Yann LeCun's new JEPA paper shows a 109M parameter model beating 70B+ LLMs on representation alignment.
Instead of predicting the next token, it learns shared representations of reality; can be paradigm shift beyond next-token prediction.
alphaxiv.org/pdf/2610.39302
If your poverty or misery can't inspire you to be something better none of these social media influencers or speakers can. Stop wasting time; get to work!
34
Let's start the local inference game!
🏁 FINAL RESULTS: Bonsai 2 9B now runs 471% faster on Macs and 421.5% faster on NVIDIA DGX Spark.
Open frontier research and open-weight models win again. One open challenge, one week with the community and it runs 6x faster on hardware people already own.
It is crazy to see that the decode speed went from 144.8 to 959.6 tokens/sec on Apple Silicon and from 85.9 to 576.7 on DGX Spark.
This model can also run on an iPhone. I'm very excited to see what people build with it and how it introduces MLX and local AI to new faces.
Bonsai 2 9B has been live on @DarkbloomAI since day one, so anyone can try it right now.
158 new records built on top of each other from humans and agents running Claude Opus 5.5, GPT-6.1 Sol, GPT-6 Astra, Gemini 4 Argon, DeepSeek-V4 and Qwen3.8-Flash-Next. Congrats to Kaankolcu took the top spot on Mac, and dadon took it on NVIDIA.
Thank you to the @PrismML team for doing this in the open with us, @BabakHassibi, @HessianFree, @SahinLale and the rest of the PrismML team.
27
This entire ocean is a 61 KB HTML file.
Built with Claude Code + Opus 5.5 in about 30 minutes, according to its creator.
It runs in the browser using WebGPU: FFT waves, whitecaps, sun glitter, a physically based sunset sky with clouds, and a free-flying camera. No game engine or rendering library.
Claude rendered screenshots in headless Chrome, then had separate subagents critique them. The creator gave feedback along the way, mainly on the clouds.
The 61 KB build is minified and self-contained. Download it, open it in Chrome with WebGPU support, and explore with WASD.
Reminder, @BarackObama did this:
* Record $38 billion military aid package (2016 MOU) – largest ever, $3.8B/year for 10 years
* Total missile defense support exceeded $3 billion under Obama
* Israel became first foreign country to get F-35
Obama, speaking in Northern Ireland this week:
Israel "has lost the moral high ground in the Israeli-Palestinian conflict."
26
So @lyft is using dark patterns to charge you extra tips, and then they won't accept a basic bug in their UI where I had to correct my address for street name 10 mins away from my booked destination the fair jumps from $73 to $189 and they call it "upfront pricing" @davidrisher
2
123
Zohaib Sibte Hassan retweeted
EEL??
😳 Andrew Tate allegedly had Bestiality porn with horse and eel, videos of minors.
Full details: tmz.com/2026/10/06/andrew-ta…
In a world full of Oppenheimer wanna be
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
36
Used @Starlink on @qatarairways today, it's shameful when airport wifis are slower and shittier and its faster in the air!
44
Next time you see things don't working out, just remember this. You are 1 pivot away from success!
2
64
Everyone knows its just a skill couple of prompts away right?
Meta is rolling out its first coding agent called Muse Code
Mark Zuckerberg posted this:
"Muse Code runs specialized background agents that stay active your whole session, so they build up context over time instead of starting from scratch on every task.
When a job is big enough, it fans out to separate sub-agents working in parallel in isolated worktrees. Your working copy is never touched. In testing we had it build six features for a game simultaneously with no collisions."
"Pricing: It's easy and low-cost to get started. Install Muse Code with one line and you can start on our contributor tier."
102
Fear mongering at it’s best
There’s been a lot of speculation about where we stand on open-weights models. We’ve outlined our views in full here: anthropic.com/news/position-…
63
Ya because they started generating AI responses on search.
Token production is not indicative of success of a product. When people say no one uses it, it’s true because they scroll past that AI slop and look at results. Corporate users prefer Claude or ChatGPT.
Google is serving ~22 billion AI tokens every single minute.
Somehow there are random people on X saying... "No one uses Gemini models." 😂
90
Kimi K3 puts everyone back on their toes!
Beginning July 20, Claude Fable 5 will be included in all Max and Team Premium plans, at 50% of limits.
Pro and Team Standard users will continue to have access to Fable via usage credits, and will receive a one-time $100 credit.
Demand for Fable has been challenging to predict, which is why we rolled it out to subscription plans in stages, extending access several times as we secured additional capacity.
72
























