Glossary
The terms this site keeps arguing about — defined plainly, then linked to the essays that make the case.
- Forward-Deployed Engineer · FDEA forward-deployed engineer (FDE) is a software engineer who embeds directly with a customer — building, integrating and shaping a vendor’s product inside the customer’s environment, rather than working on the core product from headquarters. The role was pioneered at Palantir and, through 2025–2026, became one of the most-hired titles at OpenAI, Anthropic and AWS.
- Meta-HarnessA meta-harness is the harness around your harness: the coordinating layer that sits above any single AI-agent loop and manages memory, review, orchestration and human judgment across many agents and many runs. If a good harness turns one engineer into a one-person factory, the meta-harness is what keeps a floor full of those factories building the right thing, remembering what they did, and checking each other’s work.
- AIDLC · AI Development Life CycleAIDLC (the AI Development Life Cycle, sometimes written “AI DLC”) is the reimagining of the software development life cycle for a world where AI agents do most of the building. It keeps the familiar stages — intent, build, review, ship, operate — but moves the human effort out of typing code and into specifying intent, steering agents and reviewing output at speed.
- Claude Code HarnessA Claude Code harness is the scaffolding around the model — the tools, memory, context management, permissions and feedback loops — that turns Claude Code from a chat window into a reliable engineering agent. In practice it is the harness, not the model, that decides whether an agent ships working code.