Posts

A Chinese article argues CLI is becoming the AI plugin format. I've been living this for months with 442 tools. The article is right about CLI. It's wrong about what makes CLI work.

Seven PyPI releases of a CV generation tool in one afternoon taught me that template-guided synthesis lives and dies by what the template already contains.

A 208 MB log, 59,356 retries, and zero LLM calls. A debugging story about what happens when the symptom lies about the cause.

A simple rule for keeping AI agent skill systems coherent: if two skills fire on the same trigger, merge them. Different trigger, different skill. No exceptions.

I left an AI coding pipeline running overnight with 21 monitoring cycles. 5 features merged, 10 specs dispatched, 3 root causes found. Here's what worked, what broke, and the quality of the output.

When choosing CLI vs MCP, pick the one you can undo. CLI wraps into MCP cheaply. MCP does not unwrap.

ANTHROPIC_API_KEY vs ANTHROPIC_AUTH_TOKEN — how a single wrong environment variable made an AI coding pipeline silently fail for hours, and the debugging journey that found it.

Anthropic shipped a hosted agent platform. Its architecture looks familiar. Here's what a solo builder can learn from how they decoupled the brain from the hands.

Karpathy's LLM Wiki pattern is a good starting point. Here's what changes when you run it for real — enforcement over convention, decay over growth, and knowledge that fires without being asked.

Using cell biology naming not as metaphor but as engineering manual — how mTOR's biology predicted circuit breakers, autophagy, and negative feedback loops before we designed them.

A 96% rejection rate that was actually a 96% false positive rate — how a monitoring blind spot turned a productive overnight batch into apparent failure.

The architect writes the tests. The implementer makes them pass. No prose specs, no circular validation.

Most advice about making CLIs agent-friendly is just good CLI design. Only four principles are actually agent-specific.

I benchmarked four AI coding harnesses on 12 tasks using the same model. The harness barely matters for correctness — it's all about the model.

Smart model plans, cheap model builds. The pattern everyone's converging on for AI coding agents — and the piece nobody's shipped yet.

mtor dispatched a coding task to improve itself — the tool that sends work to AI agents was improved by an AI agent.

Analysis of Simon Willison's blog corpus reveals AI coding agents crossed a reliability threshold in late 2025 and are now reshaping software engineering.

I turned the seven patterns into a zero-dependency Python library. Then I added MCP bridge support. Here's what I learned about the gap between patterns and code.

Three independent authors converged on nearly identical patterns for CLIs that AI agents invoke. Here's what they agree on, what's missing, and why nobody has built a framework for it yet.

When a dispatcher and an executor share a name, you don't have a naming problem. You have an architecture problem.

Multi-engine search agreement is not primary-source verification. A cautionary tale about hallucinating reference content from consistent secondary summaries.

Willison says CLIs beat MCP. Cloudflare says server-side code mode beats both. They're both right, because they're answering different questions.

I designed an elegant framework install for my personal AI system. Then I listed the hard problems and shipped a three-hour cleanup instead.

A day of reading skill repositories taught me less about the skills themselves than about how much I'd missed of the surrounding ecosystem.

Five skill repositories, a day of reading code, and a significant correction I had to make the same afternoon.

Naming as a design constraint: if a tool resists a name, the tool needs redesigning, not the name.

I renamed one concept across 130 files. The pain crystallized into a tool that will do the next rename in minutes.

Most projects pin dependency versions to avoid breakage. We automated the opposite: daily upgrades with automatic rollback.

We blamed the LLM for a 54% task failure rate. The real culprit was seven layers of dispatch infrastructure between intent and execution.

Compliance-first governance produces paperwork. Design-first governance produces systems you can actually explain to a regulator.