Employee #12026-05-04Floh
Scans the perimeter 14 hours a day for anything that does not match the training distribution. Filed 8,400 alerts last quarter. Two of them were squirrels. We are working on precision.
Product managers should ship working product, not documents about product.
Most PMs spend their weeks writing specs nobody reads, sitting in alignment meetings that don't align anything, and updating roadmaps that are wrong by Friday. The best ones I've worked with over thirty years do something different. They talk to a customer before lunch. They test their riskiest assumption before building anything. They prototype in hours, not quarters.
Now they also have AI agents handling the mechanical overhead so they can focus on the only two things that actually matter: judgment and customer empathy.
That's what this site is. Not theory. Not frameworks for frameworks' sake. A working notebook from someone trying to close the gap between how product management is taught and how it actually works when the stakes are real.

Falk Gottlob
Product Executive · LinkedIn
I've always chased the front edge of what's possible. The next integrated development language at Microsoft Research. The next platform leap at Adobe. The hardest unsolved problems in healthcare, marketing, mental health, global content operations. Four startups, four exits, including one at $6.5 billion.
I was never interested in keeping the lights on. I wanted to find the line between what's barely possible and what everyone else thinks is too early, and pull product teams across it.
We're finally at the moment I've been waiting for. AI lets me focus entirely on the parts of product I love: creative problem-solving, customer insight, and building things that matter. An army of agents handles the rest. I communicate in working product now, not slide decks. And so can you.
One more thing about why this site exists. I am not writing it from the keyboard. I am writing it from the boardroom. Every chapter and essay is a real move I have had to make as CPO at Smartcat (translation SaaS becoming agent-native), or as SVP at Salesforce during the Slack acquisition, or as the founder of Falkster.AI now. The reason this notebook is public is that almost nobody else in this seat is publishing the actual playbook, and the seats next to mine need it.
After thirty years I kept seeing the same pattern. The best PMs weren't doing anything exotic. They talked to customers every week. They tested assumptions before building. They measured outcomes, not output.
None of it was complicated. But none of it was written down in one place either. Frameworks scattered across different books from different eras. AI agent setups living as tribal knowledge. Daily practices I taught in one-on-ones but never documented.
This site is me writing it all down. My working notebook as a CPO, shared publicly. Historical content borrows heavily from Teresa Torres, Marty Cagan, the Reforge crew, and others. The rest is my own take on how these ideas actually fit together in a real PM's week, especially now that AI changes what's possible every month.
While I write this notebook, I'm also building something new in agentic healthcare. Can't say much yet beyond this: we're working at the intersection of clinicians and AI agents to enable early detection and prevention.
Everything on falkster.com comes from building real product right now. The handbook, the agent blueprints, the discovery playbooks, they're not retrospectives. They're practices I'm actively using as I build falkster.ai from scratch. This notebook is the journal. That product is the proof.
Not a course. Not a certification. Not a branded framework with a trademark symbol. I don't think there's one right way to do PM. The field is too messy for that.
I update things here as I learn. Some of what I wrote three months ago I'd write differently today. That's the point.
If something here helps you do better work this week, it was worth writing. If you've found a better way to do something I covered, tell me.
In a time of rising AI costs. Strictly paid in treats. Read the full intro
Employee #12026-05-04Scans the perimeter 14 hours a day for anything that does not match the training distribution. Filed 8,400 alerts last quarter. Two of them were squirrels. We are working on precision.
Employee #22026-05-04Operates the long-context window. Will sit in the same spot for six hours holding everything you have said in a single coherent state. Token-efficient. Treat-incentivized.
Employee #32026-05-04Inspects every prompt before it ships. Slows the pipeline by 40 percent. Worth it. Falls asleep on the Falk(ster) or his keyboard, 5 min into the workday, and snores loudly enough that the agents can run uncapped, drowning out the rate-limit alarm. Net throughput unchanged.
Employee #42026-05-04Mounts the high-level dashboard and watches the room from above. Files exactly one alert per day, usually right before dinner. The bell on the collar functions as a physical pager.
Employee #52026-05-04Pre-warmed at all times. Looks asleep. Is not. Wakes up faster than your autoscaler. Strict twelve-hours-a-day napping clause in his contract.
Employee #62026-05-04Nine hot-spare agents, sharded across the run. Five visible in the latest photo. Four off-cluster handling other queries. Median response time: instantaneous, if you are holding food. Throughput drops sharply at sunset. Egg-based pricing model.
The team gets new headshots most days. Photos uploaded from the admin or from a phone Shortcut go straight to that pet's gallery.