# Ryan Zidago's Blog

> Product engineering and AI/LLMs

## Posts

- [AI code reviewers should provide executable evidence](https://ryanzidago.com/posts/ai-code-reviewers-should-provide-executable-evidence.md): Why AI code reviewers should validate findings with executable and reproducible evidence.
- [Do we need referential integrity for prompts?](https://ryanzidago.com/posts/do-we-need-referential-integrity-for-prompts.md): What happens when a user-authored prompt keeps referring to an application entity after that entity has changed?
- [Parallelising guardrails for faster feedback loops](https://ryanzidago.com/posts/parallelising-guardrails-for-faster-feedback-loops.md): How to run independent Mix quality checks concurrently while keeping their output readable and their failures actionable.
- [Each test should own its setup](https://ryanzidago.com/posts/self-contained-tests-in-elixir.md): Why I prefer explicit, self-contained Elixir tests over shared setup callbacks.
- [How to build software engineering guardrails: An example with Bylaw](https://ryanzidago.com/posts/bylaw-a-concrete-example-of-software-engineering-guardrails.md): How to turn engineering policies into deterministic guardrails by inspecting code, database schemas, and composed Ecto queries with Bylaw.
- [AI coding agents need software engineering guardrails](https://ryanzidago.com/posts/ai-coding-agents-need-software-engineering-guardrails.md): Why deterministic software engineering guardrails are more reliable than asking AI coding agents to remember every rule.
- [Reducing Elixir worktree setup time by 83%](https://ryanzidago.com/posts/reducing-elixir-worktree-setup-time-by-83-percent.md): How to reuse Mix build artifacts across Git worktrees safely, with a reproducible Phoenix benchmark.
- [What I learned shipping a visual dashboard builder in 3 days](https://ryanzidago.com/posts/what-i-learned-shipping-a-visual-dashboard-builder-in-3-days.md): Notes on using multiple LLMs to research, design, implement, and recover a UI-heavy feature under time pressure.
- [Incrementally enforcing code quality rules with coding agents](https://ryanzidago.com/posts/incrementally-enforcing-code-quality-rules-with-coding-agents.md): A practical strategy for adding linting and code quality rules to a legacy codebase using coding agents.

## More

- [About Ryan](https://ryanzidago.com/about/)
- [RSS feed](https://ryanzidago.com/posts/feed.xml)