These are the ideas I keep returning to in talks, workshops, and demos: how AI tools earn trust, what a useful harness actually does, and how teams can ship with more care and less ceremony.
The Merge: Your tests pass. Can your agent finish the job?
Cua AI co-founder Francesco Bonacci explains why developers need to check whether AI agents finish the job, even when the tools they use work correctly.
AI evaluationClaude Opus 5.5 for code review: More catches, different misses
We tested Claude Opus 5.5 against CodeRabbit's production reviewer to see which bugs it catches, which it misses, and what higher reasoning effort changes.
AI code reviewA software factory needs a review gate it can trust
Why agent-driven delivery still needs a visible, risk-based gate before code ships.
Product engineeringWhat would the last software engineer still need to do?
Why product judgment, system understanding, and ownership matter when agents write the code.
The MergeThe Merge: Why Building AI Chat Is Harder Than It Looks
A conversation with Assistant UI founder Simon Farshid about the hidden work behind AI chat interfaces, product judgment, and his coding-agent loop with CodeRabbit.
Developer workflowCodeRabbit CLI 0.7.1: improved performance and reliability
CodeRabbit CLI 0.7.1 brings faster local reviews, configuration validation, safer agent authentication, and CodeRabbit Skills for coding-agent workflows.
AI evaluationOpus 5 for code review: Cleaner actionable comments, noisier overall
Opus 5 produced CodeRabbit's cleanest actionable-comment stream, but caught fewer known issues and generated roughly four times the baseline's nitpicks.
AI code reviewBuilding an effective code-review harness
The twelve primitives around the model, explained through a code-review agent.
Agent systemsLoop Engineering: When is it worth the hassle?
What loops are, the primitives they need, and where they are worth the complexity.
AI code reviewWhat AI code reviewers catch — and what they still miss
A frank account of where automated review earns its keep and where human judgment still matters.
AI evaluationWhy building your own benchmark will get you ahead
A practical three-stage guide to evaluating whether an AI feature is ready to ship.
Developer workflowA field guide to pre-merge checks
How to build a check stack that protects quality without slowing the team down.
Climate-aware engineeringCarbon-aware CI: greener pipelines, same merge speed
Shift pipelines toward lower-carbon windows without making developers wait longer.