Skills package capability. Harnesses determine how that capability performs in production. This article explores three nested control loops and shows how observability becomes governed adaptation, with evidence passing through evaluation and policy before it can safely steer execution.
The scaling law is breaking: more software no longer requires more engineers. For most of the history of software engineering, scaling software meant scaling engineering teams. If you wanted more software, you hired more engineers. More products required more teams, and more features required more developers. More systems required more
Skills provide capability. Workflows direct the work. Harnesses control execution. Factories make the whole system repeatable. This article proposes the architectural boundaries between them and extends the factory model beyond software delivery into enterprise work.
The difference is between an LLM-powered feature, an orchestrated workflow, and a system that can act with bounded agency.
Everyone is talking about agent loops, but most are just retries, workflows, or DAGs with model calls. Real loops add another dimension: feedback, state, judgment, and abstraction gain. They do not just repeat work. They climb.
Every major shift in software changes the unit of engineering. I believe we're watching it happen again. Software engineering has always evolved through abstractions. We rarely notice the transition while we're living through it. We optimized assembly language right before compilers made hand-tuning irrelevant. We perfected
Most organizations don't have an AI capability problem. They have an AI absorption problem. AI doesn't fix broken processes, unclear ownership, or weak workflows. It amplifies them. The winners won't have better models. They'll have better operating systems for turning intelligence into outcomes.
HydraFlow as a live test of FLUID systems and the operating patterns that emerged from running one.
Most teams are building AI agents. Few are getting real value. The difference is not better prompts, it’s better systems. Here’s how to land your first agentic AI use case using a Vibe-to-Value loop, with evals, guardrails, and measurable outcomes.
AI makes features cheap, but value comes from outcomes. Most AI projects stall because they lack orchestration, governed autonomy, and evaluation. The shift is from building software to operating decision systems that improve over time.
Software was never deterministic, we just couldn’t afford to explore alternatives. AI makes variation cheap, shifting the focus from writing code to validating outcomes. The future is probabilistic creation constrained by systems that ensure reliable results.
Last year I explored the idea of FLUID software as a design philosophy. This piece explores the next layer of that shift: how AI changes the software delivery model itself. For decades, software engineering assumed one fundamental constraint: Humans had to write the code. Entire design philosophies grew from that