Elon Musk just revealed something huge about Grok's new SI training and inference infrastructure
The engineering behind Starlink is now helping power the development of Grok
SpaceXSI has completely rewritten Grok's reinforcement learning and inference software from scratch, replacing JAX and SGLang with its own system built primarily in C, with a little C++
And the most important part is where this engineering expertise comes from
The new stack is built on the incredible work of the engineers who created Starlink's incredibly complex satellite network
Elon explained:
"At SpaceX, we write our own headers and calculate our own time (includes leap seconds). Not many people will know what this means, but those that do will appreciate it."
Here's why that's significant
SpaceX engineers build and control critical parts of their software themselves, even down to the code interfaces and precise time calculations needed for satellite operations
This gives them much tighter control over how their systems operate, reducing unnecessary software dependencies and allowing them to optimize performance and reliability
Now that same engineering philosophy is being applied to Grok
dumisanitatiana retweeted
Fable is right more often than Carbon, but Carbon is wrong less often than Fable
*according to omniscience
dumisanitatiana retweeted
Masayoshi Son’s SoftBank reserved enough gas turbines to generate the power of six Hoover Dams. It cost the company billions of dollars in losses. on.wsj.com/4AkINFX
dumisanitatiana retweeted
Taps the sign
Masayoshi Son’s SoftBank reserved enough gas turbines to generate the power of six Hoover Dams. It cost the company billions of dollars in losses. on.wsj.com/4AkINFX
dumisanitatiana retweeted
Replying to @DailyIranNews
@grok did Muslim involved in any genocide throughout history?
dumisanitatiana retweeted
JUST IN: New York man pleads guilty to running a fake adult day care operation as part of a $127 million Medicare & Medicaid fraud scheme.
dumisanitatiana retweeted
Our teams put Gemini 4 Carbon to work on a major security challenge: mathematically proving the code underlying the Signal protocol.
Using AlphaProof Code, @GoogleDeepMind researchers guided agents to verify the protocol down to compiled device code, closing the gap where real bugs hide. The result:
- Gemini 4 Carbon generated over 6M lines of machine-checked Lean proofs
- years of expert manual verification compressed to just weeks
~1% of the codebase required human review (Lean automatically checked the rest)
This is helping unlock scalable formal verification for building secure software, laying the groundwork to fundamentally shift the advantage to defenders. We’re open sourcing the results for the security community.
dumisanitatiana retweeted
Bukele’s possible success could reveal there was NEVER real dedication from the international community to help change Haiti’s status quo, only band-aid solutions. There was no real plan to fight the gangs.
We’re on our own, & have always been.
It just it was never that clear.
dumisanitatiana retweeted
I'm starting a new thing here:
I'm requesting all labs to make their system cards available in jsonl or whatever.
just make the data accessible so we can stop wasting time, power and water on extracting all data from the system cards
dumisanitatiana retweeted
What about power consumption?
AI power use is trivial compared to other industries like steel or cement production (and no one is upset about those).
dumisanitatiana retweeted
Every time I think European voters cannot be more in denial about pensions, they prove me wrong: they are even more in denial than in my worst nightmare.
The average French person believes there are 6.4 workers for every retiree. The real number is about 1.8.
dumisanitatiana retweeted
Anyone with a laptop and a hundred dollars a month can change the world now. While on holiday.
My AI agents found an unmapped Maya site in Mexico, using LiDAR data from a NASA dataset.
72 hours from a random thought on vacation to a reply from Mexico's national archaeology institute.
What a time to be alive…
Readers added context they thought people might want to know
This site is not unknown or unmapped. Per Dr. Sarah Parcak, Dr. Francisco Estrada-Belli of Tulane University identified, researched, and published this site years ago.
bsky.app/profile/indyfr…
liberalarts.tulane.edu/anthropology/p…
dumisanitatiana retweeted
Option 1) build a factory that outputs infinite Nobel-level knowledge on tap that could save lives and ascend humanity to the next level
Option 2) preserve abandoned concrete parking lot eyesore
They chose option 2. 🤦🏻♂️
dumisanitatiana retweeted
2000 soldiers from El Salvador under the command of Bukele have done more for Haiti then $25 billion in "humanitarian aid"
Let that sink in.
dumisanitatiana retweeted
Yakitori shop in Japan: “We’ll turn away anyone who doesn’t follow the rules” (in Mandarin).
Red Note: Boycott!
The rules are just normal etiquette. The anger comes from the quiet admission that following them is somehow optional.
dumisanitatiana retweeted
BREAKING: Nasdaq 100 futures fall -1.5% after US CPI inflation comes in higher than expected.
Odds of a December rate hike are surging.
dumisanitatiana retweeted
Gemini 4 Carbon is now available for users, developers and enterprises!
We are heavily using Carbon internally at Google for hardest long horizon tasks, our teams have used it to help mathematically prove the core cryptographic protocol of Signal — and it’s topping independent benchmarks like CursorBench 4.0, Perplexity’s WANDR, Zapier’s AutomationBench, and more.
The team also had fun using Carbon to analyze images and technical drawings from humanity’s first crewed lunar lander, the Apollo Lunar Module. Not a single flown component ever returned to Earth. It recreated the 3D asset from those historic documents and generated this walkthrough video.
Read more: blog.google/innovation-and-a…










































