Altay Prime

@elonmusk Can you please make it available to SuperGrok users too soon? I think lots more people will provide feedback that way.

Reminder that Grok Build is iterating extremely fast and we are highly responsive to critical feedback. Fixes & upgrades are dropping every day.

@TSLA_inside_ Schrödinger’s spring update. It exists and doesn’t exist simultaneously.

With Tesla’s new Spring Update, the LED strip on the door panel can now light up red as a blind spot warning. Really well done by Tesla.

@MicrosoftLearn Reminds me of T-600.

We’re introducing a new GitHub Certified: Agentic AI Developer (GH-600). As AI agents become part of modern development workflows, this role-based certification focuses on how developers and teams operate, supervise, and integrate agents across the SDLC. If you’re already working with tools like GitHub Copilot or exploring agent-driven workflows, we’d love your input. Learn more and get involved. msft.it/6013vRHHZ

This 1956 Bell Labs paper is the foundation of every $1B hedge fund. A blackjack player used it to make $800,000,000. Renaissance, Citadel & Jane Street still run on it. 10 pages. The single most useful research paper you can find in open access. Bookmark & read tonight.

@techdevnotes Grok 4.3 with OpenCode have been pretty unusable unfortunately. So I think they either have to make a magic harness or release a stronger model. I’m a heavy user of codex and Claude and there is still a big gap for real work.

Should xAI wait to release Grok Build when 1T+ model is just around the corner ...

May introduce you to Cybertitan? 6 motor; 1,700 hp; 1,800 lb-ft; Tri-Axle, 42,690lbs tow capacity.

SpaceX valuation to the moon?

“listen to your body” my body: go eat a ribeye the size of a T-Rex with butter and mashed potatoes and a whole cheesecake.

Well hello there. New UI?

@Tesla_Optimus Yea stop slacking!

Returning to work at Tesla Diner this Thursday

@3blue1brown you should make a video about why self supervised encoders want to be normal. The rabbit hole starts with LeJEPA paper from @randall_balestr and then there are some nice follow ups. Like weak sigreg; and “why self supervised encoders want to be normal” paper. I found LeJEPA style training works phenomenal in my esoteric training datasets.

This is so awesome.

@grok help

I’ve done extensive coding experiments with Grok 4.3 + Open Code past weekend and sorry to say it still is significantly behind Sonnet 4.6. Improvements I noticed: - tool calling much better - instruction following better Gaps: - writes code without researching - makes a lot of mistakes and gets stuck on a loop trying to fix it - i tell it to change X to Y. It tries and gives up without even tying sensible next steps. Then reverts everything and tells me to go do it myself :) sonnet does the same task one-shot - tries a lot of random things without proper debugging and makes it worse - grok 4.20 (reasoning) debugged and fixed things 4.3 struggled with with a single prompt

@zlohnref I have this dilemma every day now.

If FSD pulls in my garage, but not far enough to shut the door, is that ... Discomfort - I'm uncomfortable shutting the garage. Preference - I prefer not to hit my car with the garage door. Navigation - It didn't quite make it to the pin. Critical - My wife will kill me if the garage door hits the car. I just leave a voice saying tell @Tesla_AI how rad it is that the car pulls in my garage and ask nicely to make it pull forward until it hits the stop block 😃.

Can someone please buy Spirit Airlines assets and make American version of @Ryanair called Billy Air and dunk on their customers on X.

I used to report a lot using the voice interface. Not a big fan of the new menu because there are so many things that don’t fall into those categories and I can’t skip or report. I end up reporting incorrectly. If the menu is to stay; then add Other button or add an option for voice feedback. Voice feedback can be classified in Tesla side using audio or speech to text models, no?

Tesla recently updated its FSD disengagement reason menu to add the much-requested "Navigation" option. What other changes would you like to see in the menu that pops up when you take over? Learn more: notateslaapp.com/news/4065/tesl…

Does @dezmondOliver need to do the one-wheel didgeridoo dance again for us to get FSD 14.3 @Tesla_AI ?

What kind of things are you doing with Cowork? If it’s not exclusive to what Anthropic has; you can try OpenCode. You can hook it up to any model API; or use free ones if you aren’t doing confidential or secret work. I got A LOT done with OpenCode and their free Kimi K2.5 and Minimax M2.5 when they gave it for “free” to collect data. Idk what models they have today.

@tunapedia 💔 what a rug pull... but it makes sense unfortunatelly

Can anybody use Claude Cowork for more than 5 minutes without hitting a usage limit on the $17 tier? The moment you give it a document it seems to shoot itself in the foot. It made 3 copies of it and burnt the tokens on writing and reading them?... 🤷‍♀️