GGeorgeOct 1 · 4 min readWhat a Human Gate Looks LikeHuman gates sound like process theatre. But it's the opposite. Here's the actual mechanics, on a real feature. So, you be the judge. The setup You are building a feature. Let's say a session manager foystro.hashnode.dev
GGeorgeSep 24 · 3 min readThe Five Levels of AI-Assisted DevelopmentDan Shapiro's framework of five levels of AI-assisted development, borrowed from the autonomy levels of self-driving cars, is the most useful way to think about where your team actually is. The insighoystro.hashnode.dev
GGeorgeSep 18 · 4 min readWhy Vibe Coding Breaks at Scale: The Case for Harness EngineeringIn early 2024, the software engineering industry embraced a seductive new rhythm: "vibe coding." You open a chat window, describe an intent in plain English, watch an LLM generate 400 lines of code acoystro.hashnode.dev
GGeorgeSep 17 · 3 min readWhy the Harness Matters More Than the ModelThe same model weights that score 30% on a benchmark can score 95% with a better harness. That's not our line. It comes from Y Combinator's paper club on harnesses, and is the clearest statement aboutoystro.hashnode.dev
GGeorgeSep 10 · 4 min readHarness Engineering vs Vibe CodingWhen vibe coding, the agent generates, and you accept or reject after the fact. The decision happens in the prompt, invisible and unaudited. In harness engineering, the agent stops before it writes a oystro.hashnode.dev
GGeorgeSep 3 · 4 min readEvery agent session starts from zeroWe spent the last few months deliberating and then building software that runs on top of AI coding agents. The question that kept intriguing us: Why does every agent session start from zero? When a teoystro.hashnode.dev
