Many companies rush to automate messy processes, which only locks in inefficiency. Instead, learn and refine the process by doing it manually first, as early Amazon and DoorDash did. Only automate once the system is optimized, using technology to speed up good systems, not paper over bad ones.

The team writes very few specifications. When a spec is necessary for complex projects, it's incredibly brief—often just ten bullet points. This approach prioritizes speed and gives more autonomy to the people closest to the code, empowering them to make decisions.

While AI can accelerate prototyping, Linear's CEO deliberately uses a manual, slower design process for initial exploration. The friction of drawing things manually forces self-reflection and a deeper understanding of the problem, a benefit that can be lost when optimizing purely for speed.

Instead of over-analyzing and philosophizing about process improvements, simply force the team to increase its cadence and ship faster. This discomfort forces quicker, more natural problem-solving, causing many underlying inefficiencies to self-correct without needing a formal change initiative.

Vercel builds internal AI agents and tools, like an Open Graph image generator, to automate tasks that were previously bottlenecks. This not only increases efficiency but also serves as a critical dogfooding process, allowing them to innovate on their core platform by building the tools their own teams need.

Microsoft reportedly canceled internal licenses for a competing code assistant to force its developers to use its own Copilot CLI. This "dogfooding" strategy is a proven method to rapidly improve a product by making its creators its primary, most critical users.

The feeling of being overwhelmed by AI stems from applying new technology to old structures like quarterly roadmaps and PRDs. The real solution isn't just faster work, but re-architecting the entire product development process to natively leverage AI, much like building superhighways for cars instead of using old horse trails.

The Codex team's core mandate was to create a tool they loved and used daily for their own development. This intense dogfooding—including building the app on itself—served as the ultimate validation and quality bar before they considered shipping it externally.

The Codex team combines research, product, and engineering, allowing them to solve problems at either the product level or the core model level. This tight integration creates a flywheel where product needs drive research and research breakthroughs are immediately applied to the product.