Essays.
On systems, the boundaries between them, and how teams come to understand code they didn't write.
These notes describe products Nimbus no longer offers. They stay up as a record of how we got to suss.
Coding agents Who's on call for code nobody wrote? "The agent wrote it" doesn't settle who owns a change. Whoever approves and merges it owns it in production, and approvals, on-call rotations and incident reviews should reflect that. Coding agents When the model is down When the model behind your coding agents is unavailable, your team still has to know what its code does, so the record of it belongs in the repo where anyone can read it without AI. Reading code Senior engineers are readers now Agents made typing cheap, so most of the senior job is now reading code and judging it, and a promotion process that counts output is about to reward the wrong people. Coding agents Keeping the reasons behind a change Teams lose the reasons behind their code, and making engineers write decision records afterwards hasn't fixed it. Let a model draft the record from the threads where the decision was made, and have a person sign off on every entry. Coding agents LGTM is not a review Teams approve agent pull requests nobody read, and the green check has turned into theater. An approval should mean you read the change and you're on the hook for it, and a pull request too big to read should get a no. Reading code What a reviewer needs to see first Reading a pull request from the top of the diff was always a slow way to find out what a change does, and agents make it worse. Put the behavior change first, and generate it from the code. Coding agents The AI Landscape for Coding The coding tools teams use today, from autocomplete to agents to review bots. Coding agents Congratulations, You Gave Your Most Confident Junior Engineer a Megaphone Ownership, review and the senior engineer’s job when an assistant writes much of the code. Boundaries Finding the Unknown Unknowns A field rename passes every check, and days later a dashboard nobody knew about stops working. Boundaries Avoiding Integration Bugs: Beyond SDKs & Contract Testing Typed clients, contract tests and integration tests each catch a different kind of break. Distributed systems Serverless Isn’t Magic Most serverless trouble, from cold starts to the distributed monolith, comes from too much synchronous work and treating each function as its own service. Distributed systems Patterns for Distributed Systems Sagas, circuit breakers, API gateways and backends for frontends each fix one distributed systems problem, and each adds cost to a system that doesn't have it. Reading code Designing Dev Tools with Cognitive Load in Mind What a developer has to hold in their head to make one change, and how tools can help. Distributed systems The Hidden Costs of Just-In-Time Architecture Designing everything up front is slow, and designing nothing up front costs you later. Reading code From Code to Context: Feature Location Before you can fix a bug, you have to find the code for it, and that takes too long. Reading code Beyond Code: How Static Analysis Unlocks System Understanding From syntax trees to how components and services interact, dynamic languages included. Reading code The Secret to Building Better Software Teams tune the metrics they can count and miss how well they understand their system. Company 2024 in Review Our first year, from generated documentation to system maps. Company Case Study: From “This Should Be Simple” to “Ship It!” How one team used our earlier product to plan changes across dozens of services. Company Meet Nimbus Where we first introduced the company and the product we built at the time.