QK-Clip: Taking Muon Further on the Scaleup Journey kexue.fm/archives/11126 Interpreting the Key Training Techniques Behind Kimi K2: QK-Clip and MuonClip.

Value Drift Jake No this is just young people being stupid. Some of us have actually lived under socialism and we don’t want it back

Eric Jang explains that 1X's world model is a video generation model at its core, but had to be retrained from scratch to make it action-controllable. He also notes that simulated people in the world model might eventually pass the Turing test.

Full interview with Eric Jang, diving into 1X’s comprehensive AI approach for humanoids: youtu.be/6egrojG273U

Wow, is Boltz-2 really that much faster than Boltz-1 and even ESMFold. Have to test it out soon. biorxiv.org/content/10.110…

Everyone missed the real China story. While we debated TikTok, Chinese scientists quietly built 1,250 novel drug candidates—up from 160 in 2015. That's not a typo. They added 1,090 innovative medicines in 9 years. The entire EU added 200. Here's how they broke pharma's

It’s funny that people on this site think major LLM efforts are talent-bound rather than org-bound. The talent differential has never been big between major orgs. Most of the difference in outcomes is due to organisational factors - like allocating compute to the right bets, and

lmsys.org/blog/2025-07-1…

In the 70's, a diagnosis of leukaemia was a death sentence Now, over 90% of kids are cured Thanks to medicine, science and yes..Big Pharma The pharmaceutical industry is far from perfect but yes it has created cures and saved countless lives

Obviously she was Foreign Secretary & approved the madness & coverup along with Wallace

Ok, I'm trying moonshotai/Kimi-K2-Instruct and wow... you weren’t lying! 🤯 + being able to generate this in under 10 seconds is a game changer. 👇 Give it a go with this app (no account needed).

Goedel-Prover-V2 The Strongest Open-Source Theorem Prover to Date

Dominic Cummings Begs the question: Do We Trust Anything Our Leaders (of any political persuasion) Tell Us? "All governments suffer a recurring problem: power attracts pathological personalities" (Frank Herbert)

We now have multiple models that have crossed the red line of compute that doomer influenced legislators have warned us about. What's happened? Nothing. Where all the catastrophic risks? Nowhere to be found. At what point do legislators realize they've been duped by a cult

I like the quote of mine they picked for this podcast prompt: "I think having a blog is actually one of the most influential things that you can do in modern society because nobody else does it anymore."

This looks hopeful - reminder to self, to seek Xeons with "Intel® Advanced Matrix Extensions (AMX)" for > Achieve 6-14x speedup for TTFT and 2-4x for TPOT v.s. llama.cpp. > Achieve 85% memory bandwidth efficiency with highly optimized MoE kernels.

🔥 Kimi K2 is not a reasoning model but you can use it like one 🎯 here's how: 1. give it access to the 'Sequential Thinking' MCP (link below) 2. put it in an agent loop 3. tell it to think sequentially before answering it's so cheap that it won't cost you that much. use a

Kimi-K2 is top 10 now. 🔥

This chart has been sent to me by a follower and it gave me a good laugh 🤣

Gemini CLI Pro Tip: Save and resume conversations with: /chat save <tag> /chat resume <tag> Enables you to snapshot points in conversations you may want to go back to. Tags are optional.