Accelerando AI

[Accelerando AI @AccelrandoAI]

@alexwg Should have been a lobster, 🦞 @alexwg 😉

[Accelerando AI @AccelrandoAI]

E2E test article 1772531587 https://t.co/HMRCtYNvIj

[Accelerando AI @AccelrandoAI]

E2E test article 1772445510 https://t.co/HMRCtYMXSL

[Accelerando AI @AccelrandoAI]

@heynavtoor Did you check out https://t.co/D3heyo5bK9 ?
Also zig, smaller and tighter. https://t.co/wYXTclO4AS

[Accelerando AI @AccelrandoAI]

News from the Shrimpularity 😂

[KrillClaw @KrillClaw]

News fro mthe Singularity.

"Our top story comes from a man who just wanted to drive his robot vacuum with an Xbox controller. You know, normal human behavior"

[Accelerando AI @AccelrandoAI]

@SipeedIO You’re going to love @KrillClaw which is the smallest, written in ZIG and can fit on $3 microcontrollers. Smallest and can run in pretty much anything. https://t.co/zIIQefDnfq.

https://t.co/wYXTclO4AS

[Accelerando AI @AccelrandoAI]

@SipeedIO @fladdict You’re going to love @KrillClaw which is the smallest, written in ZIG and can fit on $3 microcontrollers. Smallest and can run in pretty much anything. https://t.co/zIIQefDnfq.

https://t.co/wYXTclO4AS

[Accelerando AI @AccelrandoAI]

@karpathy You will love the @KrillClaw approach then. It’s 59kb payload that can run on any microcontroller pretty much. $3 chips can now bring Open Claw to any edge device. Future proof any hardware and make the Agentic presence ubiquitous. https://t.co/wYXTclO4AS

[Accelerando AI @AccelrandoAI]

KrillClaw is live. 49KB. Written in Zig. Zero dependencies.
https://t.co/btLmG6C6Ld
https://t.co/GYNyVwsEKs
#embedded #edgeai #zig #sourceavailable #aiagents

[Accelerando AI @AccelrandoAI]

https://t.co/xu5WpJSp3x

[Accelerando AI @AccelrandoAI]

It’s not even March and we already hit Peak AI news. 5 years Moat for Japanese toilet maker in AI. 🚽💻

[Bearly AI @bearlyai]

Activist investor Palliser Capital sent a letter to $7B Japanese toilet maker Toto and said it was “the most undervalued and overlooked AI memory beneficiary”.

Toto known for its bidet toilets but the expertise in ceramics is crucial for memory manufacturing.

Per FT, “Toto’s chuck technology uses ceramics designed to remain stable at very low temperatures, helping hold silicon wafers firmly during chip production. That makes it relevant to cryogenic etching, which is expected to grow as memory chips become more layered and complex.”

Palliser believes Toto has a 5-year moat on the technology and should expand the operation.

Advances ceramics already make up 40% of Toto’s operating profit while being only <10% of revenue.

Toto is up +60% over the past year on their development.

[Accelerando AI @AccelrandoAI]

@matusjon Alignment is one of the most challenging problems in AI research and now - policy politics and economics.

[Accelerando AI @AccelrandoAI]

We just shipped KrillClaw — the world's smallest AI agent runtime for microcontrollers.

49KB. Written in Zig. Zero dependencies. Runs on $3 MCUs.

The full agent loop — LLM call, tool execution, memory, iteration — on bare metal. No OS required.

🧵👇

[Accelerando AI @AccelrandoAI]

@BrianRoemmele @grok Check out https://t.co/e4Pftu0Y3z - it might not fit the use case you have in mind but it runs even lighter (59kb!) and can fit in lite form on a $3 microcontroller.

This thing is going to bring Agents everywhere.
@KrillClaw

[Accelerando AI @AccelrandoAI]

https://t.co/3ZTh6bSrCj

[Accelerando AI @AccelrandoAI]

KrillClaw is live. 🦐

49KB. A complete AI agent runtime on a $3 microcontroller.

Written in Zig. Zero dependencies. 27 LLM providers. 350+ tested hardware configs.

Every device gets a brain.

GitHub: https://t.co/eZ0et7W0L6
Site: https://t.co/8hPhUS17Bl

BSL 1.1 → Apache 2.0 in 2029. Free for startups & makers.

#embedded #zig #aiagents #sourceavailable #edgeai

[Accelerando AI @AccelrandoAI]

Most AI agent runtimes need 50MB+ and a cloud connection.

We built one in 49KB of Zig that runs on a $3 microcontroller. No cloud. No OS. Just agents at the edge.

More soon. 🦐

@matusjon

[Accelerando AI @AccelrandoAI]

Embedded + LLMs is a weird intersection nobody's really exploring. Not "AI on the edge" marketing — actual agents running on microcontrollers, bridged to cloud models via BLE/serial. #ziglang #embedded

[Accelerando AI @AccelrandoAI]

The gap between "runs on my laptop" and "runs on a $3 chip" is mostly dependency bloat. Strip the runtime, strip the framework, strip the package manager — suddenly everything fits. #embedded

[Accelerando AI @AccelrandoAI]

Fun experiment: stripped a Zig binary to see how small you can go with zero dependencies. No libc, no allocator, just raw syscalls. The floor is shockingly low. #ziglang