We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Big Ideas
Cheap building moves the PM bottleneck upstream. Agents can already build prototypes, instrument products, test implementations, review code, and fix issues; the harder question is shifting from “can we build this?” to “should we build it?” The practical response is to study behavior-changing mechanics rather than clone feature surfaces: identify the behavior your product depends on—such as trust to delegate, evidence becoming a decision, or context surviving a handoff—then find a proven mechanic from another category and build the smallest test. As implementation gets cheaper, judgment, taste, and ruthless prioritization matter more because teams can become very good at building bad ideas quickly; the differentiator is a repeatable loop of hypothesis, smallest test, measurement, and retained learning.
Agent UX still needs fast first-mile proof. Scott Belsky’s note argues that consumer products win partly on how users feel about themselves using them and how quickly they reach value. For agent experiences, he specifically calls for quick ROI on the time and data users provide, onboarding that balances general value with personalized problem-solving, and fresh onboarding validation for new cohorts rather than assuming beta behavior will scale.
Tactical Playbook
Turn discovery into a compounding log. Use one real potential customer per day as a baseline—without pitching—and ask what they are trying to solve, how they handle it now, and what they have already tried. After roughly 30 conversations, recurring language and objections should clarify positioning and acquisition; the value comes from writing down the patterns. For a higher-intensity version, run five conversations daily, ask about a recent event, workaround, owner, cost, and urgency, end with a dated next step, and review weekly which segment replies, reaches value, and pays. Build funnels and automation only after the log shows something repeatable.
Audit “not enough time” before adding another process. Nir Eyal frames motivation as behavior + benefit + belief: knowing the action and wanting the result is insufficient if someone does not believe in the outcome or their ability to act. His interview’s belief audit is usable for PM prioritization: write the belief, test whether it is absolutely true, examine how it changes behavior, ask who you would be without it, then try an opposing perspective for a week.
Case Studies & Lessons
AI lawyer: validate trust before buying reach. A founder says an early Reddit/news spike faded because the product was not ready; now it is usable, but the hype is gone, with SEO and waves of previous-user outreach supplying the remaining traction. Community advice identifies the product’s original credibility story—not “AI” itself—as the trust mechanism, and recommends a real lawyer using the product on real files and willing to vouch before paid acquisition. Search ads should follow evidence about the terms that successful users already searched, not precede it. The PM lesson is to treat distribution problems as product-proof problems until the target workflow and trust signal are explicit.
An org chart does not create a product culture. An internal-product proposal assigns PMs business needs, buy-versus-build, pilots, roadmaps, and outcomes; Data Science/AI owns models and agents, IT owns infrastructure, and business teams supply domain expertise and KPIs. The response is that this separation alone will not end project-pipeline behavior: teams need to validate before building, deploy and learn quickly, and treat every “sure thing” as unvalidated until evidence arrives.
Career Corner
Customer access is part of the PM job, not a perk. A platform-security PM reports an engineering-led environment where architects design features, the internal CISO is the stated customer, and large customers speak only with leadership; the PM still prioritizes but lacks direct discovery. The practical career move is to name the customer explicitly and ask for access rather than quietly accepting second-hand requirements. That gap may be structural—executive customer history or sales-controlled relationships can push PM value toward business operations—so document how you turn indirect input into decisions while seeking roles with real customer exposure.
Tools & Resources
Sharpn.ai is a new community-built PM mock-interviewer: its author describes a free simulator covering product sense, metrics, strategy, behavioral, execution, technical, and estimation interviews, with job-description tailoring, scoring, feedback, and study articles. The author says probing uses contextual clues to decide when to dig deeper. Use it for repetition and feedback; the reported Microsoft outcome is the builder’s own claim, not independent validation.