← Industry Signals
Video·David Ondrej·Agentic CodingFeatured

Matt Pocock's Agentic Engineering Workflow

A working developer's view of AI coding as a structured practice, not a one-shot generation trick.

Signal summary

This video matters because it shows the behavioural layer of agentic engineering: how experienced developers give context, shape tasks, inspect outputs, correct course, and maintain control over a fast-moving AI-assisted workflow. The interesting part is not simply that code gets generated. The interesting part is how the human operator structures the work so that generation remains useful.

Why it matters

A lot of AI coding discourse treats the model as the main event. This signal points somewhere more durable: the workflow is the product. Good agentic engineering depends on clear task framing, tight feedback loops, review habits, local context, and explicit boundaries around what the agent may change.

LGI reading

This maps directly onto HKE's broader pattern: agents are more reliable when they operate inside explicit containers. In software, that container is a mix of repo context, tests, specs, diffs, review checkpoints, and developer judgement. In organisational work, the same pattern becomes envelopes, ledgers, evidence packs, and approval gates.

Use this if you are thinking about

AI coding adoption, developer productivity, Claude Code-style workflows, engineering operating models, and how to turn individual AI usage into repeatable team practice.

Related LGI concept

Agentic Coding

agentic codingdeveloper workflowClaude Codesoftware engineering