Ask HN: What are you working on? (February 2026)

Trying to use ESNs as a random projection for audio data and potentially rendered text data for some AI workflows. Seeing it I can use the echo states running both forward and backward through the data as a holographic representation which would act as a temporally dense token for potential use in LLM or audio encoder inputs.

An open source runtime governance engine for AI https://github.com/jnamaya/SAFi

I'm working on a website that lists veterinary practices owned by private equity or large corporations to help people make more informed decisions about where they take their pets. It started as a small passion project after our dog (who was sideswiped by a car) died at the hands of a vet practice recently acquired by private equity. We were billed over $13k for 2 days of care where his diagnosis and the opportunity to treat it was missed, there was zero continuity of care, no medical leadership and predatory billing practices.

The site has become quite a hit and gets thousands of unique visitors each day. https://www.privateequityvet.org/vet-list

vibebin is a self-hosted platform for running persistent, isolated AI coding sandboxes on a single VPS using Incus/LXC containers:

https://github.com/jgbrwn/vibebin

It automates installing and managing Incus, Caddy, and SSHPiper, provides a TUI for container lifecycle and quick actions, a web admin (built/compiled on the container) for toggling and updating AI coding tools (Shelley, Claude Code, OpenCode, etc.), and a background sync daemon that keeps Caddy routes and container metadata in sync.

Each container exposes coding tool web UIs on isolated ports and supports direct SSH/VS Code Remote access, so you can run multiple independent coding agents against real project files without exposing your local machine.

The project emphasizes simplicity and recoverability for running agents locally: containers are persistent, optionally routed via reverse proxy with basic auth, and tracked in an SQLite DB so setups auto-heal after restarts. It’s written in Go, includes an install script for one-line deployment, targets modest VPS specs (4–8GB RAM recommended), and bundles helpers for DNS and provider automation.

Ideal if you want a lightweight, opinionated way to host multiple isolated AI dev environments on your own server instead of relying on hosted agent platforms.

That’s a great idea and I wonder if you’d consider including the UK. My vet got taken over and now they keep spamming me with “offers” - so I no longer know if my dog really needs a checkup or not.