you making this so difficult for us