AI coding workflows

Workflows that matter.

Agent choice gets easier when you start from the job. These are the recurring workflows worth testing before a team settles on a tool.

Greenfield development

Building new software where there are no legacy users, contracts, or behaviours to preserve.

Spike

Writing quick throwaway code to test feasibility, compare options, or expose unknowns.

Prototyping and MVP

Creating a minimal version of a solution to test a product, technical, or workflow hypothesis.

Maintenance

Keeping existing systems running through small updates, dependency work, operational fixes, and routine changes.

Refactoring

Restructuring code without changing behaviour, while preserving tests, contracts, and user expectations.

Performance optimisation

Improving speed, memory usage, latency, cost, or throughput with evidence from measurement.

Rewriting

Replacing old code with new code while managing feature parity, migration risk, and rollout.

Porting

Moving code to new platforms, runtimes, frameworks, or programming languages.

Debugging

Tracing issues through existing code, logs, tests, reproduction steps, and runtime behaviour.

Bug fixing

Diagnosing defects, making targeted changes, and proving the fix with a focused verification path.

Testing

Writing unit tests, integration tests, regression tests, or repairing failing test suites.

Code review

Evaluating code for correctness, quality, maintainability, security, standards, and missing tests.

Documentation

Writing or updating technical docs, comments, READMEs, runbooks, API notes, and migration guides.

Infrastructure

Managing deployment, configuration, environments, pipelines, and infrastructure as code.

Legacy code comprehension

Understanding unfamiliar, old, under-documented, or high-risk codebases before making changes.

Learning

Learning a new programming language, technology, tool, framework, or technique through guided practice.