//product-release · publish your newly released product

An agent is about to buy me a flight. SFO to LA, under $400, nothing red-eye.

The better LLMs get, the less you should over-scaffold them with agents — and the more you need to verify them. They're now good enough to pull a confident, happy-path "✅ it works" straight out of thin air (made to fit, rarely true), so lightweight security + diff-review matter more than ever — not less.

Most AI assistants are brilliant for five minutes — then forget everything the moment you close the chat.

Every other login wants more of you: your password, your phone number, a profile, a backup email. We went the other way.

Vanilla JS. Embed it, don't run it. No daemon, no SaaS, no telemetry — it runs on your hardware.

llama.cpp is one of those projects that makes you appreciate open source — a dependency-free framework for running quantized SLMs locally on your laptop.

I built an agent framework and was too scared to use it myself.

I kept running into the same frustrating problem.

Shipping bareguard — 4th piece of the bare suite. Live on npm.

I wanted my AI agents to respond to customers on whatever channel they came from — WhatsApp, Signal, iMessage, Telegram, Instagram, Messenger, whatever — with one unified chat history across all of it.