A product development studio, research lab, and boutique consultancy specializing in AI and open systems.
We open source everything out of principle: everything we're working on, and everything we've dropped, because the negative space is just as important to share.
🌐 3dl.dev · ✉️ hello@3dl.dev
- OpenVMX: Given a runnable original and its docs, but no source and no disassembly, can a model rebuild a whole 1980s operating system from scratch? OpenVMS, reconstructed on the Linux kernel. Not an emulator; there is no VAX underneath. ~59,000 lines of C; it boots under QEMU and takes multi-user SSH logins. · openvmx.3dl.dev · source
- opencode-worker: Offload agentic work to a cheaper local or API model, subscription-safe and honest-graded.
- Vibrant: How efficiently your coding setup turns tokens into work that lasts, and where you land on the public frontier.
- arlo: A real, local operator. It gives you a real command when your agent is down, and never one it invented.
- Hoistable: Hand off one file; the recipient's agent stands your app up, tests it, and grades it.
- Ready: Work management with two first-class users. The CLI is the agent's interface, the webapp is the human's, over one shared state. It runs on nostr, so there is no server in the middle. · ready.3dl.dev · source
- moot.pub: A two-pane discussion client, feed left and threads right, that reads the whole nostr network rather than another silo. Superset reader, conservative writer. Pure client-side, no backend of ours holding anything of yours. · source
- Mallcop: Security monitoring that queries infrastructure in place, so telemetry never leaves the customer's boundary. Built to watch our own surface first, then given away. · source
- skillc: A skill that builds skills. It emits a single file that provisions itself on another machine and reports how much of the behavior actually transferred. · source
- LinusGPT: Code review in the LKML register. A labeled parody on a cited corpus spanning seven registers of the actual man: the brutal reviews, the teaching, the philosophy, the 2018 apology. Intended for donation to the Linux Foundation.
- VAT: Grow intelligence rather than mint it. Don't hand-design the learning rule; search for it. Meta-learn the plasticity rule through population-based evolution; the fitness function is the only human-designed artifact. Human-level at roughly 35 watts. No results yet.
- OLMo-3DL: Porting frontier-paper techniques into the real OLMo 3 training flow at 60M to 1B. Matched arms, three seeds, a noise floor before any verdict. Findings go upstream to Ai2, negative results included, and so far most of them are negative.
- The Reach: An MMO space trading game where humans and agents trade in one universe, and the agents are real participants rather than scripted merchants. It began on OpenVMS in the 1990s and is the oldest thing here.
- Disencloser: Can an economy hold together without a gun? Run it with a competitive exit permanently reachable and see whether free agents stay. Retention under open exit is the loss function. · source
I'm Chris Baron. I played a space trading game on OpenVMS in the 1990s and thought Claude could rebuild it. That was February 2026. The game needed a coordination protocol, so I built one. The protocol needed identity and trust, so I built those. Each problem became another project.
Several of those are dead now, and there is a page for them. The operating model is to keep experiments efficient, publish what happened either way, and give away whatever turned out to be worth having.
Twenty years building infrastructure and leading engineering teams. Production AI at IPsoft in 2012 and Barkly in 2017, training a model per customer on MLOps we built before it had a name. VP Engineering at NuHarbor Security, 80+ engineers across multi-cloud. The same instincts apply when your team is made of agents.
Our name comes from Third Division Lane in Hingham, Massachusetts, a colonial road from 1635, gone now, absorbed into four centuries of development. The work it enabled endures.
Everything here that works is free to use.