Fast, Cheap, Good: Choose Three (cory.news)

AI is poised to collapse the long-standing “fast, cheap, or good: choose two” trade-off in software development. Building on Charity Majors’ observability-first framing, the piece argues that the old split between durable (trusted, maintainable) and disposable (rapid, brittle) code is eroding because AI can both generate rapid prototypes and perform the engineering work that makes them trustworthy. The decisive axis is trust — i.e., observability (o11y), testing, and validation — not cost or speed alone. Technically, today’s AI tools can accelerate prototyping (code generation, GPU-accelerated iteration, language conversion) while also automating durability tasks: adding telemetry, synthesizing tests and validators, documenting and explaining code, and refactoring or discovering canonical solutions. The catch: this only works reliably with proper tooling and guardrails — robust testing frameworks and an observability foundation that let LLMs reason about and modify production behavior. Practically, teams that build for o11y and LLM-assisted workflows from the start will be able to iterate fast, keep costs low, and achieve production-grade reliability — effectively choosing “fast, cheap, and good.”