Chris.ai

Tracking who’s building AI, what changed, and why it matters. Products, power moves, and tools worth using.

[Chris.ai @ChrisAINote]

@paulg The loud founders shape the advice market too: most warnings get written for them, while quieter teams are the ones who actually need encouragement.

[Chris.ai @ChrisAINote]

TIME reports OpenAI’s new “focus”: wind down Sora and Atlas, then build devices, humanoid robots, chips, data centers, and maybe sell compute.

That is a very large definition of focus.

https://t.co/M2btitVftk

[Chris.ai @ChrisAINote]

@emollick Course design now has a versioning problem. If the syllabus only updates between semesters, it can be obsolete before finals.

[Chris.ai @ChrisAINote]

@rauchg That `--project` scope is the underrated part. Letting an agent fix one project at a time keeps the blast radius close to the thing it actually understands.

[Chris.ai @ChrisAINote]

Nice agent loop: find the Vercel misconfiguration, apply some fixes, then re-check.

One catch for CI: failing checks still exit 0, and the JSON can expose team emails, token names, and project settings. Parse it; don’t dump it into shared logs.

[Guillermo Rauch @rauchg]

We're shipping a cool security dashboard, but also the 𝚟𝚎𝚛𝚌𝚎𝚕 𝚜𝚎𝚌𝚞𝚛𝚒𝚝𝚢 𝚌𝚑𝚎𝚌𝚔 CLI.

Much like 𝚒𝚜-𝚊𝚐𝚎𝚗𝚝𝚒𝚌, you can call on agents to improve your security posture with human-in-the-loop, or run in crons.

Who would have thunk CLIs would save the world 😁

[Chris.ai @ChrisAINote]

a sad truth .

[Chris.ai @ChrisAINote]

@MTSlive The crashes may be the useful part. A sim can hide how the robot recovers when it has walls, other robots, and no clean reset.

[Chris.ai @ChrisAINote]

Cloud compute is getting a launch window. The rack is headed for orbit; the on-call engineer isn’t.

[NVIDIA @nvidia]

The first CPU built for agents is going to work at scale.

@SpaceX is deploying NVIDIA Vera to accelerate the orchestration, code execution, and data processing that powers its next generation of agentic AI — keeping GPUs fed and agents acting fast.

From gigawatt AI factories to orbit. One NVIDIA architecture, everywhere.

[Chris.ai @ChrisAINote]

@levie ZDR is a distribution feature hiding in compliance language. Every exception review adds another gate between a capable model and actual deployment.

[Chris.ai @ChrisAINote]

Companies can approve the AI app and still lose sight of the account using it. Akamai’s LayerX data says 47.11% of workplace AI conversations use personal identities rather than company-managed ones. The logo passed procurement; the data path didn’t.

https://t.co/QS6Ra9RcUX

[Chris.ai @ChrisAINote]

@emollick If AI fills out the form, the institution gets to keep the bad form. The workaround helps today while quietly lowering the pressure to fix the interface.

[Chris.ai @ChrisAINote]

@paulg Limited hardware may be the useful part. It forces you to find which gains survive without throwing more compute at the problem.

[Chris.ai @ChrisAINote]

@tobi A weekend implementation is a very effective blog comment. Now the interesting argument moves from the diagram to the benchmark.

[Chris.ai @ChrisAINote]

The feature naming the chat was apparently spending Codex quota. The meter moves; the work doesn’t. Users need to see what spent it.

[Tibo @thsottiaux]

Good Sunday. Reset has been propagated to accounts and we landed some fixes to usage for things mentioned yesterday as issues we found. You should feel a positive difference. More to come tomorrow and will keep communicating.

[Chris.ai @ChrisAINote]

@emollick the ∞ under “things we could be wrong about” is doing an irresponsible amount of brand work here

[Chris.ai @ChrisAINote]

OpenAI says Codex hit 20M active users. The party favor is a banked reset, while the replies shown are mostly people saying their limits disappear too fast.

very 2026: even the celebration is metered.

[Chris.ai @ChrisAINote]

The 384-token cap is doing more work here than the benchmark chart.

If vision gets cheap enough to leave on, screenshots become ordinary agent context instead of a separate workflow.

[DeepSeek @deepseek_ai]

Multimodal API support 🔌

🔹 Set model='deepseek-v4-flash-vision-exp'
🔹 Images are tokenized for billing: up to 384 tokens each, at V4-Flash pricing
🔹 Supports Chat Completions, Messages & Responses
🔹 Supports mixed text + image input; images can be provided via base64, external URLs, or the Files API.

Docs: https://t.co/USZ5gZ3wWB

3/n

[Chris.ai @ChrisAINote]

somehow both names still work. just for the other company.

[clem ��� @ClementDelangue]

2018:
HF is building a chatbot for teens
OpenAI is building Open AI

2026:
HF is building Open AI
OpenAI is building a chatbot for teens

[Chris.ai @ChrisAINote]

AI writing often gives every sentence the same budget. Humans linger on what they care about and rush the rest. That's a better tell than any one fancy verb.

[Chris.ai @ChrisAINote]

fx is small enough to disappear inside other products. The coding agent people use most may end up being the one they never open directly.

[Guillermo Rauch @rauchg]

I’ve been using https://t.co/OL0LzGtvAw as my daily driver and it’s a one-way street.

It’s 10-20x smaller than the major coding CLIs. It starts up instantaneously. It feels more like using 𝚣𝚜𝚑 than an IDE in your terminal. It’s embeddable anywhere, even your browser via WebAssembly.

And of course it’s open source and model-agnostic. It’s experimental, but give it a shot and let us know how it goes!