We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Canva's restructuring as a PM job description
Canva CEO Melanie Perkins told staff that roughly 22% of the workforce would be affected. She cited growth slowing "for the first time in a decade" and said Canva needs to "refound" itself around small, empowered teams . The changes:
- Product, growth and platform merge into one product org.
- Narrow sub-specialties fold into broader roles, and management layers collapse.
- Sydney becomes the main home for more product teams .
Mind the Product's read is that many current layoffs come from slower growth, not AI alone. Companies are reorganizing around fewer bets . It cites US tech product headcount down 28% from the 2022 peak. Managers are down 31%, VPs 38% and senior ICs 21%, while product openings are up 75% from the late-2022 low . Its conclusion: the PM may soon own an outcome "from first idea to the numbers on the dashboard." The coordination work in between (status updates, first-draft specs, meeting recaps) is what AI handles well, so judgment is what's left . Two exercises it suggests:
- Write the one outcome you want to own in a sentence or two .
- Prompt Claude or ChatGPT to do your product's core job. If the result is 80% as good, rethink your moat .
It also notes that AI features add running costs, which makes what's included, metered or charged a product decision .
Salesforce on build vs. buy
On a16z, Salesforce's M&A lead said acquisition ideas start from customer feedback on product gaps . Each deal is judged on three things:
- strategic fit and the customer success it can drive
- whether it speeds up the business, fits the tech stack and fits the culture
- value against valuation, which she called the hardest part right now
Build vs. buy comes down to whether product teams have, or can hire, the people for their three-year roadmap. That gets compared with the 6–9 months a typical acquisition takes to integrate .
Salesforce says it is moving toward outcome-based pricing, such as paying per resolution instead of per seat . When it piloted Fin, the AI support agent it acquired, on its own help site, it was running in 14 days. It now handles over 40% of traffic and resolves 65% without a human . For founders, she warned that ARR "masks attrition." High attrition often points to product-quality problems and is a red flag in diligence .
Boom: proving product-market fit without shipping
On YC's channel, Blake Scholl argued that software proves product-market fit by launching and growing. You can't test demand for supersonic flight that way, so hard tech needs more money before it reaches a proof point . After the Rolls-Royce engine deal fell apart, Brian Chesky's crisis advice gave him the confidence to build the engine in-house: become more of what you were meant to be, rather than compromise . That engine work then pointed to a first-revenue product. Turbines are a bottleneck for AI power, so Boom spent two months confirming the market and the technical plan before going all in . With the fan removed, the engine produces 42 MW, a size that suits modular data-center deployment .
Practitioner tactics
- PMM roadmap influence. Experience varies a lot by company. Some PMMs can only relay customer and competitor input . Others have moved features up the roadmap when deals stalled or renewals were at risk . Quantifying lost revenue from churn, stalled deals and competitive losses helps get action . One commenter's advice is to learn how ideas actually become PRDs .
- Customer-funded development. Treat payment as one input, not a priority override. Write a one-page exception brief covering strategy fit, full funding including maintenance, the roadmap work that will slip, and a kill date. "One large cheque shouldn't quietly create a permanent second product" .
- Loss feedback. Ask for a ten-minute debrief during the evaluation, before the decision. After a loss, the request looks like another sales attempt. Also ask new customers what almost sent them elsewhere .
- Cryptic bug reports. Log device, session and user metadata in the background. Use complaints you do understand to build a profile, then find affected users who never complained. One PM turned 28 complaints into evidence that a bug affected every iPhone user on AT&T .
Also
- Paul Graham told a startup to sell to its YC batchmates, as a real test of the product and the best input to how it evolves. It has now crossed 10% of the batch .
- Ryan Hoover, at Product Hunt's 13th anniversary, says 3x as many makers launch there as a year ago and about 25% of unicorn founders are on the site .
- Canva said roughly 22% of its workforce would be affected after growth slowed for the first time in a decade; CEO Melanie Perkins described a “refounding” that brings product, growth, and platform together, reduces specialization and layers, and gives smaller empowered teams broader ownership. More product teams will be based in Sydney, with relocation and office-attendance changes. The PM role signal is end-to-end ownership of an outcome—from first idea through dashboard results—with less coordination work and more emphasis on judgment about what to build, cut, or prioritize. The presenter argues that AI is not irrelevant, but many recent layoffs reflect slower growth and organizations reshaping around fewer bets.
- Data cited in the video put U.S. tech product headcount 28% below its 2022 peak (managers down 31%, VPs 38%, and senior individual contributors 21%); product job openings were reported 75% above the late-2022 low, indicating shrinking layers alongside a recovery in openings.
- To assess AI-related moat risk, the presenter recommends testing whether general AI models can perform a product’s core user job and reconsidering differentiation if the result is about 80% as good. He also warns that AI features add costs, making decisions about what to include, meter, or charge for part of product strategy.
- During the Rolls-Royce breakdown, Boom’s CEO applied Brian Chesky’s crisis principle—become more of what the company was meant to be, rather than compromise—and chose to develop its own engine. Controlling the engine design later let Boom optimize for boomless cruise, which the Rolls proposal could not support.
- Hard-tech product-market fit needs a different proof strategy from software: demand for supersonic travel cannot be tested by shipping early, so reaching a proof point requires more investment before first revenue, even if hard tech is not necessarily more capital-intensive overall.
- Boom found an adjacent product by matching its engine capability to the reported turbine bottleneck for AI-related power demand: the CEO tested interest in a 40-MW gas turbine with Sam Altman, whose response connected him to a team member; Boom then spent two months checking the market and technical plan before committing. The adapted engine could produce 42 MW and was suited to modular data-center deployment.
- A cabin layout the team preferred would have cost 1,000 miles of range on a 4,000-mile aircraft, so Boom explored the trade-off rather than abandoning the design. It simulated a dozen engine variants with its Makeboom software and found a clear winner that recovered most of the lost range through joint aircraft-and-engine optimization.
Shah sees AI as a major shift but warns that much current investment may not earn a decent return; he also points to data-center expansion drawing community pushback over electricity and infrastructure, raising both business-viability and external-impact questions for AI products. He flags trust as a concern when systems trained on an internet optimized for what spreads fastest deliver confident answers, and sees an opportunity for personal AI that does not require users to surrender their entire digital lives. He also warns that AI could produce more sameness and “industrialization of mediocrity,” while asking what happens to users’ taste and judgment when machines make many small choices.
Casey Winters reports an Uber Eats driver picking up his order, driving away from the office toward the East Bay and Marin, and then heading toward Sonoma; support said it could do nothing. He sarcastically frames the episode as part of the “AI support wave,” but does not specify whether AI handled this particular case.
Hiten Shah announced a “Coding Agents 101” session for people who don’t write code, using tools he recommends and real work from his website—a product-adjacent signal of coding-agent workflows being introduced to non-engineers, though the post gives no specific tools or product-management practices.
An AI-first teleradiology startup in India positions AI as a radiologist productivity layer: the founder says doctors proofread or edit and sign every report, while the company takes liability. The founder also frames AI as extending radiologist hours despite a stated 2–5% error rate, which they expect to address through process.
A commenter identifies liability, night/weekend volume, and integration with hospitals’ existing PACS as adoption concerns, warning that painful setup could deter switching even if reports are faster. The founder says emergency time-based pricing is still being worked out and weekend volume remains unresolved.
An India-based Senior AI Product Manager with 10.6 years of experience across technology, product delivery, AI, data, and financial services reports being onboarded to Toptal. They initially entered a rate of $40/hour and are considering $70/hour, but had requested a rate consultation; this is a contemplated rate, not a reported client engagement rate. One commenter says clients they encountered valued having shipped similar work more than location, and that an AI-focused profile can help attract interviews—anecdotal guidance, not a platform-wide benchmark.
- Technical founders are advised to validate with a real first customer rather than build for a hypothetical prospect, uncovering workflow nuance, regulatory requirements, incumbent-tool gaps, adoption friction, and whether the buyer will pay enough for the solution.
- A suggested delivery loop is to avoid overengineering: ship the minimum product needed to make its value clear, offer a paid pilot, validate the distribution channel and solution hypothesis, then use traction and feedback to expand or pivot before investing in full-scale engineering.
- For discovery calls, one practitioner recommends asking how customers do the job today and where it hurts, skipping demos when unnecessary, and recording customers’ exact words to inform landing-page copy.
Typeahead promotes inline AI-assisted Slack replies: it suggests the next words at the cursor, which users can accept with Tab, as an alternative to opening ChatGPT to draft a message.
A solo founder building a mental-health startup sought a partner with marketing, product, or clinical experience; after contacting about 200 people and holding roughly 20 high-quality meetings, they found a clinical partner through their personal network. They report AI helped identify profiles but was poor at outreach; personally written, researched messages got a 30% response rate, and they recommend using conversations to refine the pitch and get feedback.
A commenter cautions against treating a cofounder as an easy shortcut, arguing for thorough due diligence rather than rushed onboarding and pointing to informal, unwritten equity splits as a risk.
The post raises a product-strategy scenario in which frontier-LLM coding tools could make most complex software inexpensive to reverse-engineer, weakening implementation cost and code protection as competitive moats; it questions whether B2B buyers could instead build products themselves. In that scenario, differentiation might rely more on distinctive solutions to established problems or creating new categories, while brand, trust, network effects, data, and calibration or training loops could protect incumbents—advantages greenfield products may lack initially.
For vague or sparse bug reports, collect diagnostic logs and device, user, and session metadata in the background; use recognizable complaints to identify patterns and search for other affected users, including people who never reported the issue. This can turn cryptic feedback into reproducible QA scenarios and give product teams evidence to prioritize beyond the complaint count; the commenter illustrated this with 28 complaints but a data-indicated cohort of iPhone users on AT&T.
Digital-service benchmark: Daniel Ek asks why 30-minute company formation is not the European default. The linked post says Estonia lets companies be set up online in 30 minutes and leads the EU in startups and unicorns per capita.
- Salesforce’s product opportunity and build/buy/partner analysis starts with customer feedback on product gaps and a three-year roadmap review. Acquisition criteria include strategic fit and customer success, speed, technology-stack fit, talent and cultural alignment, and value relative to valuation; the decision also compares a typical 6–9-month acquisition integration with how quickly the capability could be built internally.
- Salesforce says it is moving toward outcome-based pricing, such as payment per resolution rather than per seat. It reports that its acquired AI support agent Finn was running internally within 14 days, handling over 40% of support traffic and resolving 65% without human intervention.
- Salesforce positions Slack as a conversational interface for AI and agents, and as a way to access its products; it says Slack revenue is nearly three times its level at acquisition and Slack is now its fastest-growing product.
- ARR alone can mask customer attrition; Salesforce’s M&A diligence experience treats high attrition as a possible signal of unhappy customers and product-quality issues, making retention an important companion to ARR.
For win/loss learning, ask prospects during the evaluation to agree to a brief debrief whichever way they decide; after a loss, a “why didn’t you pick us” email can feel like another sales attempt, when they may no longer be engaged. Follow up with the person who ran the evaluation and ask one question they can answer briefly; avoid gift cards because buyers may be prohibited from accepting them. Ask new customers which alternatives they considered and what nearly sent them elsewhere. Try to learn what might cause a loss and to whom or what before the decision; expect few responses if you only ask after losing.
For customer-funded development that competes with roadmap work, treat payment as an input—not an automatic priority override. Assess fit with product strategy, whether the customer covers delivery plus support and maintenance, and what committed work would slip; compare expected margin and learning against that displaced work. A one-page exception brief can capture the trade-offs and a kill date or sunset plan, with explicit executive sign-off. Repeated requests from multiple customers are a reason to revisit the roadmap; one large payment should not quietly create a permanent second product.
For the choice between a hybrid PM/support-manager role in Austin and waiting for a full-time PM role in California, a commenter advises against turning down a concrete offer for a role that may not materialize. They frame the hybrid role as operating leadership—accountability for factory operations and business targets—and say hands-on work with factory leaders can build differentiated experience. As an example of measurable operational impact, the commenter says their on-site warehouse technology team uses floor walks and package-picking to find inefficiencies, and reduced warehouse costs by over $500,000 that year, including through headcount reductions.
- PMM roadmap influence is organization-dependent and often consultative: some practitioners report mainly relaying customer and competitive input or not attending roadmap calls, while others say they have helped add or reprioritize features when deals stall or renewals are at risk, and shape packaging. One practitioner describes the usual scope as refinement and prioritization rather than setting strategy or creating net-new products.
- To improve influence, learn how ideas enter the roadmap and become PRDs, how priorities are set, and which cross-functional stakeholders matter; tailor customer, competitive, or goal-based evidence to decision-makers’ priorities. Quantifying revenue at risk from churn, stalled deals, or deals lost to competitors is one suggested way to build a case.
Boom Supersonic's reported strategy is to finance airliner development by selling gas turbines after losing Rolls-Royce as its engine supplier. Paul Graham notes that supersonic engines resemble power turbines because they must sustain full-power operation in demanding, high-temperature environments.
- A founder building a vertical short-drama app said they stopped posting to focus on the product, after which their YouTube engagement flatlined; a commenter advised treating the existing audience as an asset and continuing to post—even weekly—while building.
- For products competing with large platforms, commenters recommended identifying a distinct wedge the incumbents cannot offer and considering a pivot to an underserved niche; one warned that the market was losing money even for large players, making the same model difficult for a small entrant.