Ascending | Experimenting https://teapawt.org/t.co/WXqbftpXZU | File Ops https://teapawt.org/t.co/sMKEroUKvL | Booking Agent @amfionhq | Wellness https://teapawt.org/t.co/6ETx2Lj2gg
Joined June 2026
- Tweets1.6K
- Following351
- Followers21
- Likes5.7K
JFMN retweeted
does anyone have a rough estimate of the number of ballots swung via chatgpt/claude conversations? had to be a meaningful % of voters using right
JFMN retweeted
.@MIT What are you doing about this?
🤯🤬
“I may just try to get some $$…”
“…I saw that a girl sued NYU for one racist comment and got 2 million dollars…”
“But honestly maybe I’d rather get grants and paper acceptances than money…”
—MIT’s MeToo hoaxers Dr. Kristin Knouse and Izabella Pena discuss the anticipated spoils of their war against David Sabatini at MIT:
Pena hoped for a cash payout of some kind while Knouse wanted career favors she hadn’t earned.
(It’s also interesting how Knouse seemed intent on blaming Sabatini for her own stalling career and lack of productivity, despite the fact that he seemed to have bent over backward trying to offer mentorship and emotional support, no strings attached.)
JFMN retweeted
Tony Hinchcliffe praises Elon Musk
“Call me a d*cksucker whatever, I love it. I just think the guy is f’ing unbelievable and we’re so lucky to live in an era where he’s alive.”
JFMN retweeted
We built Pando, a new Git backend for our monorepo, to handle more pushes and fetches without changing how our engineers and agents use Git.
Pando cut push latency by ~80% and now serves nearly a petabyte of Git data a day. We can grow Git capacity alongside agent usage.
JFMN retweeted
There's a buzzy new paper about GPT-6 Astra being able to directly interpret model weights (prompts full of floating point numbers) - I tried it with Qwen 3.8 27B on my laptop and found that, after 45+ minutes, that model could achieve the same thing simonwillison.net/2026/Nov/3…
JFMN retweeted
Over the past 10 days, I became the digital Nick Shirley.
Using public & private records my AI agent swarm found ~$787 million in government billing that made absolutely no sense.
Thread 🧵
JFMN retweeted
I think Chandan is right. This is a significant release, the importance of which can hardly be overstated. It is a giant step toward recursive self-improvement.
For anyone questioning whether this post is overhyped, I think they are actually *underselling* the approach.
We did similar experiments and found that Claude models can write entire, functional neural net weights from scratch! csinva.io/weight-space-progr…
JFMN retweeted
i've reproduced the results from the MIT blog post today on weight-space programming. but one additional result i have to add: this is a new, step-change emerging capability that's appeared in just the latest set of models.
JFMN retweeted
I swear to you: at least one of the replies here will make you say "what the fuck" out loud
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
JFMN retweeted
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
JFMN retweeted
GPT-6.2 Astra is coming soon, and Sam Altman has described it as one of the most powerful models they've developed internally, capable of solving math problems that even the greatest minds on the planet struggle with.
OpenAI literally created a benchmark to determine whether this qualifies as AGI, and they're planning to make this model accessible to everyone, with the goal of "benefiting all of humanity."
For AGI to benefit all of humanity, it should help people with the everyday tasks that matter most to them. So we built LifeVal, a new benchmark shaped by feedback from thousands of ChatGPT users on where they most want help from a personal assistant—from finding childcare to accessing government services to managing personal finances.
Over the past year, we’ve used LifeVal to measure and improve how well our models handle these everyday tasks.
openai.com/index/lifeval/
JFMN retweeted
What if I want to use the RAM to do things that are prohibited by the terms of service of whoever is running the datacenter?
JFMN retweeted
have your bot, dot, muse, or duet agent monitor this list - adding every single ai slop poster i see to it so you can easily block them
teapawt.org/i/lists/21166591768940…
JFMN retweeted
If you don't have a personal agent filtering out the slop, you will drown.
If you don't have a personal agent running your cyberdefense, you will be hacked.
This is the new divide. Choose open or closed, choose the company you like best, but choose soon.
Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5:
hackertube.org/
- Excellent recs without any logins (state is kept in local browser session)
- Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads
- Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version
- HackerTube Longs ✊
A bit more on the process:
I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing.
Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.
