If you want it, build it 🦊 ⏳ ⚔️ ⚛️ ▲ Field Eng @SpaceXSI prev: @cursor_ai @vercel
NYC
Joined December 2016
- Tweets454
- Following770
- Followers436
- Likes15K
If you're applying for jobs, @bot just became your new best friend 🤝
Big ship from the team today.
Filling out repetitive info for your Bot has been a huge source of friction. It just got *much* better.
2
158
Ah yes, Notion - the GPU company
Notion AI lead @sarahmsachs argues that a lot of what runs on GPUs today should actually run on CPUs:
"We have a strong perspective at Notion that a lot of things should be served on CPU that are served on GPU. There’s a lot of benefit and determinism in executed code."
"Agents don’t need LLMs to reinvent that process every time, and that’s super token exhaustive, it’s super slow, and there’s no observability or determinism in it."
"Giving the agent a local file system to store local thoughts, to write short Python scripts, that’s really important when there’s so much information. That has been probably one of the most important tools outside of edit and search that our internal agents rely on."
@NotionHQ
1
2
135
Christian retweeted
everything i know about managing agents i learned from the amazing programmers and computer scientists that came before
constraints in your codebase are freeing for both humans and agents! this has always been true, but if you were a small dev team you likely never had to think that deeply because you could trust your fellow programmers to write code and review well
but large companies have had to solve this problem since even before agents. because before you had agent slop, you had human slop. lots of engineers wrote bad code out of necessity (eg backend engineer forced to write frontend or reverse). the solution to this was constraints: lint rules, smarter compilers and diagnostics, high quality tests, investments into observability, and so on
the arrival of agents just means that big company problems are now everyone's problems. the good news is that none of this is really that novel: it's just good engineering
youtube.com/watch?v=GqmsQeSz…
compute go brrrr
Next week, @SpaceX will be bringing online 220,000 Nvidia GB300 GPU (worth $15+ billion), with another ~$15+ billion in November, and then potentially another $15 billion in December, for a possible total of 660,000 GB300s coming online in Q4 alone.
Upwards of $50 billion of GPUs.
(below is an old pic of SpaceX's new Minihard GPU facility, which is where those 220,000 Nvidia GB300 GPUs will be coming online, the same number as the other two buildings, but in a much denser configuration, allowing for a significantly smaller overall building footprint)
2
424
I'm not the biggest fan of these "I gave both models the same prompt" benchmarks. Often the model that does more (takes more turns) wins. 🤷♂️
The model that looks better here is often dramatically different than which is legitimately most useful day-to-day.
First Opus 5.5 vs GPT-6 Astra test is 3D.
One prompt, Blender only, all procedural. Render the 10-second shot, and record your own build timelapse.
Opus 5.5: 35 min, 199.6k output tokens, about $13.3 in API terms.
GPT-6 Astra: 28 min, 56.6k output tokens, about $14.5 in API terms.
Feels like a worthy rival, and honestly I am glad about that. From what I watched, Opus juggles way more at once and is faster overall. Astra is still good imo.
1
286
