AI for software engineers
AI training programme · 7 modules · 4 Skilledge Hyper labs
How fast AI feels and how fast it actually is aren't the same. That gap is what we train against.
In METR's controlled trial, experienced developers using early-2025 AI tools were 19% slower on real tasks, while believing they'd been 20% faster. The tools have improved a lot since. The gap between how fast AI feels and how fast it is hasn't gone anywhere. Our job is getting a team from "almost right" to shipped and correct.
At a glance
- Who it’s for
- Software engineers, tech leads, engineering managers, new joiners getting up to speed on a codebase.
- Tracks
- Engineering
Modules
- Context engineeringRepo instruction files like AGENTS.md, and giving an agent what it needs and nothing it doesn't.
- Agentic workflowsPlan, implement, review. Handing a ticket to a coding agent and getting back a PR you can review.
- Reviewing AI codeSpotting "almost right": wrong edge cases, invented APIs, tests that pass for the wrong reason.
- Tests as guardrailsWriting the tests first so the agent has something to fail against.
- Legacy codeReading, documenting and refactoring code nobody on the team wrote.
- SecuritySecrets, made-up package names, licence risk, and what never goes into a prompt.
- Measuring the changeDORA-style delivery metrics, so the team knows whether AI helped.
Skilledge Hyper lab exercises
Lab designs: what a cohort does in Skilledge Hyper, on mock data.
- Unfamiliar codebaseFix a bug in a mock service nobody in the cohort has seen, using an agent. Graded on the PR, not the time taken.
- Clock yourselfThe same task with and without AI, timed by Hyper. Learners compare the real numbers with their own guess.
- Write the AGENTS.mdSet up a repo so an agent behaves well in it, then prove it with three tasks.
- Onboarding trackNew joiners use the agent to learn a codebase faster. Our full-stack onboarding for a leading general insurer: client-reported time to productivity of 3 to 4 weeks.
Tools covered
GitHub Copilot (agent mode and coding agent), Claude Code, Cursor, OpenAI Codex, and GitHub Agent HQ under admin control.
Sources
Related programmes
- Enterprise tools with AIAI training
- AI orchestration and agentsAI training
- AI for QA and testingAI training
Want this for your team? Tell us who it’s for and what they need to do differently after it.