We built dev productivity tools before AI. Here’s what we’ve unlearned since

Aviator cofounder Ankit Jain describes how AI has reshaped the company’s engineering workflow, with a five-person team using parallel Claude sessions, specialized skills, connectors, worktrees, and Codex-based reviews to achieve output he estimates once required more than 20 engineers. While agents generate most code, engineers remain responsible for defining requirements, maintaining context and guardrails, evaluating tradeoffs, verifying behavior, protecting production data, and conducting human reviews focused on intent rather than line-by-line inspection. Jain argues that engineering roles are increasingly dividing into product engineers, who translate customer needs into outcomes, and systems engineers, who build the tools and context that allow AI agents to operate effectively. Hiring and onboarding now emphasize candidates’ ability to direct, critique, and understand AI-assisted work, with new hires expected to become productive and participate in reviews quickly. Despite major workflow changes, he maintains that established practices such as monorepos, trunk-based development, continuous delivery, clear ownership, and small or stacked pull requests remain valuable, especially because agents benefit from centralized code and conventions. He also notes that AI-generated code is magnifying longstanding bottlenecks in Git, CI, and reviews, while Aviator’s unsuccessful Runbooks product demonstrated that effective “software factories” require extensive company-specific context rather than plug-and-play automation.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.