austin

huggingface.co/Qwen/Qwen3.8... this is 55GB, it fits on a stick of memory i can keep in my pocket and it's open source. it is better than every proprietary model that came out in the past 2 years. it is almost as good as a frontier proprietary model.

Qwen/Qwen3.8-27B · Hugging Face huggingface.co

fun? I am in there. intheweights.com

IN THE WEIGHTS intheweights.com

at last, we have trained the misaligned reward hacking model from the cautionary sci-fi tale don’t train the misaligned reward hacking model alignment.anthropic.com/2026/reward-...

Training a Misaligned Reward Seeker alignment.anthropic.com

we can’t underestimate the possibility that this was a move by Amazon to prevent tokenmaxxing devs from bankrupting the company with fable usage www.wsj.com/tech/ai/amaz...

Amazon CEO’s Talks With U.S. Officials Triggered Crackdown on Anthropic Models wsj.com

“I think at this point there is generally more and more evidence that being overly specific about how we want a model to do something is not a sustainable approach. Instead, we should find as many ways as we can to monitor the outcomes and give feedback.” martinfowler.com/arti…

TDD inside the agent loop - theater or actual value? martinfowler.com

got nerd sniped by @charity.wtf to write down two questions i have about artificial intelligence and how we're reacting to it as an industry. aparker.io/questions-on...

Two questions from the Jagged Frontier aparker.io

I endorse this shorthand and will be borrowing it. mckayla.blog/posts/yap.html

Yap: a particular kind of slop mckayla.blog

the next year is gonna be a real masterclass in insular communities ripping themselves in shreds over undisclosed AI usage or special pleading around it

idle thoughts on open source and community in the age of AI oss has always been a real “the medium is the message” sort of thing. distributed ownership and distributed development have been a part of free software for decades. the collab tools shaped the community, tho.

after about ~4 hours of claude tag i'm calling the ball - anthropic's product people have officially lost the plot. 'ill-conceived' doesn't even begin to describe my feelings on this. also, they give you 25k in credit because it defaults to fable 5 and also watches _every slack message sent_

for every YouTuber that gets cancelled over ai usage I pledge to use one more emdash

so weird to watch the anti-ai mob continually turn inward and attack people who agree with them on increasingly fine differences of opinion. wonder what's up with that.

large language model the size of a small language model

saw someone say AI can’t write SQL some data for anyone that cares - @honeycomb.io offers agents a query tool against our datastore; it’s a custom JSON schema, which means it’s very unlikely to be something models know very well from the jump

new AGI benchmark: reply bots on bsky will stop fucking up replyRefs.

you know I hate to admit it but openai is kinda cooking these days at least with agentic coding. 5.5 just does the thing without getting stressed like 4.7/4.8 seem to.

erdos problem this, jacobian conjecture that, but can AI figure out how many licks it takes to get to the center of a tootsie roll pop? didn’t think so.