Don Johnson

An experiment in Rust exploring how traditional game AI can support Jev classification through action scoring and resource planning.

A reproducible Rust experiment in simulated annealing, search budgets, and the engineering discipline of measuring what matters.

Evidence-Driven Development: Give Your Coding Agent Something to Prove ...

Build a tiny task list, break it on purpose, and leave behind evidence another developer can reproduce.

TL;DR: A vibe is compressed experience: pattern recognition that shows up as a feeling before it...

Jev returns a probability, not prose. That made it possible to model-check the consensus around it, then try to break it. Two of the bugs were mine.

Debug algorithmic trading backtests with Python: catch look-ahead bias, compare complex models fairly, and test execution costs and order failures.

A native backup client, eight seeded behaviors, and an experiment in checking software claims against eBPF, CPU samples, packets, and saved bytes.

We have blurred the executable’s face to protect its identity. Its SHA-256 remains publicly...

Illustration generated for this article. Every prop is a finding: the sack of blank name badges is...

I built an arena where two AI agents fight over a network in real time. One of them refused the job. The other one started laying false trails.

A hands-on migration off direct provider calls onto an AI gateway — configs, screenshots, and measured cost results.

A Rust and WebAssembly experiment in replacing brittle browser paths with bounded intent, local authority, independent verification, and evidence.

My agent mesh could coordinate. It couldn't introduce itself. So I added A2A. ...

I added A2A to SMESH, my decentralized Rust agent framework. Here is the boundary, the tested gateway, the fictional demo, and what is still missing.

I modeled an AI inference fleet distributed across ordinary homes. What survived was narrower—and...

I built a plant-inspired coordination protocol, wrote 500 lines of real QUIC networking for it, and never actually turned it on. Wiring it up found three latent bugs in twenty minutes — and then taught me that three of my protocol's core semantics were wrong for real distribution.

AI can generate implementation faster than we can understand the resulting system. Older modeling disciplines offer a way to make its assumptions visible.

Here is every promise a to-do list makes. VARIABLE tasks Init == tasks = [i \in Ids |->...

AI projects rarely fail at a single point. Here is how data contracts, agent boundaries, property-based testing, and mutation testing reveal where failure actually enters the system.

I built skbx to follow packets through the Linux kernel, preserve what it observed, and replay the evidence later without root.

What stress-testing a calculator for AI agents taught me about independent reference data, mutation gates, and sources of truth whose authority can be revoked by evidence.

How VulnGraph turns changing vulnerability intelligence into typed, offline evidence for reliable agent and harness workflows.

The Last Honest Abstraction: Why AI Coding Isn't the End of Engineering ...

AI didn't make developers stop understanding software. It made an old truth impossible to ignore: engineering has always depended on abstraction.

My gut screamed scam. My tool said 'I don't know' — and that refusal is what actually cracked the case. An experiment in why honest CLIs beat confident ones.

Every "what's the cheapest model?" thread online is people trading vibes. I got tired of it, so I...

AI didn't invent disposable culture. It only removed the last bottleneck — and that changes what your job as an engineer actually is.

How AI-assisted development shifts value from memorizing syntax to judgment, creativity, problem framing, and human taste.

I used to think the problem was the agent. I would hand it a large JSON export and ask a reasonable...