I build things. Mostly software. Increasingly things that make other software. Occasionally things with circuit boards, motors, and questionable power requirements.

I'm an engineer at 2389 Research, an applied AI lab in Chicago, where we're exploring what happens when AI agents become collaborators in the software development process rather than just another tool in the toolbox.

I've been building things for the internet since before GitHub existed. I'm still not entirely convinced we've figured out how computers should work.

At 2389 Research, we're building tools, conducting experiments, and trying to make AI-assisted software development more reliable, understandable, and useful.

Some of the things I've been working on:

• Tracker — Orchestrating multi-agent software development workflows with dependency graphs, human checkpoints, error recovery, and enough instrumentation to figure out what the agents actually did.
• Dippin — A language for expressing AI pipelines. Because eventually every sufficiently complicated collection of prompts becomes a programming language.
• Agentic development infrastructure — Context management, long-running coding agents, evaluation, observability, and the surprisingly difficult problem of making nondeterministic systems behave predictably.

I'm especially interested in what happens when we stop treating LLMs as chatbots and start treating them as components in larger computational systems.

My interests have a habit of escaping the computer.

• Embedded systems & electronics — Microcontrollers, sensors, PCB design, and building hardware from first principles.
• Audio engineering — Networked audio, high-fidelity digitization, and the occasional attempt to make vintage audio equipment participate in the modern world.
• Local AI — Open-weight models, Apple Silicon, inference performance, and running useful coding agents without renting somebody else's GPU.
• Retrocomputing & reverse engineering — Old architectures, lost software, obscure instruction sets, and figuring out how people made computers do impressive things with almost no resources.
• Home automation & infrastructure — Networks, self-hosted services, distributed sensors, and making physical spaces programmable.

I like projects that require learning something new, especially when several disciplines collide.

Before 2389, I worked on software at General Galactic, including:

• Galactic Camera — An experiment in cameras, identity, and the internet.
• Floppydisk.xyz — An NFT experiment inspired by the aesthetics and culture of early computing. Technical write-up.
• EIP-4361 tools — JavaScript tooling for Sign-In with Ethereum.

Go back far enough and you'll also find Python, Django, Ruby, JavaScript, APIs, distributed systems, and a respectable amount of web-development archaeology.

Some of my oldest GitHub projects are practically fossils. I'm rather fond of them.

• How do we make autonomous software systems observable, correctable, and trustworthy?
• What does good software engineering look like when much of the code is written by agents?
• How much useful intelligence can we run on hardware we actually own?
• What interesting ideas did we abandon in computing's history?
• Can we make the physical world a little more programmable without making it unnecessarily complicated?

And, occasionally: Surely someone has already built this?

📍 Chicago / Colorado

🌐 clintecker.com · 🧪 2389 Research · ✉️ clint@clintecker.com

he/him · Curious about almost everything.