Software Engineering
-
What Sixteen Models Did With the Same Eight Principles
313 trials across sixteen models, one Rust scenario, the same eight engineering intents. Lift ranged from 0.00 to 0.89 depending only on which model read them — and the strongest model in the set was nowhere near the top.
-
The Spec for the Work Isn't the Spec for the Thing
I looked at GitHub's spec-kit when it launched a year ago, found it incomplete, and carried on with my own line of work. Coming back to it, I can finally name what was missing: it builds a project spec, and I keep wanting a product spec.
-
Modelling Engineering Intent Made My Guidance Measurable
I'd been trying to test my agent guidance for a couple of years, and every attempt died in the same place. What changed wasn't the harness — it was modelling engineering intent as structured records, which made the guidance specific enough to measure at all.
-
Seven Coding Agents, One Space Invaders Clone
Anthropic is splitting their subscription on June 15, and unattended Claude Code is off my menu. I ran the same Bevy build through seven coding-agent profiles to see what could replace it. Three of them produced playable games. Four shipped binaries that crash on launch. The cost spread between the working ones is six-to-one, and the gameplay is indistinguishable.
-
Every Codebase Is an Uncompiled Knowledge Base
Karpathy's LLM Knowledge Base architecture is a powerful contribution, but it draws a false boundary between code and knowledge. Your codebase is the largest uncompiled knowledge base in your organization — and agents can now extract what's inside it.
-
Code Is Knowledge
Code isn't just instructions for machines. It's a record of how an organization thinks — about customers, domains, risks, and trade-offs. We've always known this. Now we have tools that can make it explicit.
-
GenAI Didn't Kill Rigor — It Moved It
The old bottleneck was implementation. The new bottleneck is ambiguity. And the rigor that used to live in coordination now needs to live in specification, context, and evaluation.
-
Know Your Principles
Comprehension debt isn't just about AI-generated code you don't understand. It's about not being able to articulate the principles that make your codebase navigable in the first place.
-
Finding Your True Worth
Your value isn't in typing code—it's in what you know. In a GenAI world, that tacit knowledge is more precious than ever.