Sneha Keshwani

I just finished a deep dive with Shreyas Doshi on Product Sense and I’m still processing the "time warp." Spending 9 hours in late-night sessions should have been a grind, but the density of the content is so high that time just evaporates. It’s rare to find an environment where the energy actually builds as the night goes on. The Q&A sessions were where it got real. Listening to product leaders from every time zone pull apart their actual, messy problems felt less like a lecture and more like a high-stakes war room where there is no hiding behind "it depends" and answers are direct. It’s a rare chance to see how world-class product sense is applied to high-ambiguity, real-world scenarios in practice. Most impressively, it felt like distillation of years of Shreyas' corporate and advising experience and learnings from countless productive and counterproductive meetings/ discussions into a simplified, actionable toolkit for sharpening strategy and making better product decisions. Huge shoutout to Shreyas for the intensity, content and putting this all together!!

33

2 Comments

Excited to share my first post on the Airbnb Tech Blog. I'm really proud of the work our team has been doing on observability, and this post digs into some of it. We faced a major migration of 300,000 alerts to Prometheus alongside growing alert noise. The noise looked like a culture problem. It wasn't. The post covers how we rethought our Observability as Code platform, introducing Change Reports and bulk backtesting to let engineers validate alert behavior at scale before deployment. Development cycles collapsed from weeks to minutes. Alert noise dropped 90%. We completed the migration. And the culture shifted. https://lnkd.in/eixNKqJG

89

3 Comments

As I've been mentoring PMs lately, I'm hearing the same thing over and over: folks have hit a wall. They're not just looking for a new job - they need a new approach to work entirely. In product, we talk about invisible asymptotes when growth plateaus. But I'm seeing this play out in careers now too. When rising tides were lifting all boats, these ceilings weren't obvious. Now that companies are under pressure, the career growth that felt free-flowing has stopped. I started cataloging what I was hearing: "I'm already spending every waking moment on my day job." "Everyone I know is struggling in the same space as me." "My manager only cares about hitting ship dates, not my growth." "I've been pigeonholed - just doing more and more of the same." "I'm having a hard time coming up with big wins for my resume." "I think this role has put my career in a dead end." Sound familiar? After working through these with dozens of PMs, I've identified 6 invisible asymptotes that combine to create career plateaus: 1/ Bandwidth - don't have time to think 2/ Network - don't have connections 3/ Mentorship - don't get guidance 4/ Learning - don't get to expand skills 5/ Storytelling - don't know how to brand 6/ Opportunity - don't know options The key is framing these as problems within your power to affect. When I hear "my boss doesn't care about my career growth" or "this place is just a feature factory," those are valid frustrations. But solutioning requires control. I prefer to ask: what can you actually change? Each of these asymptotes shows up differently for different people. Maybe you have bandwidth but no network. Maybe you have mentors but can't structure a learning plan. Maybe you've learned a ton but don't know how to showcase it. The good news? Once you identify which ceiling you're bumping into, you can start chipping away at it before you desperately need to. What invisible ceiling are you bumping into right now?

19

2 Comments

How I built my grocery list app PRD using Claude (and why I let it interview me) After last week's post, some of you asked: "So how are you actually doing this?" Here's where I started: the PRD (product requirements document for my non-tech friends). I remembered Evan Eisert's presentation at a VentureFizz event where she talked about writing your PRD in one tool, then building in another. That separation felt right (and a less expensive way to do it). My first attempt: I dumped my initial requirements into Claude—interaction modes, item fields, crossed-off items, meal planning. It was a start, but surface-level. Then I tried something different. I asked Claude to interview me about the product—one question at a time. And it worked. And it was easy. Here are some actual questions Claude asked: "Categories: You mention sorting by category—how are categories assigned? User-defined, auto-suggested, or a predefined list?" "Store field format: Free text entry or a predefined list of frequently used stores?" "When multiple users share a grocery list, should changes sync in real-time, or is periodic syncing acceptable?" "Can users maintain multiple separate grocery lists (like 'Weekly groceries,' 'Party shopping,' 'Costco run')?" One great aspect of this was how Claude asked follow-up questions when my answers were vague. My favorite part though, was when it asked me a question and I wasn’t sure how to answer, so I asked it for advice on how to answer its own question. 😎 The result: A 5-page PRD with clear decisions on a bunch of features. I knew exactly what to build first (and what to save for later iterations). When I moved to Replit to build, I had real requirements to code against. Next week: What happened when I took this PRD into Replit and started actually building. Have you ever tried using AI to interview yourself? What for and how did it go? Link to last week's post, the first in the series, in the comments. #ProductManagement #AI #BuildInPublic #VibeCoding

30

9 Comments

Specs don't compress ambiguity anymore. Working prototypes do. A year ago a PM would write a six-page PRD and trust the team to interpret it. Now they open Claude, build the thing in an afternoon, and hand over something a designer and an engineer can argue with on Monday. The doc is still there. It's shorter, and most of it is generated upstream: a customer proposal turned into structured requirements by an agent, before a human touches it. Our PMs' jobs are to pressure-test those requirements and build the prototype that proves they're real. This matters more in regulated industries. When you're shipping AI into healthcare or insurance workflows, the gap between what the doc implies and what the model actually does on a real case is where projects die. A prototype closes that gap in hours. A spec review closes it in weeks, and usually closes it wrong. What I've noticed leading an AI-native PM team: PMs who prototype move faster than PMs who write more carefully. The artifact is the conversation. Engineers stop asking clarifying questions about edge cases when they can run the edge case. Legal and compliance stop reviewing the paragraph and start poking at the screen. Our PMs' jobs are shifting. The writing that matters now is the prompt, the eval, and the failure case. A properly tuned agentic development workflow plus a real set of contextual PM skills (a post for another time) does more than any 12-section doc ever did. If you're hiring product leaders for an AI company and still screening on PRD samples, you're screening for the wrong decade.

9

Post #5: “Migration Lessons” (an off-topic detour from my “post once a week for 10 weeks” 2026 challenge) This past week I started at Pinterest following the tvScientific acquisition. I’m onboarding to new tools, figuring out new processes, and learning new words like “Pinformation.” All these changes can feel unsettling, and they reminded me of a previous chapter in my career where I led the most complex migration I’ve ever experienced. We migrated to: • A new database, with a new data architecture and reporting stack • A new ad engine and serving layer • And just to top it off, a completely new UI What I’m feeling right now is probably how my team members felt when I was at the helm of that end-to-end migration: everything changed for them in a very short time frame. Here are my top three unexpected lessons from that experience (I’ll skip the obvious ones like “plan for delays”). Lesson 1: Reporting changes are more destabilizing than I gave them credit for. We tackled the unbelievable task of migrating our entire reporting stack, logging, database, architecture, the works. And then we made a small “improvement”: we changed a graph from daily to hourly. I thought we did well. We gave higher granularity. It backfired. The visual story changed; People were used to a certain slope, they were unable (or unwilling) to adjust to anything else. To be clear, the data didn’t change. But the visual story did. I learned that people anchor on shapes, they internalize the curves they work with, they need it for the data to feel trustworthy. Once trust in numbers was questioned, every conversation became harder. Lesson 2: “Good enough” is the real bottleneck. I hit this wall many times. Is the logging accurate enough? Is the new UI stable enough? Can we flip the switch to the new ad engine? The definition of “good enough” determines the pace of the migration more than engineering velocity does. Unlike PM interviews, where decisions are neatly data-driven, real life is grayer. Some areas can tolerate slack. Some absolutely cannot. Knowing the difference is leadership. Lesson 3: The biggest risk is usually hiding in the area no one wants to own. The messy reconciliation logic. The legacy integration no one fully understands. The stakeholder group that’s “fine for now.” An old mapping table no one wants to touch. You can almost feel the avoidance around it. In one retro, an exec said: we need to swallow the frog. No more dancing around it. Just dive in. This is not a groundbreaking lesson, but it changed how I approach things at work and in life. Don’t avoid the friction. Let it point you to where your focus should be. To close this rather sentimental post: I'm not keen on handling another mega migration like that one, but the experience shaped how I approach change far more than any launch checklist ever could. I’m wishing myself and the rest of the tvScientific team an easy transition 💚

18

8 Comments

I've worked for many startups over the years, and many times reporting directly to VP/President/CEOs. Particularly in startups with founders as the heads, this article very clearly explains how you should approach working with leadership in these types of organizations. This is terrific to read if you're currently interviewing. https://lnkd.in/g-Np-J7u

3

Any other PMs out there who've become half PM, half AI enabler in the real life (not the title change, YET)? Three weeks ago I dropped my ego about being a "good traditional PM" and stopped asking "is this my job?". Then when I saw this, I felt very seen.

15

2 Comments

I wrote my first PRD at Google in 2007. I’ve likely already written my last. Not because I've stopped building. But because the way we build products has changed. I recently spoke with Randy Silver and Lily Smith on The Product Experience podcast about what’s replaced PRDs for me. It turned into a really thoughtful conversation on how we’re building at Floto, and what this shift actually means for product teams. Worth a listen if you’re thinking about how your own product development process is evolving. Link in the first comment. If you want a TL;DR, here are the key takeaways: The product is now the experience layer. When you're building an agent, the UI is no longer the product. The real work is defining the experience on top of the agent. That requires a different kind of spec. I'm calling it the Product Experience Document (PXD). Invert the process when engineers are faster than the spec. At Floto, engineering goes first. PMs & designers come in after the prototype to shape the experience. The old flow—spec, design, build—breaks when capabilities evolve faster than you can document them. Synthetic personas are useful early, but limited. They reflect average behaviour, not edge cases. The most useful technique: ask negative questions. “What would make you pause?” is sharper than “Why would you click?” Discovery principles haven’t changed. The best products still come from talking to real people. AI helps you scale signal (Reddit, G2, Perplexity), but the gold standard is still direct conversations. The PXD is written for agents, not just people. It covers: why, success criteria, experience principles, example interactions (good/bad/ugly), critical moments, and success metrics. Non-deterministic systems need ranges, not requirements. You can’t specify a single correct behaviour. The PXD defines acceptable ranges, hard stops, and anti-patterns—because guardrails matter as much as ideals. Product sense is now universal. When everyone is shipping code, everyone needs to understand the user experience. We’re now testing product sense across every role. I've also included the PXD we used for a real agent, so you can see what this looks like in practice. If you’d like a copy you can use as a template, just comment “PXD”—happy to share!

24

10 Comments

All PMs and designers at Check now read and write code, including folks that have a non-technical background and had previously never written a line of code in their life. And I feel like we’re moving faster than ever as a result, leading to customer feedback like this Slack message we received. How did we get here? As a software engineer by training, I’ve always felt like I had an edge as a PM. Being able to dive into the codebase, or even just set up a project to run locally, tightens feedback loops, helps me get to the root of problems more quickly, and leads me to facilitate better discussions with engineering teams. And if I feel sufficiently compelled by a customer pain point, I can always just fix the darn thing myself. But that edge is evaporating. The AI tide is lifting all boats on our team, and now PMs at Check regularly use Cursor and other AI tools to explore our codebase, and answer questions for themselves about the behavior of our systems. By no means is this a substitute for having team members who do understand the system inside and out — but when it comes to self-serving answers to straightforward questions about product behavior, being able to get an instant answer from the codebase instead of shoulder-tapping an engineer helps the whole organization move a lot faster. The thing that gets me the most excited though is the ability PMs now have to write code and build features. Recently, a PM on our team noticed an acute customer pain point and recognized it could be solved in a straightforward way. The MR itself was probably 20 lines of code. Using Cursor, he spun up a working prototype, and then worked with an engineer to land the change. The result? A happy customer and a data point that PMs who can take a feature from discovery to definition, through to implementation, testing, marketing, and release may become standard as the tools we use to build product become stronger.

71

2 Comments

“Bold. I love it!!!” That’s what I wrote in my interview notes when a designer pulled up their Figma file mid-interview, unprompted, and walked us through their work. It stood out. Recently, I read another story: someone landed a role by showing up with a working mockup instead of slides that sparked the usual debate: Is it too much? Does it have to be polished? What if the culture isn’t ready for it? Here’s my take: it’s always a risk/reward decision. I’ve taken that risk too. Years ago (long before today’s AI tools), I started whiteboarding my answers in an onsite interview when no one asked me to. But I also remember the panel walking in and asking each other, “So… how do we do this?” 🤦‍♀️ In that moment, I decided to take the risk. Partly because I had little to lose, and partly because I knew that playing it safe wouldn’t make me memorable. When the stakes are high, doing something bold isn’t always about desperation. Sometimes it’s about recognizing that unless you shift the dynamic, you blend in. Showing your work can elevate you. And yes, it can also fall flat. It won’t pay off every time. But if you want to stand out, sometimes you have to be willing to take the risk. Have you ever done something unconventional in an interview to stand out? How did it land?

7

Today’s tip isn’t about how to build a product; it’s about how to get the job so you can build one. After sitting through dozens of PM interviews over the last few years, I’ve noticed three recurring mistakes that separate the "good" candidates from the "hired" ones. If you’re interviewing right now, keep these in mind: 1. The "We" vs. "I" Trap 🙋‍♂️ Product management is a team sport, and we are trained to say "we" to credit our developers and designers. However, in an interview, "we" is anonymous. If you only use "we," the recruiter doesn't know if you made the decision or if you were just in the room when a C-level executive made it. The Fix: Be specific about your unique contribution. What did you facilitate? Which conflict did you resolve? Own your impact. 2. Know Your Numbers (The Impact) 📈 You should have your metrics ready before you even log into the call. What actually happened after your feature launched? Did it move a North Star metric? How did it solve a specific user pain point? Without clear data, your stories are just output without outcome.Impact is what makes your experience believable. 3. Radical Honesty > Overselling 🤝 If you were at a company for 6 months, don't claim you rebuilt the entire product and generated millions in revenue. We know that onboarding takes time; we know how reality works. When you exaggerate, you lose credibility. Bottom Line: Show your impact, own your contribution, and stay grounded. A great PM is a mix of humble collaborator and confident leader.

9

Startup CEOs: With very few exceptions, the first PM you "hire" at your company should be an engineer or designer who's already on the team. I wrote the article below a decade ago, and this is 10x more true today than it was then, due to the product, design and engineering roles increasingly converging. Article: https://lnkd.in/gXYN5ETj

166

9 Comments

My Startup Was Working… Until It Wasn’t We had paying customers. We had local press. We were gaining traction. It was 2016. Then UberEats, DoorDash, and Postmates entered Portland with billions in funding. At the same time, Intel—where I worked full-time—announced massive layoffs. Overnight, my 40-hour week almost doubled. Midnight fab shifts, weekend work, endless overtime. My startup didn’t die because the idea was bad. It died because I didn’t have the financial runway to compete. Here’s what that painful chapter taught me: 1. Passion + grit aren’t enough. Startups need time, capital, and breathing room to survive. 2. Corporate is just built for stability. It can’t nurture the creative chaos that startups need. 3. Adaptation beats stubbornness. Sometimes the smartest move is to regroup, save, and come back stronger. So I made a choice: doubled down on my W-2 career, moved to Seattle, and joined Amazon to maximize income. Every dollar earned—and every loan I could pull—went straight into real estate. One property a year turned into two… then into double digits in a single year. Deals kept coming, I kept closing, and I built a local team to keep momentum alive. By the end of 2022, my portfolio was generating more passive income than my paycheck. That’s when WoolooAI.com was born—not as a flashy startup, but as a necessary system to simplify and automate the chaos: acquisitions, CRM, portfolio management, project management, all streamlined. The truth is: - Without the grind of my early startup days, I wouldn’t have landed the Amazon role. - Without Intel’s layoff threats, I wouldn’t have learned the importance of building a financial runway first. - And without the expertise I developed at Amazon, I wouldn’t have gained the skills that helped me build WOOLOO AI. The irony isn’t lost: the same challenges that once crushed me ended up shaping the freedom I enjoy today. 💡 Takeaway: Sometimes what feels like an ending is really just the foundation for the next chapter. 👉 If you want the full story, I wrote it here: https://lnkd.in/geEVye36 💬 I’d love to hear your story too—what twists or setbacks ended up shaping where you are today? #Entrepreneurship #Startups #RealEstateInvesting #AmazonAlumni #FinancialFreedom #LessonsLearned #HumbleGrowth

226

8 Comments

Everyone said AI would level the playing field in product. It did. Just not in the way anyone meant. A year ago, a well-crafted PRD, a tightly structured strategy doc, a compelling research synthesis - these took time. The PM who could produce them consistently had a visible edge. The quality of their thinking was legible in the quality of their output. AI collapsed that gap overnight. Now everyone has a polished PRD. Everyone has a structured strategy doc. Everyone has a research synthesis with the right headers and the right framing. The artifact that used to signal capability is now table stakes. It tells you nothing about the person who produced it. So what's left to differentiate one PM from another? Influence. Narrative. Relationships. Executive visibility. The ability to walk into a room and make people feel like following you. The least meritocratic parts of the PM role just became the only parts that matter. The PMs who were already politically strong got stronger. The ones who relied on the quality of their thinking being visible in the quality of their work lost the thing that made them visible. We've done this before. Every time a tool democratizes output quality, power consolidates around the people who were already winning on the dimensions that can't be automated. AI didn't change that dynamic. It accelerated it. The uncomfortable question nobody is asking: if the artifact no longer signals the thinking, how does a product organization know who its best thinkers actually are? Most can't answer that. So they default to whoever is most confident in the room. That was always partially true. It's now almost entirely true. #ProductManagement #AIinProduct #PMLife

6

2 Comments

My next guest, Eno (co-founder of Factory), built this insane product management AI skill that blew my mind. It includes: → Product principles and positioning → What an 11-star experience looks like → PRD templates and review rubrics → Prioritization frameworks ...and more It's 700+ words and I'll admit it made me a little worried about my profession 😅 Watch the full interview now: https://lnkd.in/gapCAhf4

279

10 Comments