The Growth Podcast

Aakash Gupta

Join 500K+ for deep dives on AI + product management. After spending a decade plus in product, I now interview PM's most insightful experts.

www.news.aakashg.com

Play

Open in

31 episodes
weekly
Avg 1 hr 12 min
English

Episodes 31

Newest Oldest Longest

All20 Unplayed20 In progress0 Played0

Counted on this page — what you have heard stays on this device, so it is not something the list can be paged by.

November 3 · 57 min
What an AI-Native Product Team Actually Looks Like at Salesforce
This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe Brought to you by Fingerprint: Identify good and bad visitors with industry-leading accuracy Ariso: Ship AI agents and features faster, with fewer regressions Customer.io: Send smarter messages using your product data Land your next PM role: Coached live by PM leaders who've hired at Meta, Uber, Microsoft, etc Product Faculty: Get $150 off their #1 AI PM Certification via code AAKASH150 Today's Episode Salesforce and Anthropic announced Claudeforce in August. Now a sales team can ask Claude about its pipeline without ever opening Salesforce. That worries me. LLMs are probabilistic, and you can't have two pipeline numbers. So I sat down with Jayesh Govindarajan, who runs AI engineering at Salesforce. His teams build Agentforce, the foundation customers use to build their own agents. He walks me through how Salesforce gets one right answer out of an agent every time. Then he lays out the ADLC, the life cycle that replaces the SDLC once your software is an agent. If you're trying to turn your team AI native, this is the one episode to watch. 10 Key takeaways 1. Manage agents instead of doing the work - Jayesh says the biggest change in his org is mental. People stopped asking how to do the work by hand. They now ask how to manage a team of agents that does it. 2. Keep tool choice open - His engineers started on Cursor, then moved to Claude Code. Every tool clears security review first, then teams get a choice for a while. Real usage data tells leadership what to back next. 3. The value was never the UI - Going headless puts Salesforce's business logic behind an API. Customers pull decades of tuned workflows into Claude through skills and build whatever interface they want on top. 4. Let deterministic tools hold the ground truth - You describe an agent in plain English and it transpiles to Agent Script, which spells out the tool calls and context lookups. The model orchestrates tested APIs instead of writing fresh code. Hallucination drops sharply. 5. Go eval-first - Jayesh's data science teams were eval native from day one. Now his engineering teams are catching up with custom scorers tied to outcomes and built for each specific agent. 6. Make Slack the context layer - Slack is the one place people are available asynchronously all the time, and that history is gold dust for agents. A PM can mock up a prototype in a Slack thread with a customer and get feedback before a PRD exists. 7. The product trio lines are blurring - Engineers now put prototypes in front of customers directly. PMs no longer wait on engineering. Team ratios haven't shifted yet, but Jayesh already sees work bleeding across roles. 8. AI PMs need prototyping and evals - Jayesh calls these two skills the magic bullet. When he hires, he looks first for agency, then for craftsmanship that refuses AI slop and turns skepticism into failing test cases. 9. Adopt the ADLC - The agent development life cycle covers how you build, test, evaluate, deploy and observe an agent. The model underneath keeps changing, so without evals you are too afraid of breaking things to upgrade. Salesforce's new optimization agent watches every live agent for drift. 10. Start simple, then go wall to wall - Define value and an ROI metric before you build. Minimize variables on the first implementation and get one win. One of Europe's largest recruiting firms started with text agents in 17 languages and is now moving to voice. Go Deeper How to Build a Full AI Dev Team in Claude Code A Company OS on Hermes and OpenClaw AI PM Crash Course with Aman Khan How Zapier's CEO Grades PMs on AI Fluency Listen and watch on Apple, Spotify, and YouTube. If you want to advertise, email productgrowthppp[at]gmail[dot]com. To hear more, visit news.aakashg.com!
- Transcript
October 26 · 55 min
How to steal your way to a hit product, with Mark Pincus (founder, Zynga)
This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe Brought to you by Mobbin: The internet's best library of real design patterns Get a full year of Mobbin Pro, along with 8 other AI tools, at bundle.aakashg.com. Today's Episode I quote Mark Pincus more than almost anyone, and I still feel a little guilty every time I do. For years, the internet calls him a "FarmVillain," the founder who copies his way to the top. Today he owns it. Mark (founder of Zynga, author of Life at the Speed of Play) steps up to my whiteboard and draws Proven Better New, the framework I reach for every time I plan a product. He shows me how Zynga Poker copies the leading tables down to the chip sounds and still wins on one small new bet. Then he makes the case that most AI founders today are being too clever for the moment. If you're building anything consumer right now, this is the one episode to watch. 10 Key takeaways 1. Copying is what every product maker already does - Mark argues that all music, movies, stories and games are derivative work. The real skill is the magic trick: you learn from someone else's proven success and make it feel new. Steve Jobs did it better than Bill Gates, even though both copied Xerox PARC. 2. Trust your instinct, doubt your idea - Mark puts your instinct at about 95% right and your specific idea at maybe 25%. Proven Better New lets you act on both at once. 3. Proven means pixel-level, for this audience on this platform - You copy the experience click by click, legally. "Instagram" can't go in your proven column. Today's best version of the specific experience can. 4. Better is what 10 out of 10 existing users would choose - What you think is better is actually new. Better is usually small: no download, free, fewer clicks, a cleaner onboarding. 5. Isolate your new so you fail for the right reason - Zynga Poker copies the leading tables down to the chip sounds, drops the download and adds one new thing, your real friends from Facebook and MySpace. Mark says friends end up at the same table 25% of the time. 6. Onboarding belongs in proven - Zynga's PMs call social Civilization and SimCity Social dead on arrival because of their long tutorials. In casual games you need to reach the good stuff in three clicks, not three minutes. 7. All new fails - Zynga still repeats this mantra. Expect your new to fail so you kill it fast, and treat new that works in the market with reverence, because it just became proven. 8. Early in a platform shift, ship less new - Supercell's Hay Day launches with 0% new, just a FarmVille built for mobile. In the AI cycle, users sit in discovery mode, so an AI version of something they already do beats an idea that's too clever for the moment. 9. Copy everything by default, then cut on audience - FarmVille copies Farm Town's gifting and its two-way neighbor bar, friction included. It cuts only the stranger marketplace, because that doesn't fit its core audience of middle-aged women on Facebook. 10. Test the top of the funnel before you build - Mark's problem with the MVP is the word viable, because it fills you with hope. Test names and web ads first, and set a click-through bar before you write real code. Go Deeper How you need to think about product strategy in the age of AI (with Tim Herbig) The Ultimate Guide to Onboarding Ultimate Guide: Activation The Multiplayer Lever Ultimate Guide: Growth Loops The PM's Guide to Governance with Eric Ries Listen and watch on Apple, Spotify, and YouTube. If you want to advertise, email productgrowthppp[at]gmail[dot]com.
- Transcript
October 20 · 1 hr 7 min
Claude Code for Growth: Complete Masterclass with James McAulay
This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe Brought to you by Fingerprint: Identify good and bad visitors with industry-leading accuracy Ariso: Ship AI agents and features faster, with fewer regressions Customer.io: Send smarter messages using your product data Amplitude: The market leader in product analytics Product Faculty: Get $150 off their #1 AI PM Certification via code AAKASH150 Today's Episode "We no longer need to wait for engineers to build growth experiments." That's what James McAulay told me. Halfway through our recording, he showed me what he meant, on my own site. He ran one Claude Code skill, and a few minutes later I had a list of SEO fixes I'd been ignoring while my search traffic slid. So I made this episode for every PM and founder doing growth with no team behind them. James led Creator Growth at ElevenLabs, and now he runs his whole business from one folder in Claude Code. I asked him to open it up. He walks through: The goal cascade that stops Claude (and you) from building features nobody asked for An SEO audit skill wired to Firecrawl, DataForSEO and Lighthouse, run live on my site Multi-agent refine, the panel of critic agents he runs on every plan before anything gets built He runs his actual business on every piece of it, and that's why it holds up. If you own growth and never seem to have time for it, start here. Key Takeaways 1. Run growth from one folder - James's growth OS is one folder with three layers: tools with real APIs, skills he uses every day, and weekly routines that improve the setup. If the agent can't read a tool's data, the tool doesn't make the cut. That rule alone keeps the whole stack usable by an agent. 2. Tie every session to your goals - A goal cascade runs from a yearly mission file to quarterly OKRs, monthly and weekly retros, and a daily standup. A session hook shows Claude the mission at the start of every session. When you drift, it pushes back. 3. The weekly retro is the one that matters - James sometimes skips the standup, but never the Friday retro. Connected to his to-do list and GitHub, it caught him refactoring for fun while outreach tasks went stale for weeks. Your gut says one thing, the data checks it. 4. Do the long onboarding - The full setup is an interview with Claude that can run two hours. It feels like a slog. But once your context lives in files, every answer after it gets better. 5. Close every experiment - Each test starts from a template with a hypothesis, baseline, primary metric, confounders and stop conditions. The weekly retro and daily standup flag finished and stale tests. Every one ends in a documented learning instead of a forgotten dashboard. 6. Pull baselines, don't type them - When you set up a test, numbers like baseline conversion and traffic should come straight from PostHog, Mixpanel or Amplitude. James says nobody should be answering those questions by hand. Connect the analytics tool before you design the test. 7. Rebuild your SEO audit with cheap APIs - James's SEO audit skill adds Firecrawl to read any site, DataForSEO for keyword research at a few cents a term, and the Lighthouse CLI for page speed. Hand the report to the agent in your repo and it fixes the issues. What used to take an agency and a backlog weeks now takes hours. 8. Put critics in front of every build - Multi-agent refine sends a plan to 3 to 10 subagents, each with a different lens, then ranks the findings P0 to P3. The "what's missing" critic finds the most. Use Sonnet or Opus subagents so you don't burn through your limits. 9. Define every analytics event - James had Claude write definitions for all 350 of his PostHog events in about two hours. A CI test now blocks any PR that ships without tracking. Clean event data lets any agent or contractor measure your funnel correctly. 10. Ask your agent to improve itself - Ask Claude how to improve its own Claude.md, prompt or skill, and act on the answer. Most people miss this because chat can't edit its own instructions. Claude Code and Cowork can. Related content 1. Podcasts How to build a Company Operating System with Hermes and OpenClaw How to Build Effective Product Loops in Claude Code, with Tyler Folkman This CPO Uses Claude Code to Run his Entire Work Life, with Dave Killeen 2. Newsletters I spent 100s of hours building a PM OS for you How to build product-led growth in 2026 AI Agents: The Ultimate Guide for PMs Where to find James McAulay LinkedIn: https://www.linkedin.com/in/jamesmcaulay X: https://x.com/thejamesmcaulay YouTube: https://www.youtube.com/@JamesMcaulay0 Website: https://www.jamesmcaulay.co.uk Agent Accelerator: https://agentaccelerator.ai Where to find Aakash X: https://x.com/aakashgupta LinkedIn: https://www.linkedin.com/in/aagupta/ Newsletter: https://www.news.aakashg.com PS. Please subscribe on YouTube and follow on Apple & Spotify. It helps! If you want to advertise, please email productgrowthppp[at]gmail[dot]com. To hear more, visit news.aakashg.com
- Transcript
October 14 · 1 hr 12 min
How you need to think about product strategy in the age of AI
This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe Brought to you by Fingerprint - Identify good and bad visitors with industry-leading accuracy Ariso - Ship AI agents and features faster, with fewer regressions Customer.io - Send smarter messages using your product data Amplitude - The market leader in product analytics Product Faculty - Get $150 off their #1 AI PM Certification via code AAKASH150 If you want access to my AI tool stack, grab Aakash's bundle. Today's Episode Shipping features has never been cheaper. Which means shipping the wrong ones has never been more expensive. Most teams have cranked output up 10x, even 100x. Very few can tell you if any of it moved a customer or the business. So I brought in Tim Herbig. He was a PM for eight years before becoming an adviser, coach, and the author of Real Progress. He's worked with hundreds of PMs on making their product strategy actually make an impact. I've been reading his Product Practice newsletter for years. It's one of the few I never skip. In today's episode, Tim walks through: The one test every strategy has to pass The stupidity flip, to check if you've actually made a choice Audience trees, worked through a full marketplace case study Which differentiators still hold when everyone has the same AI models The metrics that tell you whether your strategy worked If you own strategy for your product area, this is the one to watch before your next planning cycle. 10 Key Takeaways 1. Cheap features make wrong bets expensive - AI made shipping fast, so picking the wrong feature now costs more, not less. You still have to get people to use everything you ship. The real question is whether any of that output moved customers and the business. 2. A strategy is only real if it helps you say no - Deck, one-pager or framework, the test is the same. If it doesn't help teams say no to opportunities, it isn't a strategy. A healthy one has most of your team's energy going to work connected to it. 3. Every strategy answers the same core questions - A narrative for why you should act now. A playing field for who you serve and who else competes for them. Winning moves for how you deliver value and differentiate. Canvases and templates are just windows into that messy work, so start with the questions. 4. AI is a resource, not a value proposition - Everyone has access to the same frontier models, so you can't compete on the technology alone. Use AI to amplify the reason customers already choose you. Brand and data are the moats that remain. 5. Run the stupidity flip - Borrowed from Roger Martin. Flip a strategic choice and check whether the opposite is obviously stupid. If it is, you haven't made a real choice. Flip individual components like audience or value prop, not the whole statement. 6. Strategy doesn't have to cover everything - Focusing on paid users doesn't mean turning off the product for free users. The segment choice just has to fit the narrative behind it. Every component has to reinforce the others. 7. Build an audience tree - Land somewhere between too broad and comically narrow. Slice by region, buyer type and value until you can say no. Tim's fictional marketplace lands on repeat buyers of cars worth more than 100K in Germany, and AI speeds up sizing whether that market is big enough. 8. Differentiate on what takes longest to build - Run your winning moves through Hamilton Helmer's 7 Powers. Brand and network effects hold up, while process power and AI features get copied. No one buys a car from you because your marketplace has AI search. 9. Measure behavior change, not tokens - First say in plain English what you'll do differently next quarter, then how you'll know it worked. Tokens used is a value exchange for the model provider, not your customers. Ask whose behavior has to change, and by how much. 10. Tie metrics to your specific choices - Generic KPIs like revenue are too noisy and too slow to tell you if a strategy worked. Pick metrics tied to the exact segment and value prop you chose. Then map every roadmap item back to those metrics. 🎁 Win a copy of Real Progress As a lil treat, I'm giving away 5 copies of Tim's book, Real Progress, in just 3 steps: 1. Subscribe on YouTube and comment on the episode 2. Follow the show on Apple or Spotify 3. DM me a screenshot of both on LinkedIn The first 5 people get a FREE copy! Where to find Tim Herbig LinkedIn: https://www.linkedin.com/in/herbigt/ Website: https://herbig.co/ Where to find Aakash Twitter: https://www.x.com/aakashg0 LinkedIn: https://www.linkedin.com/in/aagupta/ Newsletter: https://www.news.aakashg.com Check out the conversation on YouTube, Apple, and Spotify. If you want to advertise, email productgrowthppp[at]gmail[dot]com.
- Transcript
October 8 · 1 hr 8 min
The Roadmap to Becoming a Product Builder, with Ankit Shukla
Brought to you by Mobbin - Discover real-world design inspiration Get a full year of Mobbin Pro, along with 8 other AI tools, at bundle.aakashg.com. Today’s episode LinkedIn shut down its Associate Product Manager program and replaced it with an Associate Product Builder program. If you’re a regular reader, you know I’ve been writing about this shift for a long time. What’s new is that the market just gave it a full-fledged title. And a title in a job post comes with 3 things: a JD, interviews, and a comp range. So I brought back the one guest I knew would bring the best-case data. Ankit Shukla, founder of HelloPM, ran 12,500 PM job descriptions through Claude and GPT to see what top companies hire for. His AI PM roadmap is still my most popular episode ever. His evals episode is #2. This one goes further than both. 10 Key takeaways 1. The product builder role is real, and the data says so - 12,500 PM job postings analyzed with Claude and GPT. More than 30% now ask for hands-on AI building. The title has not caught up everywhere, but the requirements already have. 2. Judgment is the skill being paid for, not the tools - Across those postings, RAG, agentic AI and prompt engineering were not the number one skill. Identifying the right use case for AI was. The engineering part is what AI can already do for you. 3. The premium is real and it is bigger in India - More than 20% over traditional PM pay globally, with a $195,000 median in the US and senior roles reaching $250,000 to $560,000. Indian AI PM roles carry an even steeper premium over their non-AI counterparts. 4. Start with possibilities, not problems - Traditional product management starts with the problem. AI inverts that, because some problems you never framed as problems at all. You were treating them as harsh realities. Map what AI can do before you map what hurts. 5. Never start with the tool - The biggest takeaway Ankit gives is that POW comes before E. Possibilities, opportunities and workflows first. The moment you open n8n before you have done those three, you are building AI slop with extra steps. 6. The five levels, and why most people stall at level one - Prompting, then reusable Gems and custom GPTs, then skills and connectors, then vibe coding, then production builds with MCPs. Most people live at level zero and one. Level two is where the ROI actually starts. 7. Claude Code versus Codex is a settled debate - Ankit benchmarked both plus Cursor and a DeepSeek harness across e-commerce, social, and gaming builds. All of them shipped functional sites. For 90% of use cases the choice does not matter, so stop optimizing it. 8. The build was 20 minutes of human time and 20 hours of agent time - A production email platform on Node, TypeScript and AWS, built by someone who deliberately avoided the languages he already knew. He pulled the AWS MCP so he never touched the console, then slept while Claude shipped slices. 9. Audit what the agent spends - Claude proposed a $420 a month stack. Ankit knew AWS well enough to ask why it provisioned a machine that large and what Fargate was for. The revised stack came back at $110. The judgment is in the review, not the build. 10. The last 20% is where the job moved - Claude takes you 80% of the way overnight. Bug fixing, edge cases and errors are what remain, and in most companies the PM hands that part to an engineer. The PM's new job is producing the prototype that makes the conversation real. Please subscribe on YouTube and follow on Apple and Spotify. It helps! If you want to advertise, please email productgrowthppp[at]gmail[dot]com. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
October 5 · 1 hr 9 min
n8n CEO: Why we aren’t dead, how to build great AI products, and what I look for in an AI PM
Brought to you by Customer.io - Send smarter messages using your product data Kameleoon - Leading AI experimentation platform Bolt - Ship AI-powered products 10x faster Jira Product Discovery - Plan with purpose, ship with confidence Product Faculty - Get $150 off their #1 AI PM Certification via code AAKASH150 Today's Episode n8n was the most written-about AI tool in this newsletter last year. I covered it 25 times and ranked it A tier, above Zapier and Make. Then Claude Code and Cowork arrived, the Google Trends line bent downward, and a tweet went around declaring n8n irrelevant. So today I went to the person with the most to lose. Jan Oberhauser is the founder and CEO of n8n, last valued at $5.2 billion after SAP's strategic investment. He says n8n has been declared dead roughly a thousand times, including the week OpenAI shipped its own agent builder, which turned out to be one of his best growth weeks ever. His argument is not that Claude Code is bad. He uses it. His argument is that they are different products and you need both. Claude Code is where you prototype in ten minutes. n8n is where the thing goes once it has to run every night, survive a model outage, pass an audit, and be handed to someone who did not build it. So he opened the product and showed me. A live agent build, an eight-minute AI assistant run, an approval gate that stops an agent mid-action, and an execution log that shows every decision the agent made. Then we got into the growth story. 10x revenue in a year. No lead gen target. No per-seat pricing. A goal of a billion users with fewer than a thousand employees. 10 Key takeaways 1. Claude Code and n8n are different products, and you need both - Claude Code runs Anthropic models in your terminal as an agentic tool. n8n is the orchestration layer connecting your tools, models and data sources on a visual canvas. Jan's own users prototype in Claude Code and migrate to n8n once the workflow has to be reliable. 2. Reliability is fallback models, self-hosting and maintained integrations - Every provider goes down, so you set a fallback model inside the same workflow. You can self-host next to your own data. Each integration is code n8n writes and maintains, so an API change gets fixed once for everyone instead of separately by every builder. 3. Auditability is the thing generated code cannot give you - With code you see the input and the output and nothing in between. n8n replays every past execution step by step, with the data going in and out of each node, and lets you rerun half a workflow from a single data point. 4. AI plus deterministic logic plus human in the loop - An if statement is cheaper, faster and 100% reliable, so it belongs next to the model rather than being replaced by it. Destructive actions sit behind an approval gate, so the agent asks before it sends the email or books the meeting. 5. The AI assistant builds the workflow for you - Jan prompted it the way you would prompt Claude Code. It asked clarifying questions, thought for eight minutes, and returned a working multi-tool agent. Extending it with one-on-one scheduling and Google Contacts took another five. 6. Start small, and skip AI when you do not need it - Companies that try to transform everything at once spend weeks building the wrong thing. Jan's favorite example is a company that automated employee password resets and saved multiple full-time employees a year. The small wins are also the ones that get colleagues interested. 7. n8n wins wherever the work is business-critical - Security orchestration, compliance, employee onboarding and offboarding, DevOps, monitoring. The rule Jan gives is that the more reliability and security matter, the better n8n fits, which covers nearly everything that is not a personal use case. 8. Sprinkling AI on top gets 10 to 30%, being in the value chain gets 10x - Adding an AI button somewhere is not a strategy. n8n's bet was that when someone decides to build an agent, they build it in n8n. That choice is what produced 10x growth in a year. 9. Deleting the metrics that would make money faster - No lead gen target and no per-seat pricing, because profitability means n8n can think in years instead of quarters. The company does not push free self-hosted users onto paid hosting either. The internal goal moved from a billion in ARR to a billion users so nobody confuses the mission with money. 10. Talent density over headcount - The target is a billion users with fewer than a thousand employees, which keeps the hiring bar high. Jan wants tinkerers who run home automation and understand scale, evals and reliability. Live problem-solving beats take-home tasks now that AI can do the take-home for you. Go Deeper Build your first n8n workflow with Pawel Huryn’s guide. Then follow Mahesh Yadav’s learning path in How to Become a Builder PM. Prototype your next agent there, and promote it into n8n through the MCP server. Write 20 test cases and run them as an eval. *** You can get my podcasts on Apple, Spotify, and YouTube. *** If you want to advertise, please email productgrowthppp[at]gmail[dot]com. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
September 24 · 1 hr 14 min
How to Get a ‘Transformative’ AI Fluency Rating as a PM, with Wade Foster | CEO of Zapier
Brought to you by Arize - Ship AI agents and features faster with fewer regressions. Get a full year of Arize’s paid plan, along with 8 other AI tools, at bundle.aakashg.com. Today’s episode I wrote a PRD using ChatGPT, with a considerable number of iterations, and gave it to the CEO of Zapier to grade. In 2024, that PRD would have killed in any product review. But Wade Foster called it a little old-fashioned. In 2026, the work that made you look smart in 2024 is now the bare minimum. Shopify, Meta, and Microsoft have all tied AI to employee evaluations. Zapier took it a notch higher and published a whole rubric on AI fluency. Then reintroduced another one just 10 months later! Read the complete update here. Wade co-founded Zapier and took it to a $5B valuation on roughly $1.4M of venture money. His rubric is the one a lot of companies have modeled their own on. So I got him to share his grading structure live. If you want to know where you’d be when your company starts grading, I got you. Key takeaways 1. Basic AI use is the floor, not the bar - Summarizing, drafting, looking things up and generating a basic PRD are table stakes at Zapier in 2026. Everything above that line needs structured, reusable systems for specs, prototypes and research. Wade is looking for people operating at a meaningfully higher level than the tool's default. 2. Adaptive is where the curve should sit - Zapier wants most PMs at adaptive most of the time, with one transformative example to point to. Living at transformative full time means you are tuning your systems instead of shipping. Gaps in the capable band are normal, even for strong PMs. 3. Transformative means an n of one - Wade's bar is that you are the first person doing a new thing and figuring it out as you go. On a team of 30 to 40 PMs, that is roughly one rating a half. He grades his own fluency as adaptive in most places and admits he steals his transformative tactics from his team. 4. You should be ahead of your CEO inside your own discipline - Wade's rule of thumb is that he should be behind his PMs on PM-specific AI fluency. When he spots a technique that is obvious to him and new to the PM, he gets nervous. You breathe this every day and he does not. 5. Never stop at the PRD - A document alone reads old-fashioned in 2026. Fire up a coding agent and build the prototype, even though production code is not the PM's job. A picture is worth a thousand words and a prototype is worth ten of those. 6. Evidence got cheap, so there is no excuse - Sift Gong calls, Zendesk tickets, subreddits, LinkedIn and X before asking anyone for buy-in. AI makes a mountain of sources searchable in the time it used to take to read a handful. The buy-in conversation now happens after you have proof, not before you start. 7. Judgment separates a turbo brain from a slop cannon - AI plus judgment looks like watching a new species get invented. AI without judgment generates so much noise that it is harder to work with than a teammate who uses no AI at all. At least that person is quiet. 8. Label the effort level when you share AI-assisted work - Say whether you skimmed it once or stand behind every sentence. Handing over unreviewed output just moves the judgment onto the reader. Speed is a fine reason to share something rough, as long as you say it is rough. 9. The job shifts from doing tasks to auditing the factory - When agents handle spec, prototype, eval and first-draft code, your job is to ask why the machine produced a bad batch, not to fix widgets by hand. Where the human gates go is the design question nobody has solved. Wade sees teams still scratching their heads on the loop design. 10. The new no-code is code, and the user is now the agent - Wade's daily brief runs on generated code he has never read and does not care about. Zapier's bet is deterministic execution in the cloud, so workflows run reliably without burning tokens or keeping a laptop open. If your software cannot run headless, you are swimming against the current. Go Deeper If you ask me, start by shipping your first pull request. Then learn loops to automate your repetitive tasks. Give your system a memory on Claude Code or Hermes or OpenClaw. Then version everything you build for PMs on GitHub. You can also bring the whole org on the same page with an all-in-one team OS. And once the system is running, build your first eval, as auditing is the whole point here. I’ll see you in the next one :) If you want to advertise, please email productgrowthppp[at]gmail[dot]com. PS. Please subscribe on YouTube and follow on Apple and Spotify. It helps! This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
September 14 · 59 min
Inside the AI Stack of an $8.3B AI Company’s Product Team | Together AI
Today’s Episode 90% of teams have adopted AI, yet 56% of CEOs say they saw no major financial benefit. Both metrics are accurate, and I’m pretty sure you fall somewhere between them. With new AI tools flooding the market every week, it is safe to say that every team has its own AI setup by now. But are any of those setups in sync with each other? I’ve been building this argument in stages. First, Carl Vellotti showed you how to build a personal OS, and Hannah layered it with building a team OS. Jiaona and Mikhail vouch for a full-blown company OS as well. So, today you get a screen share. My guests are the product team at Together AI that raised $800M at an $8.3B valuation and sells inference and fine-tuning to developers. Charles Zedlewski is their CPO who brought Necoline, Pavneet, and Hassan on the call to talk about how to build a shared context repo. Their customers are already agents and their engineers are already agent-first. Their product team had no choice but to catch up. Brought to you by Arize - Ship AI agents and features faster with fewer regressions. Get a full year of Arize, Bolt.new, Airtable, Speechify, Descript, Magic Patterns, Linear, Dovetail and Mobbin at bundle.aakashg.com. 10 Key takeaways 1. Individual productivity can move a company backwards - The team's starting question was not how to make each person faster. It was whether everyone generating unlimited code and content actually added up to progress. Charles called the failure mode flooding your coworkers' context windows, where everyone launches slop at each other. 2. The shared repo holds context and skills, not code - Markdown and YAML files covering customer intelligence, sandboxes, and the output of strategy meetings broken down by mission and milestone. Anything tied to a specific codebase stays out of it. The point is that a PM can read another team's context and draft a real proposal before taking up that PM's time. 3. Skills live closest to the work they touch - If a skill references code inside one team's repo, it stays colocated there. Everything else goes to a personal or shared repo. Test it on a branch, use it a few times, and only push to main once it proves repeatable. Niche ones never get pushed. 4. Shared context is a hierarchy, not a flat pool - The team abandoned the idea that everyone should carry everyone's context. Most people have no motivation to learn the depth of someone else's area. They want the one answer they came for. Some people live at the bottom of the hierarchy, most just traverse the top. 5. The PRD stopped being a gate - Historically it was the document everyone aligned on before building started. Together treats it as a trigger for ideation and problem solving instead. One to two pages, defining the customer problem, a few solution options, and a sample user journey. That is enough to argue about whether the thing is worth building. 6. A prototype replaces the bulk of the long document - A separate skill takes the one pager and produces a prompt for a design tool, and that visual is where the sharpest feedback shows up, from engineering and marketing alike. 7. Discovery collapsed from half a day to five minutes - The research agent pulls from the support platform, the project tracker, and internal docs at once. It surfaced 19 tickets filed in two months, flagged that the feature had been partially built and abandoned, and gave verbatim quotes with sources. The value is not the summary. It is not duplicating work someone already started. 8. Automate execution, keep decisions human - Defining the feature, the API surface area, and the abstraction layer stay hands on. Code writing is the part that runs on its own. The PRD skill is explicitly instructed to challenge the PM's assumptions rather than accept them. 9. Agents are already the majority user, so validate for them - Agent evals spins up a sandbox, gives an agent a real task against the product, and watches. It caught that agents could not find the fine-tunable models page because it was not linked from the quick start. Dozens of docs fixes came out of this. Charles calls agent success the new bar for UX. 10. They refused to oversell the gains - No story points, so no proof, but velocity is up more than 5%. Charles finds 3x claims suspicious, because discovery, debate, and coordination do not get magically better with AI. Costs stayed sane partly through open weight models, partly because optimizing for collective output never produced the runaway token budgets others report. Go Deeper As promised, I put together this section to help you become the best version in your product management journey. If you’re starting from zero on this, begin with How to build a Team OS in Claude Code with Hannah Stulberg, then scale it up with How to build a Company Operating System with Hermes and OpenClaw. For the skills layer specifically, 3x CPO Oji Udezue on the Essential Claude Skills for PMs is the one to read. And if section 3 sold you on agent evals, go to How to Build Frontier-Lab Quality Evals with Daniel McKinnon. Finally, Pavneet’s take on the one-page PRD lines up almost exactly with what Srini Raghavan showed from Freshworks. You can get my podcasts on Apple, Spotify, and YouTube. If you want to advertise, please email - productgrowthppp[at]gmail[dot]com. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
September 4 · 1 hr 8 min
How to Build Effective Product Loops in Claude Code, with Chief AI and Product Officer at JobNimbus, Tyler Folkman
Today’s episode Most PMs have automated something with AI by now. A PRD review. A weekly status update. The problem is that every session starts from scratch. Loops are what fix that. A skill that runs the same way forever is just a skill. A skill that takes the log of what happened and rewrites itself is a loop. Everyone keeps saying loops are the new prompts. Almost nobody shows you how to build one. Tyler Folkman is Chief AI Officer and Head of Product at JobNimbus. In this episode he builds a loop live on screen. He runs a preflight check on his own podcast recording, spins up three prototype variants while answering questions, writes a decision skill from scratch, and closes the loop on camera. He also covers the loops every PM should be running, the hooks that stop Claude from doing real damage, and the point where vibe PMing breaks. Brought to you by Customer.io - Send smarter messages using your product data Ariso - Ship AI agents and features faster with fewer regressions Viktor - AI employee connected to 3k+ tools with every action approved by your team Bolt - Build a complete design system from your code Product Faculty - Get $150 off their #1 AI PM Certification with code AAKASH150 10 Key takeaways 1. A skill becomes a loop when you feed the learning back. A static skill sits there and stays the same forever. The loop closes when you hand the AI the log of the whole session and ask what to improve. Skip that step and you are just running a skill.2. The gate is the most important part of the loop. Fetch inputs, do work, pass a gate, write the artifact. The gate is where you validate the work was correct, and making it deterministic matters more than anything else in the chain. 3. Agents make decisions, prompts wait for you. The difference is not the model. An agent gets a task plus a way to verify itself and executes as far as it can alone. Prompting keeps you in the loop, which caps how far you can scale yourself. 4. Write the first version of a skill by hand. Human authored skills tend to beat AI authored ones because you know more about what you actually want. Once you are on the AI loop it moves fast enough that injecting your own thinking gets hard, like promising you will still pedal on an ebike. 5. In product, the gate is a customer, and that breaks the loop. Code loops run fast because the gate is a test. You cannot lock customers in a room and iterate on them. The workaround is mining the research calls and transcripts you already have to build a cheap first filter. 6. Prototypes are free now, so generate variants instead of one answer. The internal standard is at least three variants per idea. One minimal, one full featured, one creative. Then narrow a hundred ideas down to five before anyone talks to a real customer. 7. Synthetic customers filter, real customers decide. Customer research transcripts loaded into a warehouse let AI inspect a prototype as your customer. It is not a high bar, but it catches low hanging failures fast, which is the whole point of a gate. 8. Write docs for AI, keep human docs to three pages. Part of onboarding is not written for people to read. It is context so the AI can answer questions. Anything a human is expected to read should be one to three pages, visual, and cut down by hand. 9. Hooks add determinism that a prompt cannot. Telling Claude never to delete everything or never to share credentials only works if it reads that instruction today. A hook fires on the bash command itself. Session close hooks can also force the improvement step you would otherwise forget. 10. Ship AI output you have not thought about and it costs you. Passing unreviewed AI work up the chain just moves the effort to someone busier than you. Answering a question in a meeting with what Claude said is the failure mode. Use AI to push your thinking, not to replace the part where you do it. Related Content If you don't know where to start on Claude loops, I got you covered with my ultimate guide on Loops for PMs. And then you can follow it up with The Complete PM Guide to /goal in Claude Code. Tyler mentioned the evolving role of PMs and how it is shifting into a Product Builder role, so you can read all about it in my How to Become a Builder PM deep dive. If you want to advertise, please email productgrowthppp[at]gmail[dot]com. P.S. Please subscribe on YouTube and follow on Apple and Spotify. It helps! This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
August 28 · 1 hr 6 min
How to build a Company Operating System with Hermes and OpenClaw
Today’s episode Today I’m showing you how to build a company operating system, with Mikhail Shcheglov, CPO at OLX Classifieds. After 5 months of continuous building on OpenClaw and Hermes, his entire product team now runs on it. His knowledge graph covers 54% of the company's product, business and customer context. That is a number he tracks as a personal KPI. At that level the agent already makes backlog decisions. Stakeholders pitch feature requests to it before they’re allowed near a PM. It runs his email, his calendar, his recruiting funnel and his design system. He opens his IDE on camera and shows all of it. 2 findings cut against everything you’ve been told. Summarizing your meeting transcripts costs you 20-25% recall, so he stores every single one raw. And letting Hermes write its own skills off repeated tasks produced a 31% accuracy lift in controlled testing. We also get into something heavier than architecture. What happens to PM headcount when one PM covers four domains? And what does a CPO actually screen for now when hiring? Don't miss.... Brought to you by Bolt.new - Ship AI-powered products 10x faster Product Faculty - Get $150 off their #1 AI PM Certification: code AAKASH150 Customer.io - Send smarter messages using your product data Land PM Job - 12-week live course to master the PM job search Viktor - Use $100 in starting credits to get 5x more done with this AI employee If you want access to my AI tool stack including Airtable, Speechify, Descript, Magic Patterns, Linear, Dovetail, Arize and Mobbin, that's $27,000 of value for $150, grab https://bundle.aakashg.com/. Key Takeaways Context coverage is a CPO level KPI - Mikhail tracks what percentage of the company's industry, business model and customer knowledge his agent actually holds. It sits at 54%. That is enough for it to operate like a junior to mid PM and make backlog calls. At 70 to 90% he expects strategy level work. The real problem AI solves is knowledge leakage - A domain expert leaves and takes five years of context with them. Every company has this hole and almost nobody measures it. One store of business, customer, product and technical knowledge closes it, and the better your AI knows that context the more you can hand it. Do not summarize your transcripts - Summarization cost them 20 to 25% recall. You lose the granular detail where the answer usually lives, and you force every conversation into a template it was never shaped like. Store everything raw. Memory needs three layers, not one - A knowledge graph for structure, a vector database for fuzzy retrieval, and raw daily transcripts in MD files. Exact keyword matching fails on most real queries because real questions are ambiguous. The vector layer carries the load. Auto generated skills lifted recall by 31% - Hermes watches what you keep asking for and decides on its own that a skill is worth writing. Tested across five core topics with ten questions each, control group against treatment group. Plus 31% accuracy. Imperatives matter more than prompts - Their rules file runs 700 lines. No fabrications. Think before you act. Facts over guesswork. And a ban on what he calls fake helpful, where the agent can't do the thing so it explains how you could do it yourself. CLAUDE.md stays short, SOUL.md goes long - CLAUDE.md holds under 100 lines and carries the highest priority. SOUL.md runs 800 and sits second. Some of those 800 lines contradict each other and it still produces his most accurate output, because every imperative gets tested against real queries. Make the agent the gatekeeper - Stakeholders are trained to pitch the agent first. It asks clarifying questions, checks the request against priorities already set, declines politely if it doesn't clear the bar, and routes it to the right PM if it does. The org chart is mapped internally so it knows who owns what. Half of PM time is process, not thinking - Weekly reports, stakeholder updates, demos. Delegate that layer and one PM does the work of two, pointed entirely at discovery. He now runs one PM across three or four customer facing domains, and only keeps dedicated owners on monetization and search. Own the agent yourself or lose the advantage - Feedback arrives daily and he pushes changes from his phone straight into the repo. Hand it to an AI ops hire or an engineering team and you keep the tool but lose the speed. Nobody without skin in the game iterates fast enough. Related content I’ve already created all of the resources Mikhail mentioned in this episode. But I never mapped them all to one place. That changes today in this edition. PFA the map sorted by the parts you want to build. Memory first, I Built You Memory for Claude Code, Hermes and OpenClaw Once you've got a memory layer, put it in version control. GitHub for PMs has the 3-repo setup I use for skills and eval The Hermes Agent Guide for PMs is a 20-minute setup and a 30-day rollout with persona and skill templates included For OpenClaw, start with Mahesh Yadav on becoming a Builder PM For the org-wide version, watch my episode with Jiaona Zhang, CPO at Laurel, on building a Company OS My conversation with Hannah Stulberg of DoorDash on building a Team OS is the same thing one floor down, with a free starter kit 👨‍💻 Where to find Mikhail Shcheglov: LinkedIn: https://www.linkedin.com/in/scheglovm1/ Substack: https://corpwaters.substack.com/ 👨‍💻 Where to find Aakash: Twitter: https://x.com/aakashgupta LinkedIn: https://www.linkedin.com/in/aagupta/ Newsletter: https://www.news.aakashg.com/ If you want to advertise, email productgrowthppp@gmail.com. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
August 24 · 1 hr 5 min
How Srini Raghavan helped $3.4B SaaS Giant Freshworks Embrace the AI PDLC
Check out the conversation on Apple, Spotify and YouTube. Brought to you by * Customer.io - Send smarter messages using your product data * Ariso - The AI operating partner for every manager and team * Product Faculty - Get $150 off their #1 AI Builder Fellowship. Code AAKASH150 * Land PM Job - Join me for Cohort 5 starting November 16th * Amplitude - Custom agents that monitor your funnel and file the ticket Today’s Episode Freshworks is a $3.4B giant of SaaS. They’ve been around since 2010. They have over 75,000 customers and 4,000 employees. Their 2026 revenue will be $960M. They’re a colossus. When Srini Raghavan joined them as Chief Product Officer, they had a 6 month release process. Under his tenure, they moved to a 2 week release cycle. They embraced a new way of working powered by AI. In today’s episode, he breaks down everything: * The AI PDLC they embraced * The AI Harness they created in Cursor, so it can run on any models * How the PM role shifted to Product Builder If you’re a product leader, this is a great example of how to become AI-native. If you’re a PM, it’s a great harness (using Grok models!) to learn from. I hope you enjoy it as much as I did: Apple | Spotify. | YouTube I’ve written up the key takeaways for newsletter subscribers as well. 1. The AI PDLC they Embraced How do you go from shipping every 6 months to every 2 weeks? It’s not about becoming AI first! Most teams try that: they open Cursor or Figma Make and start prompting. Srini made the point that it actually all begins with being Data First: When you’re data first, you create the right foundation to actually be AI first. At Freshworks scale, with 300 million end users, you can’t afford a hallucination. For Freshworks, that was three things: * A design system structured so agents can parse it * Coding standards written down explicitly * A single repo as the source of truth They wrapped that all into a system they call Prism. It’s a knowledge hub that knows the product and its dependencies inside out, plus a context hub that passes feature context between phases, and a central library of what they call AI builder artifacts. These are the skills, rules, commands, and agents that describe how Freshworks specifically builds. The AI PDLC sits atop all of that. And it looks like this: It’s the same lifecycle every SaaS company runs from discovery through release, with two changes. First, there’s a governed AI agent working inside each. Second, there’s an evals phase at the end. And that’s how their release cycle went from 6 months to 2 weeks. But even in the episode, there were some cracks. Figma Make skipped a few design system components. Srini noted that those are the places where humans still have a role. 2. The AI Harness in Cursor they created I began in engineering. After 14 years of not touching code, now I’m spending a lot of time in Cursor. Srini is now spending lots of time in Cursor, and he showed us. Everything starts with a slash command, /fw-innit. Here’s what that kicks off: The harness asks for the business unit, Epic ID, and feature team. Then it runs 12 phases, all the way from idea brief through to prototyping and QA. Srini did the whole demo on Grok models! He explained: I used to use Claude, but Grok seems to be working really well, and it’s really really fast. Nothing takes more than 10-15 seconds. And indeed, that’s what we found. If you really need speed, consider Grok. He live built a performance dashboard for their new EX Agent Studio. After he described the feature in plain English, the agent came back with questions a good PM would ask: * How deep should the drill down go? * Who is the primary persona? * What does success look like in 6 months? It then wrote its own SQL against Bel, their Databricks data lake with usage from all 75,000 customers. It assembled this into quantitative evidence in the PRD: It looked across all 4,358 active ITSM accounts and showed its work. At the end of the process is a step called CPO check which reviews the draft the way Srini would. His team named it after him. From there, the PRD goes straight into Figma Make. Because of being data-first, they have a design system built into it. And after a few prompts we had a working prototype: That’s the end-to-end process. No Claude Code or Codex. Just Cursor + any model. 3. How the PM role shifted to product builder Which brings us back to where we started: what roles will execute work like this in the picture? Srini manages 200+ people with titles like PM, UX researcher, and designer. His prediction is: The titles Product Manager, Product Designer, and Engineer will go away in the next 5 years. He didn’t say that to get clicks on Social Media. That’s how he’s running Freshworks. When he joined, Freshworks ran 1 PM and 1 designer for every 10 to 20 engineers. Now, their new workflow is 1 PM to 1 engineer. Some teams have no dedicated designer at all. Here’s why: those titles marked stations on an assembly line. The PM would write the PRD, hand it to design, and design would hand it to engineering. Now, like in section 2, one person can do the whole assembly line. If you want to prepare for this future, I’ve put together the roadmap: Pay close attention to step 8. Since one person can now run this assembly line, that person also owns whether the thing made any money. In the full episode, he also covers how Freshworks’ agent builder works, how CPOs can replicate his playbook more. Check it out: Apple, Spotify., YouTube. Get more of Srini Freshworks is hiring and you can find Srini on LinkedIn. He also generously shared his deck. Since the episode was recorded, Srini actually left Freshworks. With his tenure as CPO at RingCentral, SVP at Five9, and Director at Cisco, I’d expect you see him at another high-powered role soon. Go Deeper * Hear from more CPOs on becoming AI native: Rachel Wolan, Jiaona Zhang * Learn more about becoming a Builder PM: Mahesh Yadav * Explore other Coding Harnesses: Codex, Claude Code PS. Please subscribe on YouTube and follow on Apple & Spotify. It helps! This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
August 6 · 1 hr 4 min
3x CPO Oji Udezue on the Essential Claude Skills for PMs
This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe Whether you’re using a PM OS, or using regular Claude, or using another AI harness (like ChatGPT Work or Codex)… One of the most important parts of any PM’s AI setup is the skills you have. Today’s guest Oji Udezue, CPO at Typeform, Calendly, and Parsable, and he goes into a masterclass of what skills PMs should have, how to build them, and how to use them. Brought to you by: Bolt.new - Ship AI-powered products 10x faster Product Faculty - Get $150 off their #1 AI PM Certification: code AAKASH150 Customer.io - Send smarter messages using your product data Ariso - Ship AI agents and features faster, with fewer regressions Land PM Job - 12-week live course to master the PM job search Get More of Oji LinkedIn ProductMind The open-source skills repo Shipyard Building Rocketships Go Deeper I’ve done extensive testing on what makes a good skill and how to build a PM OS for paid newsletter subscribers (that go much deeper than this podcast). → Get the founding plan to get access. I also have several free podcasts on Claude Code that you may find helpful: my 3 part series with Carl Vellotti, PM OS with Dave Killeen, Team OS with Hannah Stullberg, Company OS with Jiaona Zhang, and Claude Code for CPOs. → To never miss an episode, subscribe on YouTube and follow on Apple & Spotify. Finally, in my live course, I teach you how to do extremely advanced PM automation Claude Code. → Join us.
- Transcript
July 28 · 56 min
How to Build Frontier-Lab Quality Evals with Daniel McKinnon, ex-PM at Meta, Google
Today’s Episode A developer posted this workflow in March, and it is the clearest picture of where PM is heading that I’ve seen all year. Rasty Turek spent the past year building with coding agents, and he mapped how his process changed over that time. He reckons he now spends around 90% of his time on evals. His eval started as QA, and then it became the spec. Great, now everyone agrees evals are important and will become indispensable for PMs going forward. But there is very little on how to write one. That changes today. I’ve now done 6 episodes on evals, and all of them start with an agent that is running and failing. So what do you do on day 0? Daniel McKinnon was a PM on the Llama models at Meta, a boomerang who spent around 7 years there in total. He sat on Facebook’s central AI team for the entirety of its existence. He wrote enterprise evals for Gemini, Llama, and Ray-Ban Meta. His first job at Meta was on the speech recognition team. He had to figure out how to check whether the models were any good. They weren’t called evals back then. But he’s been writing them for his entire career anyway. In this episode you’ll learn: * How to build an eval set from nothing * The floor-and-ceiling method for calibrating * How to score it and make the shipping call Check it out: Please fill out this short survey on PM salaries. 🆓 I’m doing a free webinar Thursday on getting AI PM interviews. Join me: The next cohort of my LandPMJob program starts in August. If you want my 1:1 coaching, sign up. ---- Check out the conversation on Apple, Spotify, and YouTube. Brought to you by: * SerpApi - Get started with SerpApi using 250 free credits. * Product Faculty - Get $550 off their #1 AI PM Certification with code AAKASH550C7 * Ariso - Ship AI agents and features faster, with fewer regressions * Land PM Job - 12-week experience to master getting a PM job * Pendo - The #1 software experience management platform ---- Key Takeaways: 1. An eval is a trivia question for the model - At its core, an eval is a prompt with a correct or plausibly correct answer plus a way to score whether the output is good. It is the clearest way to communicate what your product should do in the AI era. 2. Offline evals catch problems before you ship - Test the model offline against a fixed prompt set before pushing to production. If it fails, you change the model, the prompt, or the approach before real users ever see it. 3. The best eval sits between too easy and too hard - An eval that scores 100% gives your engineering team nothing to optimize. An eval that scores 0% is equally useless. Aim for a 25% to 50% success rate so there is room to run. 4. Old benchmarks are already saturated - MMLU, HellaSwag, ARC and the rest were built for a simpler question-and-answer world. Frontier models now score effectively 100% on them, which is why you have to keep building new evals and throwing away old ones. 5. Writing an eval is mechanical once you understand the problem - Come up with roughly 100 prompts that match the real distribution of tasks. The hard part is not the writing. It is deeply understanding the domain first. 6. Subject matter expertise drives everything - The cystic fibrosis and congenital heart disease evals worked because Daniel understood the genetics, not because of any template or tool. There is no eval template the way there is a PRD template. 7. Modern evals are agentic, not just Q&A - The genetics eval hands the agent a file with billions of variants and asks it to find the cause of a disease. This is a task, not a lookup, and it mirrors how real AI products now work. 8. Find the model ceiling on purpose - The easy cystic fibrosis case gets solved by most models. The harder digenic congenital heart disease case exposes where even strong models fail. Knowing the ceiling is the point of the exercise. 9. Sample multiple times before you trust a result - Models are non-deterministic. Run the same task several times so you understand the real distribution of outcomes rather than a single lucky or unlucky pass. 10. Meta and Google build products very differently - Google is seen as more engineering-led, Meta as more product-led and far more aggressive culturally. Daniel worked on both Gemini and Llama and saw everything from Llama 3 highs to Llama 4 lows. ---- Where to find Daniel McKinnon * LinkedIn * X * Gamow Labs Related content Podcasts: * AI Evals with Hamel Husain and Shreya Shankar * How to Run Evals in Claude Code with Aparna Dhinakaran * Evals are the New PRD with Ankur Goyal Newsletters: * AI Evals for PMs: Everything You Need to Know to Get Started in 2026 * AI PM’s Guide to LLM Judges * AI Evals Explained Simply ---- PS. Please subscribe on YouTube and follow on Apple & Spotify. It helps! This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
July 20 · 1 hr 19 min
The PM's Guide to Governance with Eric Ries, author of The Lean Startup
Check out the conversation on Apple, Spotify, and YouTube. Brought to you by * Land PM Job - 12-week experience to master getting a PM job * Jira Product Discovery - Plan with purpose, ship with confidence * Amplitude - The market-leader in product analytics * Bolt - Ship AI-powered products 10x faster * Product Faculty - Get $550 off their #1 AI PM Certification with code AAKASH550C7 ---- Today’s episode Every startup founder picks up Zero to One. Reads the chapter on Delaware C-Corps. Files the paperwork. Moves on. That paperwork will outlast every product decision they ever make. I sat down with Eric Ries, the man who created Build, Measure, Learn. NYT bestselling author of The Lean Startup. Co-founder of Answer.AI with Jeremy Howard. Founder of the Long-Term Stock Exchange. Dario Amodei called him before Anthropic’s seed round. His new book Incorruptible drops May 26. It is the blueprint for building a company that the financial system cannot capture. He also demoed live how he wrote the book using Solve It, the AI platform from Answer.AI. Not prompting. Not generating. Editing the model’s responses directly. If you are building anything you want to outlast the next funding round, this is the one episode to watch. Check out the episode on Apple Podcast and Spotify. If you want access to my AI tool stack, grab Aakash’s bundle. ---- Key Takeaways: 1. Governance has four dimensions - Compliance is table stakes. Purpose, coherence, and integrity are the three most boards ignore. Companies that nail all four outperform the market over decades. 2. Financial gravity destroys good companies - The unconscious reflex to comply with the values of those who have more than you. Jim Senegal called it heroin. You compromise once and it gets baked into the forecast. 3. Costco's governance fortress is the blueprint - Staggered board terms, poison pills, fiduciary hierarchy. $10K at the Costco IPO is worth $8.7M today versus $151K in the S&P 500. 4. Stone does not enforce itself - Johnson & Johnson carved values into limestone. Asbestos ended up in the baby powder. $10B settlement. Structure protects ethos but does not create it. 5. Mission lock vehicles create 6x survival - A separate entity holding the for-profit board accountable. Novo Nordisk, IKEA, Patagonia, Hershey, Vanguard all use this structure. 60% survival to year 50 versus 10%. 6. Anthropic's LTBT took two years to defend - AI safety experts appoint board seats. The trust gains power as the company hits milestones. Structural protection is why Anthropic can afford to be courageous. 7. Public Benefit Corporations write mission into the charter - Legal permission to pursue purpose over shareholder value. Not the B-Corp certification sticker. A legal structure. 8. LLMs are conformity machines - They produce the center of the outcome distribution. For competitive advantage you must change how you use AI. 9. Solve It enables human-in-the-loop writing - Edit the model's responses directly. 600 test readers, 10K structured comments, Python scripts organizing feedback per chapter. 10. Build Measure Learn works at any timescale - Not about absolute speed. Relative velocity versus your industry convention. The AI labs that release more quickly create decisive trust advantages. ---- Where to find Eric Ries * LinkedIn * Incorruptible * Answer.AI / Solve It * Long-Term Stock Exchange * Lean Startup ---- Related content Podcasts: * AI Product Strategy with Aman Khan * The Marty Cagan Episode on Product Management * How to Succeed as a Head of Growth with Dan Olsen Newsletters: * AI product strategy in 2026 * How Cursor grows * PLG in 2026 ---- PS. Please subscribe on YouTube and follow on Apple & Spotify. It helps! If you want to advertise, email productgrowthppp at gmail. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
July 13 · 1 hr 33 min
Claude PM Masterclass from Former FAANG AI PM Jyothi Nookula
Today’s episode “This one’s too complex. I’m still stuck on ChatGPT.” I get some version of that DM every week, usually right after I publish something on the PM OS or the Team OS. And every time, I feel it, because those guides do assume you’re already up and running. So I made this episode for the person sending the message. Jyothi Nookula has been an AI PM since before that was a title, through Netflix, Meta, and Amazon, and I asked her to take a PM from zero to eighty on the entire Claude stack in one sitting. She walks through: The five-layer Claude stack, so you finally know which surface and which model to reach for and when A chief of staff you build in Claude Code that reads your meetings and quietly learns your org, your people, and your politics The self-improving agent loop she used to beat 30 engineering teams at an internal hackathon, as a PM Everything she shows is something she runs at work, which is the only reason any of it holds up. Send this to the PM in your life who keeps saying they’re behind. Two hours from now, they won’t be. ---- Brought to you by: Hyper Agent: Turn your recurring PM work into reusable agents ---- If you want access to my AI tool stack - Dovetail, Arize, Linear, Descript, Reforge Build, Relay.app, Magic Patterns, Speechify, Bolt.new and Mobbin - become an annual subscriber ($150), and grab Aakash’s bundle. If you want access to my AI PM customizations - PM OS, Job Search OS, and Prompt Library - become a founding subscriber ($250). ---- Key Takeaways: 1. Match the model to the job - Sonnet handles ninety percent of PM work at the best cost. Save Opus for genuinely hard reasoning, and hand fast bulk jobs to Haiku. Defaulting to the smartest model for everything just burns time and money. 2. Stop skipping the knowledge layer - Projects, skills, and memory are what make Claude know your actual work instead of guessing from a blank slate. Almost everyone underinvests here, and it is the difference between a chatbot and an assistant that knows you. 3. A skill beats a prompt - A skill is a saved playbook Claude picks up on its own when it fits the task. It only loads when needed, so it never clogs the context window. Build one once and stop re-explaining the same task forever. 4. Write your skills yourself - Human-written skill files consistently beat AI-written ones. Draft with Claude to move fast, then layer in the domain knowledge only you have. That last step is what makes it actually work. 5. Automate your time based work - A morning brief, a standup summary, and an end-of-day wrap can all run on a schedule while you sleep. You walk in already knowing what needs your attention. It clears the busywork that eats your mornings. 6. Give your automations guardrails - Cap the length, tell them to stick to facts, and never let them hallucinate. Left unchecked, an AI brief will pad itself and invent things. A few hard rules keep it sharp and trustworthy. 7. Build a chief of staff that learns your org - Point Claude at your meeting notes and let it build a picture of your people, your priorities, and your politics over time. Feed it transcripts first, since they carry the richest signal. It compounds into something no generic chatbot can match. 8. Keep that knowledge base on your own laptop - Your most personal work data does not belong in someone else's cloud. When you leave a company, it walks out with you. You keep full control of your most sensitive context. 9. The PM job is changing fast - The ratio is shifting from one PM per eight engineers toward two PMs per one. Building is becoming part of the role, and the PMs who can ship are pulling ahead. 10. Building is easy now, taste is scarce - When anyone can build, the edge moves to knowing what is worth building and what good actually looks like. That judgment is the one skill you cannot download. ---- Related content Github repo: https://github.com/fibbonnaci/ai-builder-skills Podcasts: How to Become an AI PM - YouTube | Spotify | Apple How a VP Uses Claude Without Producing Slop - YouTube | Spotify | Apple This CPO Uses Claude Code to Run His Entire Work - YouTube | Spotify | Apple Newsletters: I Built You Memory for Claude Code, Hermes, and OpenClaw I spent 100s of hours building a PM OS for you How to build a Team OS in Claude Code ---- Where to find Jyothi Nookula: LinkedIn: https://www.linkedin.com/in/jyothinookula/ NextGen Product Manager: https://nextgenproductmanager.com/ Where to find Aakash: Twitter/X: https://x.com/aakashgupta LinkedIn: https://www.linkedin.com/in/aagupta/ Newsletter: https://www.news.aakashg.com --- PS. Please subscribe on YouTube and follow on Apple & Spotify. It helps! If you want to advertise, email productgrowthppp at gmail. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
July 9 · 1 hr 14 min
The PM's Guide to AI Design that isn't Slop
Today’s episode Today I’m showing you, advanced Codex workflows with Meng To, founder of Design+Code. He told me early on that he’s barely touched Claude. He’s been living inside Codex every day since launch, running a setup most PMs have never seen up close. Plan mode and a fleet of 20 agents running at once while he steps away from his desk entirely. Slides, charts, and full brainstorms generated inside the same chat window, reviewed, regenerated, and shipped, all without writing a line of code. We also get into something heavier than workflows. Why PMs are losing their jobs to layoffs right now? And how to future proof your career? Don’t miss.... ---- Brought to you by: Arize: Trace, evaluate, and fix your AI agents before broken behavior ships to users ---- If you want access to my AI tool stack - Dovetail, Arize, Linear, Descript, Reforge Build, Relay.app, Magic Patterns, Speechify, Bolt.new and Mobbin - become an annual subscriber ($150), and grab Aakash’s bundle. If you want access to my AI PM customizations - PM OS, Job Search OS, and Prompt Library - become a founding subscriber ($250). ---- Key Takeaways: 1. Codex is a fleet operator - Meng runs 20 agents at once while stepping away from his desk entirely. Each one works on something different, slides, charts, brainstorms, while he does something else completely. 2. Plan mode isn't optional - Skipping it means paying twice, once to build the wrong thing, once to undo it. Codex returns a full breakdown, architecture, steps, and open questions, before touching anything. 3. A screenshot beats a paragraph every time - It shows the AI what you mean instead of what you think you mean. A simple two-key shortcut drops any window straight into the chat as context. 4. The taste skill is the real differentiator - Without it, AI design defaults to generic. With it, the output looks like something a senior designer with years of experience actually made. 5. Trust is earned in tiers - Read only first, then supervised access, then full access. Skipping straight to full access before learning where the AI tends to fail is how people get burned. 6. HTML beats Figma for speed - Every extra tool is a login, a subscription, and a context switch the AI can't do for you. Keep your blast radius small. 7. UGC won because audiences are tired of corporate polish - A synthetic version of you, used honestly, reads as more human than a generic message. Ten old photos is all it takes to build a digital twin. 8. Technical PMs aren't surviving layoffs because they write code - Meng hasn't written a single line in six months. They're surviving because they're fluent enough to direct a fleet of agents and catch a wrong output before it ships. 9. Meng builds his own tools when nothing off the shelf fits - His own video editor, his own SaaS templates, his own design brainstorming app. The tool built for your exact workflow beats the popular one every time. 10. The bar isn't five star anymore - Five star is just the floor everyone clears by default now. The real question is what six, seven, all the way to eleven star looks like, because that ceiling rises exactly as fast as the floor does. ---- Related content Podcasts: How to Design with AI - YouTube | Spotify | Apple How to Use Codex Like an OpenAI PM - YouTube | Spotify | Apple The Ultimate Guide to ChatGPT Codex - YouTube | Spotify | Apple Newsletters: OpenAI’s Codex is the Best Way to Use ChatGPT I spent 100s of hours building a PM OS for you How to build product strategy in the age of AI ---- 👨‍💻 Where to find Meng To: LinkedIn: https://www.linkedin.com/in/mengto Design+Code: https://designcode.io Aura: https://aura.build 👨‍💻 Where to find Aakash: Twitter/X: https://x.com/aakashgupta LinkedIn: https://www.linkedin.com/in/aagupta/ Newsletter: https://www.news.aakashg.com --- PS. Please subscribe on YouTube and follow on Apple & Spotify. It helps! If you want to advertise, email productgrowthppp at gmail. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
June 24 · 1 hr 7 min
How to Build a Company OS in Claude Code with Jiaona Zhang, CPO at Laurel
Today’s episode Product teams have figured out AI for engineering. The PMs are using Claude. The engineers are in Cursor. If you’ve been reading this newsletter from the start, you’ve seen how the top 1% are using AI to 10x their output. But there are still teams at companies like Adobe, teams in sales, customer success, finance, who don’t have access to any of these tools. You ask them what AI to use for their next task, and you get a blank stare. That gap is the real problem. And it compounds every day. Jiaona Zhang “JZ” has built the fix. She is the CPO at Laurel, which just raised $100M in Series C, and she has led product at Airbnb, Dropbox, Webflow, and WeWork. Today she runs a product team that ships frontend and backend features end to end, without any engineering handoff. In this episode, she screen-shares everything. Laurel’s full Company OS built in GitHub, with skill files for every function from CS to legal to finance. The playbook to agent pipeline that turned 50-page docs into automated workflows. The daily Slack briefing that tells every person exactly what to do and which skill to use when. And a ton more. This is one of the densest episodes I’ve ever recorded. The knowledge per minute is as high as it gets. Don’t miss..... ---- Brought to you by: Ariso - Ship AI agents and features faster, with fewer regressions Bolt - Ship AI-powered products 10x faster Pendo - The #1 software experience management platform Product Faculty - Get $550 off their #1 AI PM Certification with code AAKASH550C7 Customer.io - Send smarter messages using your product data ---- If you want access to my AI tool stack - Dovetail, Arize, Linear, Descript, Reforge Build, Relay.app, Magic Patterns, Speechify, Bolt.new and Mobbin - become an annual subscriber ($150), and grab Aakash’s bundle. If you want access to my AI PM customizations - PM OS, Job Search OS, and Prompt Library - become a founding subscriber ($250). ---- Key Takeaways: 1. Every company has a 1% who are AI-native and a 99% who do not know what to use when. The Company OS closes that gap by encoding the 1%'s workflows into skills that anyone can use when they open Claude.2. Build the ontology before you build the OS. Map every team's work to categories and tasks first. Color-code what should get more human time vs what gets automated. The OS is built from that work map.3. Even the friction of going to a different interface kills adoption. A separate agent tool in a new tab will not get used consistently. Deliver skills and automations inside Slack and email, where people already are.4. When AI adoption is everyone's responsibility, it is no one's responsibility. Dedicate one person full-time to AI Operations. Start with one person who demonstrates value. Every other function will want their own version within months.5. The Company OS turns a 50-page playbook into a set of agents. Write the playbook first. Then audit it. What requires a human? What can be automated? Build the skill files from what remains.6. The captain model replaces the handoff chain. Every feature has one owner end-to-end. The captain is whoever has the most critical skill for that feature's hardest problem.7. PMs at Laurel ship front-end and back-end features. Not just growth experiments or copy changes. Core product features deeply integrated with billing systems and time entry logic. One PM who identifies as a designer shipped one of these end-to-end last month.8. JZ went from hundreds of reports to 5 PMs and 4 designers. They ship more than ever. Adding people adds coordination cost. In a world where one PM can take a feature from discovery to production in a day, large teams cancel out their own capacity gains.9. The new PM interview is a screen share. JZ asks every candidate to show their actual screen. In 60 seconds she knows their level of AI skills.10. The PM fundamentals never changed. Problem space first. Know why and for whom you are building before you build. The speed changed dramatically. What you are supposed to be doing at the heart of it did not. ---- Related content Podcasts: How a VP Uses Claude Without Producing Slop - YouTube | Spotify | Apple How to Build a Team OS in Claude Code - YouTube | Spotify | Apple How to Become a Builder PM - YouTube | Spotify | Apple Newsletters: I spent the last week building an OS in Claude Code I spent 100s of hours building a PM OS for you How to build product strategy in the age of AI ---- Where to find Jiaona Zhang LinkedIn - https://www.linkedin.com/in/jiaona/ Reforge - https://www.reforge.com/profiles/jiaona-zhang Laurel - https://www.laurel.ai/ Where to find Aakash: X - https://x.com/aakashgupta LinkedIn - https://www.linkedin.com/in/aagupta/ Newsletter - https://www.news.aakashg.com --- PS. Please subscribe on YouTube and follow on Apple & Spotify. It helps! If you want to advertise, email productgrowthppp at gmail. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
June 9 · 50 min
How a VP of Product Uses Claude Without Producing Slop | Matthew Wensing, Customer.io
Today’s episode There are hundreds of guides on writing PRDs with Claude. Dozens on running user interviews. Almost nothing on how a VP of product actually uses it. Matt Wensing is VP of Product and Design at Customer.io. They crossed $100M ARR, just shipped an AI agent, and are one of the fastest growing companies in B2B SaaS right now. I asked him to show me his actual documents, his actual Slack threads, and the exact sessions where Claude helped him produce leadership grade output. What he showed me changed how I think about AI for leaders. Claude has the instincts of a brilliant new hire, it wants to deliver before it fully understands what you need, and at the VP level that gap shows up fast. Matt has spent months figuring out how to manage it, and in this episode he shows you everything. ---- Brought to you by: LogRocket - Find the bugs killing your conversion before your users do. I ran a head-to-head eval to see if that's true. ---- If you want access to my AI tool stack - Dovetail, Arize, Linear, Descript, Reforge Build, Relay.app, Magic Patterns, Speechify, Bolt.new and Mobbin - become an annual subscriber ($150), and grab Aakash’s bundle. If you want access to my AI PM customizations - PM OS, Job Search OS, and Prompt Library - become a founding subscriber ($250). ---- Key Takeaways: 1. Take inventory before you open Claude - Before building anything, list every piece of raw material you already have. Zoom recordings, strategy docs, past presentations. The quality of what you feed Claude determines the quality of what comes out.2. Pivot content, do not write from scratch - Claude's best use case is transformation, not creation. Give it two inputs and ask it to reorganize one into the shape of the other. Matt calls this matrix multiplication.3. Build slides first - Build the visual story first. Screenshot the finished slides and feed them back into the same Claude session. Ask it to write a talk track that adds depth using all the context it already has, not one that just repeats the slide.4. Kill eager suggestions immediately - The moment Claude asks if you want it to generate the next thing, say stop. You control the pace. A 200-iteration session with a great deliverable beats saying yes to the first draft every time.5. Start sessions in the abstract - If you reveal the domain too early, Claude pattern matches to the nearest template. Keep it abstract. Build a clean mental model first. Reveal the domain only when the framework holds up on its own.6. Layer complexity in slowly - Start with the simplest version of the framework. Let Claude stabilize on the basics before you add exceptions. Dumping everything in at once produces a lost in the woods experience for both of you.7. AI alignment decks always backfire - When you one-shot an alignment deck, you flatten the problem. Senior executives have spent months living with the real complexity. They feel the thinness immediately, even when they cannot say why.8. Decompose the problem before building anything - Challenge yourself to explode a nasty problem into all its pieces before you touch Claude. Put those observations into the context window first. Then assemble the solution.9. The Slack scanner keeps leaders close to the ground - Customer.io built an AI scanner that monitors dozens of Slack channels and surfaces threads where a product person should be involved. It runs continuously without overwhelming. 10. Chiefys audits your strategy docs automatically - Chiefys is a Slack bot that holds Customer.io's ratified company documents and checks new work against all of them. It flags contradictions and stale documents so nothing goes invisible after you ship something new. ---- Related content Podcasts PM’s Guide to Claude - YouTube | Spotify | Apple How to Become a Builder PM - YouTube | Spotify | Apple We Built an AI Product Manager in 58 mins - YouTube | Spotify | Apple Newsletters How to Use Claude for Work How to Build Product Strategy with Claude Code The PM OS ---- Where to find Matthew Wensing: 1:1 Video Consultation: https://intro.co/MattWensing LinkedIn: https://www.linkedin.com/in/wensing/ X: https://x.com/mattwensing Where to find Aakash: X: https://x.com/aakashgupta LinkedIn: https://www.linkedin.com/in/aagupta/ Newsletter: https://www.news.aakashg.com --- PS. Please subscribe on YouTube and follow on Apple & Spotify. It helps! If you want to advertise, email productgrowthppp at gmail. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
June 3 · 1 hr 7 min
How to Use Codex Like an OpenAI PM | Abhi Muchhal, PM OpenAI (ex-Meta and Nubank)
Today’s episode Six months ago, I told you Codex is the best way to use ChatGPT for PM work. Most of you tried it. Some of you stuck with it and very few of you are running it the way the people who built it actually run it. Today we get that inside look. Abhi Muchhal is an International Growth PM at OpenAI. Before that, Meta, Nubank, and a founder building on the OpenAI API. He is one of the people responsible for ChatGPT’s growth in India, Brazil, and Japan, markets that are now driving a meaningful share of OpenAI’s 900 million weekly active users. He opened his actual setup on camera. The harness. The automations. The prompts that actually work. And the ones that failed before he figured it out. ---- Brought to you by: Bolt.new - Ship AI-powered products 10x faster Product Faculty - Get $550 off their #1 AI PM Certification with code AAKASH550C7 Customer.io - Send smarter messages using your product data Ariso - Ship AI agents and features faster, with fewer regressions Jira Product Discovery - Plan with purpose, ship with confidence ---- If you want access to my AI tool stack - Dovetail, Arize, Linear, Descript, Reforge Build, Relay.app, Magic Patterns, Speechify, Bolt.new and Mobbin - become an annual subscriber ($150), and grab Aakash’s bundle. If you want access to my AI PM customizations - PM OS, Job Search OS, and Prompt Library - become a founding subscriber ($250). ---- Key Takeaways: 1. The harness is what separates Codex users from Codex runners - The connectors, the permissions model, and the skills layer are the three components that make Codex a system rather than a chat tool. Without all three, you are using an expensive autocomplete. 2. Generic prompts hit the wrong data - Abhi's team had separate B2C and B2B tables that both matched "tell me about weekly active users." The generic query returned the wrong answer every time. Specificity is the skill, name the exact dashboard and the exact metric, looks simple but saves a lot of time when you scale. 3. Three permission levels - Read tasks get full autonomy. Synthesis and drafts get full autonomy. Anything going to another human gets your eyes first. Treating permissions as binary, all control or all autonomy, breaks. 4. The person who cares most builds the skill - One OpenAI growth team built a skill that automates their entire experiment review process. It writes the hypothesis, monitors the run, and prepares the review doc. 5. Real automations run without you - Abhi runs three automations before he opens a single dashboard: a Slack triage, a 9:30AM self-refreshing growth dashboard pulling from 7-8 sources, and a weekly stakeholder update that writes its own first draft. He reviews, makes edits if needed, and sends. 6. Prototype before you document - Build the working prototype first, then write the 10-question companion FAQ. Showing engineers something that runs changes the conversation from whether to build to how to build it. 7. India is OpenAI's second largest market and under 10% of working adults are knowledge workers - The ChatGPT use case that drove US growth does not reach the same share of people in the markets driving the most new users. Building for the world means knowing how different the world actually is. 8. The WhatsApp computer use loop ran in 68 seconds - Point Codex at the WhatsApp desktop app. It reads what you missed, identifies action items, checks your calendar, and types the draft in the composer. One tap to send. Every PM building for international markets should run this workflow at least once. 9. Speaking evals is the key to breaking into a frontier lab - Name a capability you care about. Describe how you would measure it. Say how you would know if the model improved. You do not need 50 evals under your belt. You need to understand why they exist and what a good one measures. 10. Building something real is non-negotiable for frontier lab applications - Abhi had a live Chrome extension running on the OpenAI API at the time of his application. ---- Related content Podcasts: The Ultimate Guide to ChatGPT Codex How PMs Ship 100K Lines of Code at OpenAI Evals are the new PRD Newsletters: OpenAI’s Claude Code Killer AI Agents Guide for PMs How to Land a $300K+ AI PM Job ---- Where to find Abhi Muchhal: LinkedIn: https://www.linkedin.com/in/abhimuchhal/ OpenAI:LinkedIn: https://www.linkedin.com/company/openai/W here to find Aakash: X: https://x.com/aakashgupta LinkedIn: https://www.linkedin.com/in/aagupta/ Newsletter: https://www.news.aakashg.com --- PS. Please subscribe on YouTube and follow on Apple & Spotify. It helps! If you want to advertise, email productgrowthppp at gmail. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript
May 25 · 1 hr 14 min
How PMs Ship 100K Lines of Code at OpenAI with Ryan Lopopolo, Member of Technical Staff
Today’s episode Most companies are still debating whether PMs should ship code. OpenAI is already debating the best ways for PMs to ship code. They’re living in the future. The builder behind a lot of that harness engineering is Ryan Lopopolo. He wrote the OpenAI post on harness engineering and runs a frontier team where PMs, designers, and engineers all ship using the same system. The wild part for me? His PMs shipped around 100K lines of production code. Did they open the IDE? Hell no! Their coding happened through PRDs, tests, docs, and harness rules. The model did the typing. As someone who spent a decade in PM growth roles, I’ve seen how long it takes to move a feature from PRD in a doc to code in prod. For most companies, that latency is weeks. In Ryan’s world, it can be days, and the PM is inside the loop instead of watching from Jira. So I wanted to get to the bottom of this: * What does the harness look like when PMs can ship like that? * How do engineering teams set PMs up so they don’t ship slop? * What changes in the EPD trio when code is cheap, and validation is the bottleneck? That’s today’s episode, and I come with receipts as Ryan goes deep. ---- Check out the conversation on Apple, Spotify, and YouTube. Brought to you by: * Product Faculty - Get $550 off their #1 AI PM Certification with code AAKASH550C7 * Bolt - Ship AI-powered products 10x faster * Customer.io - Send smarter messages using your product data * Ariso - Ship AI agents and features faster, with fewer regressions * Pendo - The #1 software experience management platform ---- * If you want access to my AI tool stack - Dovetail, Arize, Linear, Descript, Reforge Build, Relay.app, Magic Patterns, Speechify, Bolt.new and Mobbin - become an annual subscriber ($150), and grab Aakash’s bundle. * If you want access to my AI PM customizations - PM OS, Job Search OS, and Prompt Library - become a founding subscriber ($250). ---- Key Takeaways:1. Code is a liability, not an asset - Every engineering org was built around the assumption that code is expensive to produce, validate, and deploy. Codex inverts this. Code is now the cheapest part of the stack and the constraint moves to how clearly you describe the problem.2. The new constraint is product decisions per week - With code generation effectively free and parallel, the bottleneck is no longer keystrokes. It is the quality of the brief, the clarity of the architectural boundaries, and the speed of verification.3. A billion tokens a day is the new floor - Ryan's claim is that if you are not running this volume you are negligent. The math comes out to roughly $2K to $3K per engineer per month, which is trivial against the headcount cost of human-only execution.4. A single PR can burn 350 million tokens - One refactor that would have taken Ryan three weeks ran on Codex for 60 hours straight across three days. He gave it two prompts total after the initial spec. The output matched what he would have produced himself.5. The harness is the actual product - Codex CLI is the surface. The harness is everything that gets the agent the right context at the right phase. Pre-work, messy middle, and close. Each phase needs different context, different tools, and different verification.6. agents.md is forcibly injected context - This file lives in the repository root and is always loaded into the agent's context. Use it for the operating model and the non-negotiable rules. Everything else gets pulled in dynamically because context is a hard, scarce resource.7. The painted-door technique works inside the codebase - Ryan's team enforces package boundaries so a designer can paint a fake UI on top of stubbed APIs. Real usage signal, no backend cost. This only works because the architecture refuses to permit a ball of mud.8. The PM's PRD can become a shipped PR in one week - In Ryan's setup, the PM wrote a markdown PRD, the team reviewed it in a Monday meeting, and a working feature shipped to customers by the following week with zero PM-to-engineer back-and-forth.9. The Monday morning roadmap starts with legibility - The first move is making the repository legible to the agent. Write the implicit team decisions down in a documentation tree. Use @-mention Codex to keep that tree updated whenever a Slack thread surfaces a new guardrail.10. One agent beats multi-agent handoffs - The lossy friction of agent-to-agent handoffs costs more than it saves. The right answer is one agent with full addressability over design, backend, and frontend, powered by a model good enough to hold the whole task in context. ---- Where to find Ryan Lapopolo * X * LinkedIn * OpenAI Related content Podcasts: * How to Run Evals in Claude Code with Aparna Dhinakaran * How to Build a Full AI Dev Team in Claude Code with Gabor Mayer * This CPO Uses Claude Code to Run His Entire Work Life with Dave Killeen Newsletters: * PM’s Guide to Claude with Pawel Huryn * How to Become a Builder PM with Mahesh Yadav * How to Build a Team OS in Claude Code with Hannah Stulberg ---- PS. Please subscribe on YouTube and follow on Apple & Spotify. It helps! If you want to advertise, email productgrowthppp at gmail. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.news.aakashg.com/subscribe
- Transcript

Showing 1–20 of 31 episodes