ls ~/blog
Blog
Notes from building with AI agents: what worked, what broke, and the workflows that came out of it.
- latest · · 13 min read
Building a Sandboxed AI Agent Infrastructure, Block by Block
The real order I built it in: agent + SDK first, then isolation, then Unix sockets, then gVisor, then a proxy manager, then RAG, then knowledge graphs. Each block added because the previous one exposed a problem.
-
Agent-First Applications: Building Software for the Person, Operated by the Agent
How treating an LLM agent as the first-class operator of a product turned a forty-step deployment into a conversation and made the end customer non-technical-proof.
-
Giving OpenCode Eyes: Extending Terminal AI Agents With Browser Tools
How adding browser automation and MCP tools to OpenCode transforms terminal-bound agents into full-stack collaborators that can see, test, and debug what they build.
-
Turning an Old Laptop into an Autonomous AI Coding Sandbox via Telegram
How to set up OpenCode on an old laptop, integrate opencode-telegram-bot, and build a secure remote sandbox accessible directly from Telegram.
-
Migrating to Astryx with AI: An Experiment
An honest look at migrating a React expense tracker to Astryx using AI: what we tried, what failed, and what we learned.
-
Position: Reiterating, or How AI Turns SDLC Into a Loop, Not a Line
One of the first things to disappear when you switch to AI-first development is the notion of a linear SDLC. In its place is something that looks less like a pipeline and more like a loop. And if you don't close the feedback loop, you're not automating the process, you're just automating the errors.
-
Ship Fast, Break Intent: How a Simple Markdown File Keeps AI from Wrecking Your Feature
If you want your developer to be 100x with AI, it's more like they'll do a year's worth of work in three days.
-
The Boolean and the Watcher: Two Sides of AI-Assisted Coding
A pair programming session in a forgotten AngularJS component taught me more about AI tools than any article had.
-
Why Debugging Is the Unsolved Problem in AI Coding Tools
Everyone's racing to build better code generation. The industry is quietly converging on a fix for what happens when that code breaks but most developers haven't wired it up yet.
-
Building This Portfolio With Antigravity: What We Tried, What Failed, and What We Learned
An honest look at building a Linux-terminal portfolio with Antigravity the wrong turns, the dead ends, and what eventually clicked.
-
The Double-Edged Sword: PR Reviews in the Age of AI
How AI-assisted coding is changing the rhythm of software development and pull request reviews.
// end of blog · 11 posts