The playbook is being rewritten

For years there have been well-understood best practices for accelerating developers with quality at scale. But coding agents are forcing all teams to question the old assumptions of the SDLC.

Traditionally, many of the largest companies would standardize on one or two languages because supporting more was too expensive. With coding agents, this is probably wrong — now you should probably just use the best language for the job. Previously, entire teams existed to maintain UI design systems, but in a world where a Figma MCP can bridge design to code directly, it's not clear that layer of abstraction survives.

The playbook for how software gets built is being rewritten right now, and nobody knows the end state. Practices that were standard for a decade are being questioned in months. The assumptions that drove team structure, tooling choices, and review processes are all up for renegotiation.

What we do know: the speed at which code can be written and reviewed by agents has already outpaced the speed at which teams can agree on what to build. That gap is growing. And the tools designed for the old playbook — Jira tickets, Google Docs specs, Slack threads — were never built for a world where the primary velocity bottleneck is alignment, not execution.