Uploaded January 2026 | Updated September 2026, 1 week ago
Apologies for the audio quality in the one - I tweaked some OBS settings I should not have tweaked, and it sounds a lot more blasted than usual. Keeping this here until I find time to re-record.
Discover Ralph Wiggum, a devilishly simple approach to AI coding agents that uses just a for loop instead of complex orchestration systems. I'll show you the exact bash script I use to let Claude work through my entire backlog overnight, how to structure tasks as a PRD, and why keeping changes small with robust feedback loops (TypeScript, tests, CI) is the key to actually shipping working code. This technique mimics how real engineers work: pick a task, complete it, commit it, repeat. With models like Claude Opus 4.5 being so capable now, simpler approaches like Ralph are finally viable and honestly more intuitive than the multi-phase planning systems we were using just months ago.
11 Tips on using Ralph Wiggum: https://www.aihero.dev/s/r-wiggum-article
Original Ralph Article: ghuntley.com/ralph
Anthropic Article: anthropic.com/engineering/effective-harnesses-for-long-running-agents
Join my newsletter on AI Hero:
https://www.aihero.dev/s/y-newsletter
Follow Matt on Twitter
twitter.com/mattpocockuk
Join the Discord:
https://aihero.dev/discord
Apologies for the audio quality in the one - I tweaked some OBS settings I should not have tweaked, and it sounds a lot more blasted than usual. Keeping this here until I find time to re-record.
Discover Ralph Wiggum, a devilishly simple approach to AI coding agents that uses just a for loop instead of complex orchestration systems. I'll show you the exact bash script I use to let Claude work through my entire backlog overnight, how to structure tasks as a PRD, and why keeping changes small with robust feedback loops (TypeScript, tests, CI) is the key to actually shipping working code. This technique mimics how real engineers work: pick a task, complete it, commit it, repeat. With models like Claude Opus 4.5 being so capable now, simpler approaches like Ralph are finally viable and honestly more intuitive than the multi-phase planning systems we were using just months ago.
11 Tips on using Ralph Wiggum: https://www.aihero.dev/s/r-wiggum-article
Original Ralph Article: ghuntley.com/ralph
Anthropic Article: anthropic.com/engineering/effective-harnesses-for-long-running-agents
Join my newsletter on AI Hero:
https://www.aihero.dev/s/y-newsletter
Follow Matt on Twitter
twitter.com/mattpocockuk
Join the Discord:
https://aihero.dev/discord
