We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Big Ideas
AI-native teams need product judgment at engineering speed. Oji Udezue describes engineers adopting AI and moving roughly 10× faster while PMs continue discovery and strategy at the old pace, leaving product as the bottleneck. His ProductMind system addresses the layers often missed by engineering-focused AI tooling: 17 skills cover business and product work such as finding problems, deciding what to build, creating value, monetization, direction, and execution. The important design choice is the use of gates: because LLMs rarely say no, a skill can stop the next step and replicate a piece of PM judgment rather than merely asking a PM to edit the output afterward.
Agent value is moving toward data-rich workflow platforms. Aaron Levie argues that if agents generate 100× more code, process large datasets, and make decisions across systems, the platforms managing that data and work become more important—not less; governance, security, compliance, guardrails, and safe data access are part of the product. Scott Belsky makes the related case that a “teamwork graph” improves the efficiency and accuracy of both work and agent actions, while Hiten Shah reduces the strategy question to who controls the data. For PMs, model choice is only one part of the roadmap: define the context, permissions, workflow state, and control surface the agent can safely use.
Tactical Playbook
Put a no-build gate in front of AI-generated features. ProductMind’s Sharp Problem Test asks:
- What do people do today? An easy workaround means there may be no problem.
- How many people have it, and how often? Daily use can outweigh a market ten times larger with monthly use.
- Will anyone pay? Check what they already spend on the workaround.
Fail one question and kill the idea; require a 3× improvement over the current alternative before expecting switching. The associated 11-step scaffold deliberately keeps its first four steps away from code and produces research and product artifacts before implementation.
For B2B discovery, convert interest into a small commitment immediately: when real pain surfaces, ask on the first call whether the prospect will review rough versions. Then ask who can grant access to the relevant ticketing systems or logs; “I’d have to ask” is research about the buying path, not pilot readiness.
Case Studies & Lessons
The Savannah Bananas turned observation into the roadmap. The team videotaped fans and saw them leaving early; the evidence suggested the three- to four-hour game was too slow, so the plan changed from being a conventional baseball team to creating a faster show. Jesse Cole’s advice is behavioral rather than survey-led: put ideas into the world and watch what fans do. The team then tested “banana ball” behind closed doors—nine innings in 99 minutes—accepted that the first version was chaos, and expanded only after fans stayed more than an hour after a show.
The business model reinforced the product test: sell tickets rather than depend on sponsors, progressing from two tickets in the first three months to 4,000-seat sellouts. They still price tickets at $40–$60 with no fees and run a face-value resale market, despite secondary-market prices above $200, prioritizing fan trust over immediately extractable revenue. The PM pattern is clear: observe behavior, prototype narrowly, set a behavioral threshold, then scale—and treat pricing as part of the product promise.
Career Corner
Make judgment visible in senior interviews. Shreyas Doshi says interviewers can learn a lot from the questions candidates ask and points to the subtle cues hiring managers often miss. Prepare questions about decision context, constraints, trade-offs, and what has failed; the question phase should show how you reason, not only what role you want.
Tools & Resources
- Agent Plugins: OpenAI and partners introduced an open format that packages Agent Skills and MCP server configurations so one plugin can work across compatible agent clients. Treat portability as a product requirement for internal agent capabilities.
- PM AI workshops: Lenny Rachitsky’s Head of Education is running free workshops with Cursor, Replit, PostHog, and Supabase on using popular AI tools in day-to-day PM work; registration is available through the linked course page.