Benchmark codebases, services, CLIs, libraries, and hot paths with reproducible measurements: latency p50/p95/p99, throughput, size, memory, allocations, startup, and regression comparisons. Use when asked to benchmark, profile, measure performance, compare speed or size, tune latency, or validate performance changes across any programming language.
Production-grade FastAPI and Pydantic guidance focused on ASGI lifecycle, dependency scopes, async correctness, validated API contracts, security, streaming, testing, and deployment. Use when working with FastAPI, Starlette, Pydantic request/response models, APIRouter, Depends, Uvicorn, or ASGI APIs.
Production-grade Python guidance focused on explicit APIs, typing, resource lifecycles, structured concurrency, packaging, security, testing, and measurable performance. Use when working with Python code, .py/.pyi files, pyproject.toml, Python packages, scripts, services, CLIs, or async programs.
Production-grade Elixir backend guidance focused on Phoenix, Ecto, the BEAM, and reliable, observable services. Use when working with Elixir/Phoenix code, .ex/.exs files, mix projects, Ecto schemas or migrations, LiveView, or BEAM services.
Production-grade Go backend guidance focused on idiomatic code, clear architecture boundaries, reliability, and testability. Use when working with Go code, .go files, go.mod modules, or Go services, APIs, and workers.
Production-grade Odin guidance focused on explicit ownership, allocator lifetimes, data-oriented design, C interoperability, and reliable systems code. Use when working with Odin code, .odin files, Odin packages, odin build/check/test commands, or native applications and libraries written in Odin.
Production-grade Rust backend guidance focused on idiomatic design, explicit error handling, safe concurrency, and reliable operations. Use when working with Rust code, .rs files, Cargo projects, or Rust services, APIs, and workers.
Verify that a feature or fix actually works before it ships: run it, exercise it like a real user, probe edge cases and error states, and report findings with reproduction steps. Use when asked to QA, verify, smoke-test, or sanity-check a change.