Notes on the Software Factory | Benedict Brady
benedict.dev · 1,404 words · saved by 2 readers
Notes from a talk on vibe coding: giving agents tools, speeding up the bottleneck, critics, memory, skills, integration tests, and the agent native stack.
← Back These notes are adapted from a talk I gave at the Ellipsis Labs office about vibe coding and the software factory. July 8, 2026 Give the Model Tools The number one piece of advice is always the same: give the models tools to check their own work. That said, they are becoming much better at this naturally. To the extent you can, you want to do things like this: Give the agent CLI access to deployments so that it can deploy and check for bugs Give the agent browser access for frontend (I think this comes out of the box now) Give the agent powerful machines to run…
saved by
related reading
- Building Effective AI Agents \ Anthropicanthropic.com
- Killing Coding Agent Slop With Adversarial Self-Playusetelos.ai
- Code was our medium for thoughtwattenberger.com
- Why We Built Our Own Background Agentbuilders.ramp.com
- Shipping at Inference-Speed | Peter Steinbergersteipete.me
- Building Effective AI Agents \ Anthropicanthropic.com
- Effective harnesses for long-running agents \ Anthropicanthropic.com
- Thoughts on slowing the fuck downmariozechner.at
- Arjun Virkarjunvirk.com
- Don’t Outsource Your Thinkingteltam.github.io
- The Age of Async Agents — Cognition's Walden Yan & OpenInspect's Cole Murraylatent.space
- Vibe engineeringsimonwillison.net