How the most AI-pilled product team builds products | Fiona Fung (Claude Code and Cowork)

Key Concepts

• AI-Pilled Engineering: A shift in software development where AI tools (like Claude Code) handle the bulk of coding, moving the bottleneck from writing code to verification and product strategy.
• High Agency, High Accountability: A management framework where team members have the freedom to act autonomously but are held strictly accountable for the outcomes and hypotheses they test.
• Latent Demand: Identifying and scaling solutions for user behaviors that emerge organically, even if they weren't the original intended use case for a product.
• Just-in-Time (JIT) Planning: Moving away from long-term (6-month) roadmaps toward lightweight, monthly planning that is frequently adjusted based on real-time feedback.
• Dogfooding: The practice of using one's own product daily to identify bugs, UX friction, and new opportunities for improvement.
• Asynchronous (Async) Workflows: Utilizing AI agents to handle routine tasks (like monitoring feedback or generating PRs) to allow for more efficient, non-linear work.

---

1. The Transformation of Software Engineering

Fiona Fung, who leads the Claude Code and Co-work teams at Anthropic, highlights that Anthropic engineers are shipping eight times as much code per quarter compared to the 2021–2025 period.

• Coding as Salt: Coding is no longer the primary bottleneck; it is now a commodity. The ceiling for what an individual can build has been lifted, shifting the focus from "how do I write this?" to "how ambitious can I be?"
• The New Builder Profile: Hiring now prioritizes two distinct profiles: creative builders with product sense (who own the end-to-end experience) and deep systems experts (who handle the "hard parts" that require verification).

2. Methodologies for Managing High-Velocity Teams

Fung describes a management style adapted for an AI-first world:

• The "Remote Session" Management Technique: Fung uses a Claude Code remote session to monitor all repositories, Slack channels, and metrics. This allows her to have informed, high-level conversations with her team about impact and quality rather than just tracking lines of code.
• Bad vs. Sad Framework: To manage quality, teams categorize issues as "Bad" (irrecoverable, critical errors) or "Sad" (recoverable pain points). This framework allows teams to prioritize fixes based on the actual user experience rather than just raw performance dashboards.
• Automated Verification: Since human code review is a bottleneck, the team uses AI to validate code against established frameworks and specs. This is described as an evolution of Test-Driven Development (TDD), where the AI generates the tests first, removing the "tax" of manual test writing.

3. Addressing the "AI Divide" and Fear

Fung addresses the growing gap between those leaning into AI and those resisting it:

• Growth Mindset: She emphasizes that what made an engineer successful in the past may not be what makes them successful in the future.
• Leaning into Fear: When faced with anxiety about AI, she advises asking, "What is within my control?" and "What can I do about it?" She cites the quote: "The cave you fear contains the treasure you seek."
• Small Business Advocacy: Fung is passionate about bringing AI to non-technical users (e.g., small business owners). She suggests that listeners act as "AI ambassadors" by helping friends or local businesses solve specific, painful tasks (like invoicing or document organization) to demonstrate the value of these tools.

4. Future Frontiers: Async Agents and Routines

The next shift in engineering is toward asynchronous agentic workflows:

• Routines: Instead of manually prompting an agent, managers can set up "routines" that trigger agents to perform daily tasks (e.g., summarizing feedback, identifying bugs, or drafting PRs).
• Context Switching: While agents increase productivity, they also increase the load of context switching. Fung notes that she now blocks "focus time" specifically to review the work generated by her async agents.

5. Synthesis and Key Takeaways

• Metrics vs. Anecdotes: While metrics are useful, Fung argues that product leaders should trust anecdotes and personal "dogfooding" experiences over dashboards. If a product is hard to use, the leader will feel it personally.
• Culture as a Living Thing: As Anthropic grows at an unprecedented pace, Fung views team culture as the most critical, fragile asset. She encourages open, honest debates and a "one team" mentality where members support each other to the finish line.
• Actionable Advice: For any team, Fung suggests identifying one manual, high-noise process and asking: "Is this still serving its purpose?" If not, kill it.

Notable Quote: "In a world where you can be anything, be kind." — Fiona Fung, reflecting on the importance of human connection during high-growth, high-stress periods.