# Prototype Volume, Product Sense, and the New Shape of PM Work

*By PM Daily Digest • June 30, 2026*

This brief covers the shift from coordination to strategy and discovery, Anthropic’s prototype-heavy Claude Code practices, and Teresa Torres’s simple habit for learning where AI actually helps. It also highlights the skills becoming more valuable as AI compresses delivery work.

## Big Ideas

- **AI is demoting coordination work and promoting strategy/discovery.** Christian Idiodi argues that technology is now the business, not a support function [^1], and his “Jim” story shows the failure mode: a PM day consumed by incidents, ceremonies, and status updates [^1]. His larger point is that AI is disrupting delivery most, so the remaining differentiator is selecting the right problems and solving them well [^1]. **Why it matters:** as delivery gets easier, reactive coordination contributes less. **How to apply:** audit your week and deliberately move time toward customer problems, strategy, and discovery.

- **The role is getting more fluid, not disappearing.** Idiodi says the core job still spans strategy, discovery, and delivery [^1], while Andrew Ambrosino warns against replacing PMs with generic “builders” [^2]. At Anthropic, Aakash Gupta describes five modes PMs may shift between depending on product need: Prototyper, Builder, Sweeper, Grower, and Maintainer [^3]. **Why it matters:** value is increasingly defined by what you actually do, not your title [^4]. **How to apply:** ask which mode your product needs right now, then bias your time accordingly.

> "Without a product vision, my friends, AI is just random experimentation in your company. Without a product strategy, AI is just noise." [^1]

## Tactical Playbook

1. **Run discovery as a prototype portfolio.** PMs and designers should “build to learn,” while engineers “build to earn” with scalable, reliable systems [^1]. Anthropic reportedly builds first, skips PRDs, and expects most prototypes to die [^3]. **How to apply:**
   - Pick one product risk to reduce
   - Create multiple fast prototypes
   - Judge them on value, usability, feasibility, and viability [^1]
   - Kill weak options quickly
   - Hand only strong survivors to engineering, then later sweep out what stops pulling weight [^3]

2. **Use AI on one real task every day.** Teresa Torres’s habit is simple: pick one item from your to-do list, let AI take a first pass, then learn from where it fails [^5]. **Why it matters:** repeated trials expose AI’s actual strengths faster than chasing every new tool [^5]. **How to apply:** start with, *“I have to do this task. How can you help?”* Then respond to the first output with specific feedback, even if it is “this was terrible—what context do you need to do better?” [^5]

## Case Studies & Lessons

- **Anthropic’s Claude Code team is optimizing for volume, then judgment.** The team reportedly tried hundreds of versions before shipping agent teams; even the loading spinner took 50-100 iterations, with about 80% never shipping [^3]. AI reviews every PR before a human does, and the team prefers the general model over specialized ones [^3]. **Takeaway:** when build speed rises, the scarce skill is quickly separating the prototype that becomes a product from the one that wastes a quarter—what Gupta calls “taste at speed” [^3].

## Career Corner

- **Product sense is becoming a career moat.** Idiodi defines it as knowing what it takes to build a good product for customers within business and environmental constraints, including revenue, cost, regulation, market shifts, and company realities [^1]. Aakash Gupta makes a parallel point for designers: when interfaces take minutes to generate, the hard question is which one deserves to exist [^6]. **How to apply:** evaluate AI-generated options against differentiation and business constraints, not just whether they technically work.

- **Human skills are rising, not falling.** Empathy, influence, trust-building, critical thinking, grit, and emotional intelligence are cited as durable advantages [^1]. **How to apply:** treat collaboration and judgment as first-order skills, not soft extras.

## Tools & Resources

- **Practical inspiration:** Teresa Torres says her “AI at work” examples are meant to spark personalized use cases, not prescribe one workflow [^5]. Her monthly Claude Code show-and-tell sessions serve the same purpose by exposing people to what peers are building [^5].

- **Worth watching:** [Enter the era of the product creator](https://www.youtube.com/watch?v=10kuiJfxWHM) for the strategy/discovery shift [^1], and [AI Shaped Problems - All Things Product with Teresa & Petra](https://www.youtube.com/watch?v=L0GQVFBCahw) for a practical adoption habit [^5].

---

### Sources

[^1]: [Enter the era of the product creator - Christian Idiodi at #mtpcon London 2026](https://www.youtube.com/watch?v=10kuiJfxWHM)
[^2]: [𝕏 post by @lennysan](https://x.com/lennysan/status/2071693688661516743)
[^3]: [substack](https://substack.com/@aakashgupta/note/c-285068404)
[^4]: [𝕏 post by @lennysan](https://x.com/lennysan/status/2071628545252827579)
[^5]: [AI Shaped Problems - All Things Product with Teresa & Petra](https://www.youtube.com/watch?v=L0GQVFBCahw)
[^6]: [substack](https://substack.com/@aakashgupta/note/c-284787261)