Why This Exists

The thing that bugged me

Twenty-plus years building products. Microsoft Research, Adobe, Salesforce, four startups (including one 6.5b exit, and one acquired by Microsoft), CPO at Commure, Crisis Text Line, SOCi, and Smartcat. Across all of that, I kept spending my weeks doing things that felt productive but weren't moving the needle. Specs nobody read carefully. Meetings that should've been async. Manually checking dashboards. Roadmap decks that stakeholders forgot by Monday. Sounds like you?

Here's what's changed: AI can now do most of that mechanical work. Not in theory, in practice. Agents that pull data, draft reports, monitor dashboards, summarize meetings, triage feedback, and flag anomalies. The technology is here. The question is whether PMs will use it, or keep grinding through the 80% manually while competitors don't.

I know I'm not the only one. Most PMs I talk to say the same thing: 80% of their week goes to mechanical work that creates zero customer value. The actual PM job, understanding customers, making good calls, figuring out what to build next, gets squeezed into whatever time is left.

That bugged me. So I started experimenting.

What changed

Three shifts happened around the same time:

I started talking to customers every week. Not quarterly research projects. Not NPS surveys. Real conversations, every week, even when they were just 20 minutes. My product decisions got better almost immediately. I stopped guessing. Credit to Teresa Torres for the framework. Her work on continuous discovery gave me the structure to make it sustainable.

I started prototyping instead of speccing. Instead of two weeks on a requirements doc, I'd build a clickable prototype in a couple hours and show it to customers. Feedback was richer, faster, more honest. Some of my best product decisions came from prototypes customers hated, because I learned that in a day instead of after three months of development.

I started using AI agents for the mechanical work. Not a chatbot. Actual autonomous agents running on a schedule, reading my tools, delivering reports. Clunky at first. The first two weeks of agent reports were mostly noise. But after tuning, they changed my mornings. I went from 45 minutes of dashboard-checking to a 5-minute report scan.

Then I realized agents change the product itself, not just the process. When you can ship an AI agent that does in seconds what used to take a human analyst hours, you're not optimizing a feature, you're rethinking the value proposition. The companies that figure this out first don't just save costs. They build moats. The product IS the agent. The platform IS the orchestration layer. This is the biggest shift in product strategy since SaaS ate on-premise.

None of these ideas are original. I borrowed from the Product Operating Model (Marty Cagan, SVPG), Continuous Discovery Habits (Torres), outcome-driven planning (Reforge, John Cutler, many others), and the AI agent ecosystem (Anthropic, OpenAI, the Model Context Protocol). What I'm trying to do is document how I put them together in practice, and share what's actually working at the intersection of product management and AI.

---

Working through this at your company? I do a small number of product org audits each quarter where I write the honest assessment and a 90-day plan against the Product Builder Standard. See current openings →

---

But even that feels dated now

Everything I just described, the continuous discovery, the prototyping, the AI agents doing mechanical work, that was the first wave. We've entered something fundamentally different. The Era of the Product Builder.

Here's what changed: PMs can now build. Not "build" as in write a spec and hand it off. Build as in sit down with an AI coding agent, describe what the customer needs, feed it your source code, your design system, your product context, and have a working prototype in hours instead of sprints. The gap between "idea" and "something a customer can touch" has collapsed.

This isn't a marginal improvement. It's a role redefinition. The PM who can prototype with AI isn't just faster, they think differently. They test assumptions in real products instead of slide decks. They ship experiments instead of requesting engineering bandwidth. They show customers working software instead of wireframes. The feedback loops shrink from weeks to hours.

And it goes beyond prototyping. The Product Builder generates their own Claude Code prompts from customer requirements, actual source code, designs, and product context. They orchestrate agents that auto-analyze discovery calls, monitor production metrics, and flag when the data contradicts the roadmap thesis. They don't manage a backlog, they build the thing, validate it, and then decide whether engineering should scale it.

The PM role isn't dying. It's splitting. There will be PMs who operate the way we have for thirty years, writing specs, attending standups, managing stakeholders. And there will be Product Builders who collapse the distance between customer insight and working product to near-zero. The second group will ship 10x more validated ideas, and the market will notice.

If you're reading this and thinking "that sounds like what I want to become", that's exactly who this site is for.

What this site is

My working notebook, organized around four areas I keep coming back to:

AI Product Operating Model. Empowered teams, product trios, owning problems instead of feature requests. Now add AI to the mix: which decisions can agents make autonomously? Where do humans stay in the loop? How do you structure a team when half the "work" is done by models? Still working a lot of this out.

Continuous Discovery. Weekly customer conversations, opportunity mapping, assumption testing. Now supercharged with AI: agents that auto-summarize transcripts, extract patterns across dozens of calls, and flag when assumptions are invalidated by new data. Still the single highest-ROI practice. AI just made it 10x faster.

Outcome Orientation. OKRs that measure behavior changes, not feature launches. AARRR dashboards. QBRs that tell a real story. AI agents now monitor these metrics continuously and surface anomalies before the weekly review. Writing good OKRs is still hard. But knowing when they're off track is now instant.

AI-Native Execution. The 65 agents we are building and shipping, the MCP setup, the scheduling, the tuning. This is the frontier. Most experimental area. Things break. Agents hallucinate. But the time savings are real when it works, and getting more real every week.

The AI Toolkit. The actual agents, prompts, and workflows I use daily. Prototype Prompt Generators that take customer requirements, your source code, your design system, and your product context, and produce a Claude Code prompt that spins up a working prototype in hours. Discovery agents that auto-analyze interview transcripts. Metric monitors that flag anomalies before standup. These aren't theoretical, they're the tools I build with and share here so you can fork them.

Who this is for

Two audiences:

Senior IC PMs who want to up their personal practice. Talk to more customers, make better decisions, automate the grind, do more impactful work. If you're a PM with 3-7 years of experience and feel like you should be operating at a higher level, a lot of this will land.

PM leaders (Directors, VPs, CPOs) who want better systems across their teams. Different challenges at this level. You're not just running discovery yourself, you're trying to get 5 teams to run it. I share what's worked at Smartcat and where I've struggled.

How to use it

Start wherever you want. There's a suggested order in the handbook, but most people should start with whatever problem they're feeling right now:

• Disconnected from customers? Start with Continuous Discovery.
• Drowning in status updates? Start with Your AI Agent Fleet.
• Shipping features but not moving metrics? Start with The Impact Loop.
• Want to prototype faster? Start with Prototype Before You Spec.
• Want to build with AI agents? Start with Your AI Agent Fleet and the toolkit section.

Everything here is a living doc. I update it as I learn. Some of what I wrote three months ago I'd write differently today. That's the point.

If something helps, or you've found a better way, tell me. Most of the best ideas here started as conversations with other PMs.