We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
1) Big Ideas
AI should widen the option set, not make the product decisions. Ravi Mehta defines taste as curation: knowing the possibility space, deciding what matters, and choosing what is right for an intended audience. His AI workflow is frame → explore → critique → explore again → select → refine, replacing ask → generate → accept; use AI aggressively for exploration and cautiously for decisions. For PMs, generate several interaction or feature directions, then record the audience, constraint, and evidence behind the selection. When your preference and user preference diverge, interrogate the gap rather than dismissing the user.
The harness is part of the product. Hiten Shah argues that even a great model fails inside a product that “loses the thread.” Harness-Bench ran 5,194 agent trajectories across 106 tasks and saw average harness scores from 52.4 to 76.2 with the same task suite and model pool. Treat context continuity and agent evaluation as product requirements, not post-launch plumbing; improve the surrounding system before assuming the model is the bottleneck.
2) Tactical Playbook
Make metrics actionable, not ceremonial. A Reddit synthesis of two 2026 product surveys reports that 34% of PMs have no clear primary metric, 49% lack time for analysis and tracking, and 40% do little or no experimentation; it also reports that only 4% think company metrics reflect their work. Treat this as directional: the thread flags sample bias and KRs that PM teams cannot independently move, particularly in B2B. Start with one leading indicator believed to connect to revenue or retention, track it consistently, and replace it if evidence breaks the link. Leadership-set roadmaps can be appropriate when executives have better customer context; the test is customer evidence, not whether leadership was involved.
Qualify enterprise demand before allowing it onto the roadmap. A bank-and-insurer vendor’s checklist is concrete: find the P&L owner with budget and mandate, request the security questionnaire by the second meeting, and attach a budget line to pilots. Multi-thread above and beside a champion, and make one-off custom work paid. This filters interested-but-powerless buyers and prevents unpaid, customer-specific scope from becoming product strategy.
Try the smallest fix before rebuilding. A team with unusable Metabase image thumbnails considered a new internal tool, then used a Chrome extension to zoom on hover in five minutes; the dashboard stayed fast and the team shipped faster. Isolate the missing capability, test a reversible workaround, and only then price the operational cost of a rebuild.
3) Case Studies & Lessons
Cursor made a contrarian product-layer bet, then changed layers without losing the thesis. It focused on the human–model interface rather than a coding-specific foundation model or VS Code plugin, because users express intent in natural language and a plugin would make Cursor part of someone else’s product. It treated product quality as go-to-market and deferred enterprise. As models advanced, it moved from IDE to agent platform to model platform in about two years; user scale supplied data and know-how for its own models, and later enterprise economics justified a rapid sales build. Lesson: define the layer you can own, let it constrain scope, and revisit the layer when technology or margin structure changes.
4) Career Corner
AI PM hiring has expanded while general PM hiring contracted. Aakash Gupta’s tracking shows 16,420 AI PM listings versus 35,725 PM listings—46% of open PM roles, up from 2% in February 2024—while total PM jobs fell 18%. In a hand-classified sample of 113 postings, 69% asked for prior AI/ML experience, but only 14 required production or scale experience; median PM tenure was five years while examples often asked for one or two AI years. The author notes that 40 classifications were judgment calls. The practical opening is to build a real AI product, get it used, document decisions and evaluation artifacts, and publish it as verifiable work—not fabricated experience. One example: after 15 years in Apple hardware and a senior program role, Brian Luc built RenoSmarter.ai and became an AI PM at Cisco five months later.
5) Tools & Resources
Use a structured feedback request. Product communities’ Feedback Friday template asks for the company, URL, purpose, technologies, specific feedback requested, beta-testing need, and additional comments; it discourages link-only promotion and asks feedback seekers to return feedback. Reuse it for user research or beta recruitment: a concrete question and reciprocal ask should produce more actionable input than “thoughts?”
Behavioral-product lens. The speaker defines behavioral design as using technology to help people form healthy habits and break bad technology habits. He argues distraction is driven by internal discomfort—such as loneliness, uncertainty, and boredom—as well as external triggers such as notifications, email, and devices; products can become escape routes from those states. For PM discovery or engagement work, translate this into a trigger-to-outcome map: identify the discomfort, the intended behavior, and whether each cue serves that behavior or merely captures attention.
Indistractable execution model. Traction is intentional action toward a plan, while distraction is anything pulling away from it—even seemingly productive email when it displaces planned work. Apply the model in sequence: (1) use a 10-minute rule to delay temptation and examine the underlying sensation with curiosity and self-compassion; (2) turn values into scheduled time and synchronize calendars with a spouse or boss; (3) audit and hack back external triggers across phones, computers, chats, email, and workplace interruptions; and (4) only then add precommitments such as price, effort, or identity pacts, since using them out of order can lead to failure.
Timeboxing for product work. For perfectionism and open-ended tasks, replace completion-based to-do lists with time-boxed commitments; measure success by uninterrupted work on the task for a set period, which can provide relief and build agency through repeated follow-through.
Responsible-engagement caveat. The speaker rejects the blanket idea that technology hijacks everyone’s brains, warning that this narrative creates learned helplessness; his alternative is a systematic approach in which devices serve people rather than people serving them.
- A Reddit post citing the 2026 Product Focus and State of Product surveys reports that 34% of PMs have no clear primary metric they are accountable for ; among PMs with a primary metric, 35% are tied to revenue/P&L and about 4% to feature adoption . It also reports that 49% lack time for data analysis and metric tracking , only 4% believe company metrics reflect the value of their work , and 40% do little or no experimentation while roadmaps are set by leadership preference . Commenters cautioned that the figures may be sample-biased and are difficult to interpret without knowing the respondent cohort .
- Metric accountability varies by seniority, organizational design, and product type. One commenter associates senior PM accountability with revenue, usage, or retention , while another argues that many assigned KRs are not independently actionable and recommends cross-functional alignment around a small set of north stars . In B2B, sales-driven outcomes and immature analytics can make it difficult to connect product usage directly to revenue or churn ; internal-tool PMs may instead be accountable for delivering a capability that enables another team’s use case .
- A practical approach is to agree on one leading indicator believed to connect to downstream revenue or retention, track it consistently, and change it when evidence shows the connection is weak . In early-stage and midmarket SaaS, leadership influence on the roadmap can be appropriate when executives have stronger customer context; the problem is preference unsupported by customer evidence, which PMs should bring back to leadership .
- One practitioner pairs revenue with customer calls—at least four per quarter—attach rate, churn, NPS, delivery against schedule, support volume, outages, active partners, and active users . After outage numbers and support calls rose, the PM addressed technical debt with a “get well plan” they said worked; customer and market feedback also prompted accelerating an initiative as the market moved quickly .
- Choose the product layer where you can win. Cursor focused on the human–model interface instead of building a coding-specific foundation model or shipping a VS Code plugin. Its rationale was that users communicate with models in human language, requiring broad language understanding, while a plugin would make Cursor part of someone else’s product.
- Let the product thesis constrain scope and go-to-market. Cursor treated product quality itself as the go-to-market motion, deferred enterprise rather than pursuing it early, and maintained a focused product-company identity instead of becoming a collection of plugins, services, and adjacent bets.
- Use adoption to build capability, then self-cannibalize as technology changes. After acquiring users rapidly, Cursor accumulated the data and know-how to build its own models—a “backdoor” path that would not be available in the same way to a company starting as a frontier lab. As model capabilities advanced, it moved from IDE to agent platform to model platform in roughly two years, while tab autocomplete became less central.
- Expand into enterprise based on observed economics rather than inherited playbooks. Cursor initially rejected advice to invest in sales, then concluded that enterprise contained the bulk of customer spend and much of the margin while its product remained a strong top-of-funnel engine; it subsequently built its sales motion at exceptional speed.
- An experienced 0–1 PM with about seven years in product, mostly in fintech, transferred to a highly valued flagship project after a nine-month internal process. Shortly after the transfer, the system title changed to Senior Program Manager, product-development responsibilities were reassigned to another team member, and the PM’s scope became 50% partnerships plus events, communications, and administration—work they said did not fit their strengths or career direction.
- The practical escalation suggested in the thread is to clarify the original transfer mandate with the VP or executive sponsor: ask what the transfer was intended to achieve and what rationale supports the redesigned responsibilities, especially when the new director cannot explain the value trade-off.
- If the role remains misaligned, commenters proposed either delivering exceptionally in the new specialty, becoming the company’s internal expert, and building relationships with people who can support future advancement, or pursuing an external role that better matches the PM’s skills and goals; one commenter said that such a move should not be viewed negatively by a credible recruiter.
- A counterpoint for PM career positioning was that product management is not mainly idea generation: it involves structured human interaction, behavioral analysis and observation during discovery, communication, and decisions grounded in more than personal opinions.
- AI PM hiring has expanded sharply: AI PM listings rose from 2% of open PM roles in February 2024 to 46%, while overall PM jobs fell 18% and non-AI PM listings shrank. Hiring managers commonly seek evidence of having shipped AI products to production at scale because prior experience reduces perceived product, business, and team risk, including the ability to work with AI engineers/researchers, design evaluations, and support launches.
- A hand-classification of 113 LinkedIn AI PM listings found that 69% explicitly required prior AI/ML experience, 19% required familiarity, and only 11% had no AI bar; the author notes that 40 classifications were judgment calls and a softer reading could put the requirement at 55%. AI experience was requested nearly as often in generic PM postings as in roles titled AI PM (66% versus 70%), but only 14 of 113 listings required production or scale experience. The median listing sought five years of PM experience, while examples separately requested only one to two years of AI experience, suggesting that a focused AI track record can supplement broader PM tenure.
- The recommended career tactic is to build “synthetic experience”: independently create a real AI product, get it used, and document the product decisions so the resulting artifacts can be verified and discussed in interviews; the approach is explicitly different from fabricating employment experience. One example is Brian Luc, who moved from roughly 15 years in Apple hardware engineering and a Senior Program Manager role without a software or AI PM title to building the AI renovation tool RenoSmarter.ai, updating his LinkedIn positioning, and posting about the project; five months later he became an AI PM at Cisco.
- Lenny Rachitsky highlights Marc Andreessen’s view that AI arrived at a time when slowing growth and population decline needed to be countered. The accompanying argument is that even tripling productivity growth would restore job turnover only to historical 1870–1930 levels, with faster growth generating new careers, products, and services rather than causing mass unemployment.
- PM interview expectations are becoming more product-specific. A review of 84 published interview reports from 10 leading tech companies found product sense/design was the largest category at 29/84 (35%); examples include improving a favorite app and solving a live product challenge for leadership, with candidates expected to defend opinions through trade-offs rather than recite frameworks. Metrics/execution represented 19/84 (23%), strategy 14/84 (17%), behavioral/values 13/84 (15%), and technical AI 7/84 (8%). Preparation should therefore include diagnosing a concrete broken metric, making company-specific strategic choices, preparing values-based disagreement examples, and pairing technical AI depth with plain-language explanation.
- AI PM hiring is expanding while becoming more selective. The post reports 35,725 PM listings and 16,420 AI PM listings, with AI PM roles rising from 2% of open PM jobs in February 2024 to 46% today; overall PM jobs fell 18% over the same period. In a hand-classified sample of 113 AI PM listings, 69% required prior AI/ML experience, 19% required familiarity, and only 11% listed no AI bar; the author notes that 40 classifications were judgment calls and a softer reading could reduce the requirement to 55%. Only 14 of 113 listings explicitly required AI experience in production or at scale, while the median listing sought five years of PM experience and examples commonly required only one or two years of AI experience.
- Synthetic experience is a practical, non-fabricated route into AI PM. The recommended approach is to build a real AI product, get it used, document the product decisions, and present the resulting verifiable work across a resume, portfolio, and LinkedIn; fabricated experience is explicitly rejected because candidates must be able to explain technical artifacts such as evaluation sets. Brian Luc applied this playbook after 15 years in Apple hardware engineering and a senior program-management role: he built RenoSmarter.ai, publicly posted about it, and became an AI PM at Cisco five months later.
Harness-Bench tested 5,194 agent trajectories across 106 tasks and found average harness scores ranging from 52.4 to 76.2 across the same task suite and model pool, signaling that the software or harness layer around an AI model can materially affect agent outcomes—not just the underlying model.
Hiten Shah announced a live session inviting people to bring an AI task that repeatedly goes wrong; he will demonstrate how to identify what is getting in the way and decide what to change first.
For AI product diagnosis, “the model is bad” is not a sufficient explanation: product choices can make a capable model appear unintelligent.
Hiten Shah announced a session examining why the same Claude or ChatGPT model can feel better in one product than another, including what happens around the model and how to determine what went wrong when the resulting work is poor.
- A PM in a 31-PM, five-director organization reports that they and one teammate carry roughly half of the company’s most important initiatives: the author owns three of the top 10 initiatives and materially contributes to a fourth, while the teammate owns two and materially contributes to a third; the other 29 PMs cover the remaining five, with two receiving substantial help from them. The sustained concentration has created significant pressure and stress, although performance reviews and 5–10% annual raises over the past three years have recognized the workload.
- Career-market signal: one commenter reports that a role was put on hold or canceled midway through the interview loop and describes the PM job market as exceptionally difficult.
- AI adoption caveat from the community: commenters describe AI-generated artifacts appearing across Slack, Jira, presentations, spreadsheets, documentation, and leaderboards, while one says visible AI-written documents reduce their trust in both the document and its sender.
An India-based PM candidate seeking remote US/EU roles with dollar compensation has tried targeting companies with consistent pay across regions, direct outreach, tailored CVs, referrals, and building a project to send to prospective employers. These approaches have worked for Indian companies but have not yet produced results with US companies.
An entry-level PM career discussion raises the risk that a genuine one-year Product Management internship, such as at Amazon Europe, can be filtered out when a résumé is not ATS-optimized; it also questions whether removing “Intern” from a title or exaggerating ownership is becoming necessary to compete. One commenter’s counter-advice is to emphasize verifiable signals—education, achievements, internships, scholarships, and personality—instead of inflating experience, while acknowledging that ATS screening has made résumé visibility a “game” and that the first entry-level role is especially difficult to secure.
- In a highly regulated, domain-specific setting, the author reports that newer AI models can create features or fix bugs from specifications with far less manual correction; some developer work that previously took weeks or months now takes a day. The reported bottlenecks are shifting toward writing quality requirements, validating the increased output, and producing designs quickly enough, while refinement meetings increasingly resemble demos of already-built work.
- An emerging AI-era operating model is to specify the desired outcome and context rather than prescribing implementation, reinforce reusable design-system rules, and allow engineers to introduce ideas during execution when they fit the system. For small, low-priority features, one PM copied a request almost verbatim into a grab-bag backlog, had developers produce a preview, then reviewed and edited the result; the PM reports spending about two hours total and shipping the feature, while retaining detailed upfront planning for larger strategic work.
- Human product judgment remains a required refinement checkpoint because AI does not know the users, system, environment, or end goal; refinement verifies that a feature is buildable and can expose critical flaws or better ideas. In vertical SaaS, another practitioner reports that the most efficient teams pair one PM with one engineer to define requirements together and then build quickly, combining complementary product/domain and technical depth.
Hiten Shah cautions that AI models, like employees, can appear average when operating inside a bad system, suggesting PMs should examine the surrounding product and evaluation system rather than attributing performance solely to the model.
- AI product strategy: Hiten Shah predicts that the best AI products will eventually make users forget which model they are using, pointing toward product experiences that abstract away the underlying model.
A 10-year AI/ML engineer at a large technology company is considering a full-time move into Product Management after enjoying a part-time role on an AI product. The motivation is to gain greater influence, strategic impact, and leverage for eventually starting a company, while weighing the loss of hands-on building; the key open questions are whether the trade-off is worthwhile and whether technical expertise creates an advantage or unexpected friction in PM.
- Product taste is audience-specific curation, not an objective quality or execution skill. It means understanding the available possibility space, deciding what matters, and choosing the option that is right for the intended audience; craft is the ability to execute that choice.
- Use AI to expand product exploration without outsourcing product judgment. For interaction or feature concepts, replace “ask → generate → accept” with “frame → explore → critique → explore again → select → refine.” AI can prototype several approaches quickly, but the product builder should retain decisions about direction and selection; the recommended stance is aggressive AI use for exploration and cautious use for decision-making.
- Develop product taste through structured critique and audience comparison. Ask what you like about a product, why it works, which maker decisions were strong, who it was built for, and what you would change; then compare your reaction with the target audience’s. Investigate both audience misalignment—what users value that you do not—and creator misalignment—why something you value is not landing—to distinguish poor concept, weak expression, audience mismatch, or an unrecognized breakthrough.
- Career/interview tactic: “Tell me about a product you love” followed by questions about value, design decisions, target users, and improvements can reveal a candidate’s product taste; hearing candidates explain products also broadens the interviewer’s own perspective.
AI product experience: A strong model alone does not guarantee a good product experience: if the product “loses the thread,” the experience can still fail. Hiten Shah points to the surrounding layer around the model as a major determinant of product quality, making context and continuity important product-design concerns.