Turns out that the Starlink software, built for managing the movement of vast amounts of data through space and to the ground, is also very well suited to training super intelligence
@bot can now connect to private MCP servers on your tailnet without exposing them to the public internet.
Powered by Tailscale OAuth apps, now GA, so you can build secure agent integrations, too.
Read more here: https://t.co/cmn8tpM6w7
We have dumped JAX/SGLang and rewritten the entire RL and inference stack from scratch in C/C++, which took some extra time.
Training (aka pre-training) is done. Lot of cleanup was needed on the SFT data to get that mostly right. Now, 4.8 is in RL and the evals are starting to climb. We will do a soft rollout to Bot users later this week and see how it goes.
Grok 4.8 will bear no resemblance to 4.7 (thankfully). It is not yet incredible, but, this time, the foundations are built right, so I expect it to evolve to be something special.
@tasarren I'm finding this as well.
Just one too many times now.
@poteto Grok4.7 is getting stuck in a loop and it's eating up limits, your welcome to read my sessions ๐.
Grok @Bot (and the shared pool) got a full weekly reset. After the โcosts were unsustainably highโ
After the overnight efficiency fixes, this is the right move.
Keen to see if the token burn actually slows down this week.
@kapilansh_twt I canโt imagine anyone learning a new programming language today.
Why would you?
Choose the best language for the given situation and hand it to AI.
Sad to hear about Patrick Stewart.
I grew up watching him as Captain Picard of the Starship Enterprise - he was such a constant, steady presence.
Feels like losing a piece of my childhood.
Rest in peace, Captain.