We should all see the Government Banning Mythos / Fabel as a sign for two things:
1. We need to increase investment in securing your computer systems. Reassigning not firing for the big companies. (They're not going to secure themselves)
2. Investing in local models.
I find it interesting that companies focus on the big features, when it's the little features that drive users nuts. Eg - Teams on mac why do I need to type my name EVERY time I join an external meeting?
Managed to get codex cli working with vllm and qwen 3.6 27b. Was a lot more painful than it needed to be (custom chat templates, proxies). I think time needs to be spent improving the local model space.
@ThePrimeagen just watched youtu.be/zaGOKd4jqEk?si=gh0g… and the thing that keeps ringing in my head is a lot of enterprise companies are not ready, they're not shortening their release timeframes with automation. In the world of mythos, a month -> quarterly release doesn't' work.
I've tried codex, kimicode, claude code and opencode. So far I like the UI aspects of kimi best, but the use of rust in codex. Much faster than the rest.
I've now changed my mind on this - opencode has the best ui, codex is the fastest. I still wish opencode was written in rust. LOVE the RHS "current task view" in opencode and how it uses my todos from a backlog.
Only one chance in this lifetime…
Like watching sunset at the beach from the most foreign seat in the cosmos, I couldn’t resist a cell phone video of Earthset. You can hear the shutter on the Nikon as @Astro_Christina is hammering away on 3-shot brackets and capturing those exceptional Earthset photos through the 400mm lens. @AstroVicGlover was in window 3 watching with @Astro_Jeremy next to him.
I could barely see the Moon through the docking hatch window but the iPhone was the perfect size to catch the view…this is uncropped, uncut with 8x zoom which is quite comparable to the view of the human eye. Enjoy.
The #1 thing I built for openclaw is a rust based monitor that checks ram and disk util, along with gateway 10s response timeout, and restart counts -> telegram. Helped diagnose a lot of issues that were invisible to me . (node OOM, total OOM, disk util, Node high cpu) @steipete
We’ve identified industrial-scale distillation attacks on our models by DeepSeek, Moonshot AI, and MiniMax.
These labs created over 24,000 fraudulent accounts and generated over 16 million exchanges with Claude, extracting its capabilities to train and improve their own models.
An easy win for DOGE would be to allow booking online appointments for Social Security, (plus improve the computer systems to give you status updates). 100 minutes estimated hold to book an appointment to update citizenship jeez. @elonmusk@VivekGRamaswamy