Really enjoy the new editing tool from @imagine. The grounding capability from the multi-modal is insane.
Many robotic companies put Herman Miller into the office, but Tesla is one of the few places that put the office into the futuristic factories.
No battery left behind.
Our goal is for no Tesla battery to end up in a landfill.
Across our US operations, we’re recovering more than 90% of key battery materials like nickel & cobalt from cells that we recycle
More details in our 2025 Impact Report
Show more
In an era of human-AI symbiosis, our consciousness becomes the extension of the long memory and contextual refinery of a machine, while the machine mainly does the compute.
Six months ago, if GitHub Copilot was down, people would feel annoyed but get over it.
Today, if people cannot access a coding agent, they will stop working.
Models may come and go. The demand for compute will never go.
TLDR: If an agent is allowed to code, it could score almost perfectly on many benchmarks, even the one with AGI in the name.
Riding a 🌊 is at its best when it’s barely noticeable. It’s easier to hop on it when it’s small than to catch it when it’s big.
The icon.
Models view their world from a harness. How do humans view the universe?
If a change of harness can boost or degrade a model's performance, what about humans?
Perhaps how we view the world decides our potential, and many people are capped by their worldview.
Show more
Perfecting the overnight mode on Grok Build to keep the main agent context clean, dropping compaction from 3 times to 1.
Prompt:
"Overnight mode on: . The role of the orchestrator is to a) make high-level strategy decisions, b) make sure sub-agents deliver useful information and keep going, c) validate sub-agent results (fan out another sub-agent as a validator), and make the next move."
Show more
There are people who are extremely token-efficient at steering the model. They can work on tens or thirties of PRs at the same time and more, using only a modest amount of tokens while delivering good work.
In contrast, many people let the model run wild because it's frontier, with vague prompts so the model runs like a headless fly and tokens just blow up.
It reminded me that some professors are very good at directing labs so PhD students are extremely productive, while in other labs many PhDs spend years searching and still cannot graduate.
Show more
First thing in the morning, my kid was writing down the specs waiting to be built and even thinking about how to make a game balanced and not easy to cheat. Very interesting.
Entropy follows a predictable pattern—it wants you to lose until you give up everything. Don’t let it.
What a beautiful sight to see.
“Turn on overnight mode” is such a magical prompt to drive Grok Build to deliver the results in the morning.
“Turn on overnight mode” is such a magical prompt to drive Grok Build to deliver the results in the morning.
I was using my Grok Build; my kid was running hers.
When the digital world collapses to a single prompt, the next frontier is real-world robotics.
Aladdin is likely on the free tier plan because he could only make three wishes before reaching the usage limit.
Humans can't look everywhere at once, but FSD Supervised can
