Even Anthropic Still Needs PMs: Lenny’s Summit 2026 Review | AA274

So Om, watch every talk and every panel from the Lenin Friends 2026 summit so that you don't have to. awesome. there's 11 videos and they're varying ranges. Like some them are short, like eleven, 13 something like that. And the longer ones are think nearly half hour. but basically saved you work day. tuning in the Lenny and Friends Summit is the yearly gathering of Silicon and you're just Valley product darlings the CPO of Anthropic, the CPOs of Anthropic. think there's two. There's two where they're Ramp Atlassian CEO linear who will get into here second. And the product leads from Open AI. So there was a who's who. Oh, and not forget Marty Cagan himself. that's right. Yeah, that's right. Why I really wanted have this podcast was I wanted into all of the repackage wisdom that you're living the agile world, you're you're very familiar with. Yeah, but it's being discovered here again for first time. Deja Vu all again. That's right. OK, that's right to kick this off, let me present you with exhibit a last year I came for PM's. That didn't work. Let's do again. This year I'm coming for the roadmap. Do you still need PRD? Do we still need roadmaps [UH] never heard that one posed till this morning [UH] Despite Claire's optimism, I guarantee your roadmaps are not going away. Welcome back arguing agile. your first time. Welcome I'm product manager Brian Orlando and this my co -host enterprise agility leader and the curator of Forgotten Principles, Mr. Om Patel. Today, that's what are doing today. That's right. That's right. Boy, there's little bit agile amnesia in this podcast. But the, but the things people out this podcast. The PMs who ration engineering that they think that's their job they're ones being replaced by AI. There's strong undercurrent people these talks on one side or the other, the project model side or the product model side. And I thought about having cool little graphic with their faces or whatever and move them around we talk about fight or like the audio. It would lost on anyone just listens to this. I'm like that's much effort. some talk about feature factory. So again, project product model. Right? Give me lot And then there's be talk about feature factory. and then last thing you're hear is that strategy. and never really lived roadmap. And lot these product folks, know, if knew to read, they'd be very upset. But lot these product folks are just discovering that for the first time now that ability to throw out new products and new code, like now that has made that possibility. They're starting learn that oh! maybe need strategy. there's definitely line items their budgets for AI. but there's definitely nothing in budget for rework. So never was. Yeah. And there isn't even today, which is surprising. Right. Here are. will land on on one side of that of that like every speaker landscape. Either the east west, the north south, the something or the left or right. Yeah, So let's start with what I alluded to in the opening is that there are two kinds PMs at this both presenting and listening. Probably right is the people that believe that engineering time was always scarce resource. And now that they're ones doing engineering with AI. .I. that like their job obsolete there product builders now and they're, that's their job whatever[UH] And then there's people that are kind claiming the opposite. It's like, that wasn't really, maybe wasn't little bit issue, but that wasn't really the main issue. round one, we're watch some clips and really the thesis in this explores our product managers necessary subjects in the world. right. Let me put[UH] Katie Dale. She's head design at Stripe. I thought she had one the best talks in the whole series [UM] or play little bit [UH] from her. We're listen to a bit more than we'll end up in the video version of this just because don't want to get struck from the stratosphere by Lenny and Co There was a point view behind Villa Savoye. There were principles behind the Bauhaus Dessau. But after World War Two, with urgency to build and new construction methods, the style was essentially copied again and again without intentionality. The thinking got thinner and thinner and all that was left were generic patterns ill suited the context at hand. Katie Dill is a designer by way. At Stripe. I'm sure you are no stranger to what she just said. You're completely unfamiliar with you said. All like, oh, some people did some things maybe made us wrote some principles down and values down and they really believed in them. And other people saw those things and started copying them. didn't really, they didn't really care about principles. didn't really care about whatever. They just wanted go fast and copy them verbatim and chose what they wanted to out of it. Yeah, right. This is Deja Vu you're talking about. Absolutely. I mean like it is the nice thing is I think I watch her as first or second. can't remember [UH] don't think were any other designers that had talks. sure everyone else had talk was either like CEO or like a CPO or a, I think her being designer was one the reasons that it it just came from a completely different angle. Came from it came from completely different side to say hey you can you can do things but are they good? Like do people really care about them? they inspire anyone? know or just punching out like flat box style buildings because everyone else is. Right. Yeah. People often tend to kind just go along right with flow. And she's like said brings different perspective. Speaking going along flow, let's see what Claire Vo has to this. engineering was capacity the scarce was resource. Here the scarce we resource go. that was like I mean that's kind of why we exist honestly is like we had way more ideas than people. We had way more demand than our ability to deliver against that demand. And the product manager, the products leader's job was to prioritize all those ideas, sequence them, and then spend like all time in meetings and in Slack and in spreadsheets saying no my bottleneck has moved from like building what what can build to believing like what do I believe is actually worth building? pause there because I to give backdrop that I went into this. We did whole episode about Claire Vo and her talk from last year where she said that product managers are dead. Everyone can build. Right. And listening this talk, it seems like someone who is has only ever experienced the project model of building products is sort coming into this realization now of [UH] oh now the better ideas the better assumptions the people can test assumptions opposed just building stuff are performing better. And she's seeing this world for first time. And that's what like what flavored lot this for me when I'm watching this. It seems like someone slowly coming that realization, which think, which I will tell you flatly. did not expect that. I expected a mostly tone deaf Claire Vo like saw last year Much improved. you know, this year and this talk [UH] And was, was quite impressed. wrote down I wrote notes give her kudos for like real realizing this and then bringing forward say hey like this thing that was like brought up to understand like I don't think this important anymore. Right. Yeah, think it's good admission of know what reality is today I think it's like said kudos her doing it and doing it on stage front people. Right Fantastic. Right. I think this gap between our building capacity and like true conviction like does any this code matter is a much bigger problem than any of us are willing to admit. So here's another one with my career who is from anthropic. Now he was the apparently have two CEOs Anthropic. He was co-founder at Instagram. believe was Instagram. for me, had a realization moment recently. so earlier in the shifted to an IC role. mostly building and we're getting close to shipping something. And one the like PM leads works with Ami pulled aside like, Mike, I know love build this project going well. Like it's looking really exciting. Like we really need PM on this project. like, don't know. Like need PM? Like Claude's got it like we've got lot going on. she's no, we really need PM. And it's like such reminder. Like once joined, like all of the things and all the glue and all of the connective tissue and all things were about get dropped if I'd not done that. That would have happened. like just such cool eye opening like before and after so much week later I think and I was like Kat, 100 are percent right And thank you. And just, and think about all the things that encompassed. was, know, actually bringing along. We're a company that serves everybody from, youknow prosumers all the way to really large enterprises. Like have you brought all of that contingency along. Right. When we need figure out how enable customers success team to like talk about the changes as questions go time. somebody thought about that? Is somebody looked safeguards? It's obviously very, important. Are people getting kept track? All these things And nobody has infinite capacity even with this assistance these tools. So able have like call it what you will, but somebody playing that role of making sure that everything go everything get connected. We've kept the end user needs mind, To Ami's point, and that the builders are like heads down are you know, in full on Claude mode can keep doing that without having them pop out and the very different activity, which is how you actually make sure all those things well. think it's very essential. think all that's happened now that because can move faster, that role is like have be operationally excellent that role way that like even beyond what had be before. But it definitely not gone away. fact, I have felt this like renewed like yes, this is a like you're recruiting for team need people wear these hats, this hat is increasingly important So do just hear him say he's come from the left side spectrum the right? Yes. Right. Yeah. This is the second one that's done that. Right It first one. The second one. Interesting. Right. Yeah. And he, mean, he, he's here talking about that. know, he, he built Instagram, was builder role. know, built built. Everything was great. Yeah. And then you know, he brought that same attitude to Anthropic, started doing same thing. then one other product leads Pinged him is hey buddy, I think really need someone help you[UH] do all these other things. And know, was like, well, I need that? I just let Claude let a little, little bot do the scrum master job. Yeah, I can do that. Right. And he, if you watch whole talk, he, he does get into either he or the open eye panel. can't remember these panels gets into Claude can't like build human relationships. And Claude the [UH] Claude only has context give it. Yeah. And doesn't have, it doesn't know how certain people get along. You're making simple statement when you're saying just let chat bot do it. Absolutely. And the context, you know, may never be there. Maybe somebody technology occasionally, know, the technology will eventually get there whatever OK, maybe it'll get close that. But you're talking about like every human interpersonal relationship at company between every other person I[UH] mean, you're putting lot tech right there. Yes, it's the interesting thing about his [UH] statements there. He did say you know yes maybe somebody could play role. So he's coming around but he's also like we'll play the role. It's kind of downplaying little bit. thought so it's not full on admission that yes, this, this is needed. It's like, know fine. have somebody there to play the role. But keep building. But again, he, he came around to it, know, now he accepts Oh, know but it's funny that like somebody is full time, by way, a chief product officer, by way, is like a product lead came him and said, hey, think should look this. It's funny because it's funny but also kind makes cry because like again this This is the product career ladder. The people at the top don't even really think need a product you know it's like another episode this, but I can't. need be little less[UH] incendiary than I feel right now. Even listening this, you [UH] know get through... oh, Ami is is the other CPO. me, think she has good response here. me play think we're also living world where like the tools are really good and you can build a lot things and can build them really fast. In some ways, that just like expands the universe of what is possible to build. And you still need to have really good judgment about like what to build based on sometimes pretty limited information because feels like everything's moving really fast. So like figure out what to build. And then think in some ways have be even more relentless about making sure happens because there's so many forking paths that you could take like you just heard me is the other CPO at anthropic. Got great point. It's like, well, if you hand over chat bot, there'd be an unlimited amount forks [UH] it could take. again, you're hand all that the chat. It's just a, it's a ridiculously, it's like reductive argument that don't have when people bring up now because it's ridiculous. It's interesting. They have to [UH] co- CPOs. It seems like they're kind of balancing one another little bit that. Maybe in that you know, discussion that maybe. Yeah, absolutely. mean, do you really to push all your chips across table with[UH] the chat box. Probably not. Probably not. Yeah. right. Let's [UH] let's let Marty Cagan button up this category. if you spend your time doing some pull requests the front end. Fine, maybe should just go be a developer. But if you want to make sure that what your engineers do build and provide your customers, you should be building learn. It's a totally different approach, totally different job. Your job is make sure that when they do build that thing, achieve the outcome. need the principles of the product model, the craft of product strategy and product discovery have never been more important. that was end of that section. The section of do we really need product managers? mean, Marty Cagan just told you like it's more important ever. And heard couple people come into the realization. But like nobody really here was pushing hard against. So not a lot arguing going on section one. Right? Maybe section two we'll get some arguing going on. so which is at your company does the PM ration engineering out still? Because again, that that model is what I feel is going away. And they've all highlighted here. Yeah And that's a business you should get out of anyway because [UM] Honestly, doesn't even really work. That is business that's being automated by A .I. So you know, smart get out of its way. and if you are business of rationing out engineering and you think need software factory to help you RAMP has factory that like to show you. And I will tell you, this was the the ramp CPO says speed is everything. And he did in this talk. This was [UH] one longer talks believe. it was also pathological [UH] in parts because again, he, he really believes that like speed everything and[UH] everything can be automated. And his presentation, there were some good intriguing highlights of how they automate things And I'm play you a long section his talk. It's going get cut down for the final edited product that we put out. But[UH] I'm player for you and we're talk about so we're hear from the CPO RAMP. Geoff Charles. We're gonna hear from CPO RAMP, Jeff Charles. And then we're going to hear from, I believe he's CEO Linear immediately after that because pretty much had opposing talks. Interesting. Yeah, yeah, which is only reason we're even listening this talk. And otherwise completely skipped it because it makes the other one more interesting. Yeah. 90 percent of the parts car change every year. Only 10 percent of 16000 the parts of F1 vehicle carry over next year I want also think about what if 90 percent your code changed every year So I know, know, I know said that we would let him talk but a, I need a time out right now. Like I'm taking my lap and time out. I'm play. This flag on plate because I [UM] know a little bit about racing I know, know two things. I know rubbin' is racing. That's what know. I learned that from Tom Cruise. and I know that the parts of Formula One cars are constantly being tinkered with and redesigned for like slightest edge no matter how minor every year. So every piece car is in a constant state of improvement. And there's like like entire teams work on one piece car. And the only parts car that don't change are the parts either F1 has regulated that you must use, know, X, Y, Z component for safety whatever Or they're just so expensive like the steering column, like they're so expensive you couldn't engineer that piece car. You know, those regulated. yeah, right. Yeah. Probably to some point. Yeah. So like saying 90 percent of the F1 car changes every year because they're so fast or know, they're so fast. They're trying like build better components. mean it's sort true but also it's like not really true. Like you still need wheels. I understand what conceptually he's trying say this, which is you don't care what code is in your software and the A .I. will replace 90 percent code and don't care because the program still works. So like get mindset of caring about the actual code itself. And I don't like again, don't know go along with that. I don't know. I'm having hard time getting grips with that too because the people are trying get edge on the other, know, on the other [UH] racing team. But like said, within the guidelines and regulations of what's regulated, you can't just what you want. Right? You're right. Right the opening his talk is speed everything. Execution is everything. That's the opening. now I'm cutting you to he's showing how they business at RAMP. He's showing some internal applications. He's showing some their internal workflows. So this what I've cut to. I've got some their internal workflows stuff. Sure. 93 percent peers are now automatically handled by review buddy so that the highest quality engineers are spending time the seven percent of peers that truly matter, numbers like that me I course ignore them 93 because percent of peers are automated. OK, like 15 is that peers? Like what that? when someone throws know? Yeah, it's also what what is there in the PR itself, Yes, minor things or yeah [UH] don't have to look at them. Right? [UH] Ninety nine percent of all CSS changes. Right? Yeah, CSS color changes. That's right. That's right. It's probably not that. mean, we're probably being little flippant, but[UM] only little. next bottleneck is testing. Previously PMs. I'm sure you're familiar with this. We spin up QA environment. We try tweak different fixtures. We click through the screens. make that engineer didn't skip out that one feature. really liked. But by way, that's We how he does testing. Like he spins QA environment. just kind of pokes around, looks for his pet thing I will say like this. This does sound lot like how product managers do testing, which is why they probably shouldn't doing tests the first place. Exactly. Yeah. let professionals do anyway. They built their own test suite or they use test suite. I'm not quite sure they, think built their own, but he talked about here. automated that. We give it instructions. say, hey, pay an invoice but amortize it. And clicks around and actually gives us feedback both blocking feed sure that everyone company is loop. Things moving so fast. You're shipping so many different things. The next bottleneck becomes human attention. We've all been there. PM's being inundated with notifications. And so job becomes coronation. I see few people coronating right now. Too much process. Right. And slows down the builders. I Too little process. everyone confused. stop right now because is debate tactic he's doing now. And it's a, it's a disingenuous this debate tactic. It serves he's saying. we're PMs our main job is coordinate but like too much process. And like that's that causes problems but not enough the process that causes problems. Like OK, like he's not taking side. like water's wet. Like what's your point? Exactly. How you solve this? every question is API. Every question API. And you architect system. We build gadgets. It understands full picture. It understands is next steps and answers with evidence. But not only that, the agent updates our roadmap. The [UH] agent gives status updates. The agent pings people who So are late[UH] their I deliverables. point out what's happening. There's here. a lot Let of me fun pause ways can actually this. scale. went from testing to say as we move through rest process, we build gadgets to automate things. And then we let the agents paying people assign tasks, let them know when they're behind. It writes code. It sends stuff out. So basically they built a push agents. Yeah. That push things to the next person next decision cycle, the next whatever that's I'm hearing with this. that's I'm seeing with this. And it's, it's strong, like strong. He strongly, he's trying convince me. We just, just spent the last section exploring all these people the project model, having these realizations that they need move to the product side house. He's going other way he's going the other way, trying sell them on Well, isn't this great? How fast can go in the project model? As say, speed is everything. So yeah, the push model versus the, know, the pull model. Yeah[UH] trying pull us into the into the feature factory. So let's[UH] let's let's finish. So this page, you're probably asking yourselves three questions. The first question is, OK, how know we're moving faster? Right? move faster with quality? After all, you know, talk about product velocity. measure it? Certainly harder than, know, a lap time racetrack. some things are not easily quantifiable. Nicky Lauda in 1974 took lap with a Ferrari car and told yeah. So he's he's Enzo this piece It drives poorly and handles poorly, breaks poorly. Got car, went engineer, fixed it. That not dissimilar to the culture RAMP. I'll start there. But the point I'm trying make here is that you, you will know you're moving fast enough, if you actually hire driver that knows what speed looks like and you put them in charge and you actually allow them to challenge you. That might be hard for some leaders, but that's what need do to be challenged by people who know what great speed looks like. there because again, he threw out bunch stats. 60 percent of UX issues or whatever get solved 24 in hours whatever. And And then said well how you know the speed is helping? going [UH] stop says, you can't really measure that you need hire driver knows what what speed looks like. And I like at this point the talk, I'm sitting there and I'm thinking how like am I the have I missed a critical detail here where like I feel that line could been cut from this talk. Like if if he had previewed this talk to somebody out loud, that been like[UM] you should cut that line because kind like kind of undermines whole. exactly. Yeah, listen, these people that up there the on the podium, right on stage. guess talking about their stuff. So yes, guess, I guess the factory go fast, go fast. Well, what don't hear at least thus far, we haven't heard is what is the cost of rework. Oh, he's not he doesn't go. course he doesn't there. Yeah why is that not surprising? think I might have left that comment on the on this video. I'm not sure but I'm sure didn't get responded to. right. Let's let him finish. I think he might be done. I might left more here. Let see else In Audi 2006, Audi to bad car very slow car so they asked themselves how can we win if we can't go faster than anyone else? Their answer, fuel efficiency, fewer stops, more time track. And they won Le Mans three years row Cconstraints force you to choose dimension in which can be world class. So embrace your constraints but not bottlenecks. Find the one thing you're really good at. Find that one bottleneck that you think can 10 extra company and just start there. Yeah, yeah, stop here because know got strong feelings what just said Constraints bottlenecks. They can 10x your company. You're good luck with that. First all, it's like someone who never read goal had like and just absorbed it through pop culture That's exactly right. now repeating those concepts. Yeah. Constraints are actually restrictors to your progress. They're not going get to 10 X. Right. So that's quite opposite to what he's saying. In reality, constraints are those things that force you to think through them and overcome those. Sure. But still, you're not the same speed that you would be [UH] if you didn't have constraint the first place. Constraints are not enablers or accelerants. Right? They don't accelerate you. So do get 10 X by embracing your constraints? it's don't know. again, was my big take away from listening couple talks is like, well, they're rediscovering principles that like many people outside Silicon Valley like know and live with, And one them is this, know, constraints and bottlenecks, know, theory constraints every constraint, every bottleneck constraint. But not every constraint is bottleneck kind of thing that I posted out Yeah, whatever LinkedIn where was. Right. And was from this talk directly And it's funny that he just gave story that should be most

Audi was losing races, And they said, are there obviously like we can't like we don't have money to, know, buy better engine buy better this whatever. Right. don't have money to compete with the big players. So are there bottlenecks? This same like it wasn't out in a race. could be like I'm running a business. I'm running like factory building widgets. And I know that don't have an unlimited amount money like Coca-Cola or Pepsi these otherlike, know companies has. But is there some bottleneck that I can concentrate on and solve that my competitors have not solved? That would give me a competitive advantage where [UM] my cost becomes lower in these specific areas to compete. And so, let me solve those first different strategy. me have different strategy than the big boys have. And me solve let me solve bottlenecks that they wouldn't solve just because they're either, know, not well -run companies a million reasons why they, why they haven't solved this bottlenecks they don't have good management. Who knows? Who And once solve them, the business frees up and new lanes competition open up that weren't there for me before that can compete in. it's like someone who read the like they went store, they looked the back of The Goal and read the summary on the back cover and then gave talk. Yeah. From their perspective, And that was part of me being enraged with this talk is sorry, I'm trying make good point, but don't really have good point. The example goes wonky little bit because he's emphasizing speed whole time [UH] he wasn't winning. had crap car. They would have focused just on speed. How did they get, know, they get leg up competition by building a faster car. Basically they didn't do that. They were thinking outside box about keeping car on the track longer. How that? field efficiency one. But there are other ways as well. Yeah coordination of the team itself that 200 people coordinating this race not just driver. Didn't when the Mini came out, didn't they like destroy every race they were in? Like that because car was so light and cornered so like it's like go kart, right? Yeah, But think I'm [UM] pretty sure know I remember the story was they got banned. But I do remember they won like every race were in. And it caused a big uproar in a racing community. This was, know, many 50s, 60s, 60s, think. Yeah But kind same thing is like, well, I mean the faster engine, the handling of all these cars you know, and the size and that stuff, like the, know, how can around corners. The, the car just competed completely different terms. Yeah. So do call constraint? I don't know [UH] I think it's innovation to be honest. I also don't think it's just, know, that unidirectional go after speed thing, right? Because you'd miss out on all these other things. Yeah, had biggest problem this talk will just flat out tell let's, let's, think he's done. let me move on to Kari Sarin sir. Sorry, mean surname. Sorry. And I believe he's the CEO of Linear. I believe context now the product product leadership. When software can build itself was naming his talk which you're to, you're feel a serious gear change It is important to, to move fast and like do things. But this started even before And think before the idea was that need hire more people. We create very specialized roles so they can work in their own ways. And then create because there's lot people, we create more process. And because have process, lot people, we actually don't even know anymore like what people are doing or what they're supposed do. So run experiments. So experiments, data tell us what, what works what doesn't. And this all to scale like how we get more output from the organization But again, like my philosophy is always like it's more more output isn't better. I think like what we want to build is like better experiences customers. One my messages today is that [UH] you can build the software factories and you can make and you can automate things and that all makes sense. But just like as organization, don't become one. Don't become software factory where you are outsource all the work and thinking to somewhere else and like just focus on like how efficient can the output be because output is not product. Like customers are not buying lines code. They're not buying experiments or in end, like if sell something or make some good product, have make something someone wants. making products produces two things that the product and learning. struggling like trying think what you should build, how you should build it, which way, which shape does it take? Like when are You have all these questions you are asking and maybe you're answering them yourself. Maybe go talk to someone, maybe talk customers and try to find the answer. what he just said night day, night day, from what we just heard completely. I'm his camp. I'm in Karri's camp. Yeah The learning piece is interesting. That point that makes because if you're automating stuff and you're fixing things, your agents are fixing things. That knowledge of what went wrong and what was the restorative action, things like that you're not learning anything. how you then leverage that in the future because it's not even there to begin with So that's gap in that gap amplifies over time. Yeah So I'm in this camp. You know, I've always believed this to be true. You can automate stuff. But what you missing? Right, right. And he makes good point there. So like that, so section the section we just listened to of like should build feature factory or should we build a it's called software factory. Right? I feel were so close brand and they could just called feature factory. should have In Geoff's slides everything was output, was a PR written, PR reviewed, bugs caught, questions answered, stuff. And then earlier podcast, talked about Claire Vo coming the realization that the feature factory like she wasn't sure if any of it mattered listened to Carrie[UH] And then we who said know, hey, the heck this. Do whatever your customers like really respond to and care about. And like speed great but it isn't the product. Right. And world like in the in the normal world of the companies we talk about podcast where they're, know, they're heavily influenced by accountants and finance folks boy, I feel that's a night day I wonder where they were in the talks I hope they weren't next each other. It really been gear change if If Karri was first and then Jeff came after him, like travesty Yeah, absolutely. If they were next each other, may have been in fisticuffs. And course, thing that already talked about was left out it was does company track rework? Because I need know because also with the, the AI factory being like the feature factory being like so explicitly created. There's not just rework on the work that has go back into the [UH] into factory get fixed. There's also rework on the process of the factory itself and all that stuff. Like do they track that because that's not in anybody's budget. Let me assure you boy. Yeah, we got time that. Right. understand. get back work. If there's away section, get a find your AI throughput numbers, whatever, whatever your throughput numbers are you're measuring. And let us know if it's rework rate either work itself or in process that takes maintain that AI factory doesn't know you're tracking that and how you're doing. Yeah. Most interesting. All right. break this into two parts to be honest. Like we can because next be built build it throw away. call Like pause here and lab. that the the R&D guy here. And we're only halfway through. We're exactly halfway through. think you're right. This video should two parter. I think we're going turn this into a two parter basically through half of what had ready to present. And the next episode We'll talk about the build throwaway mentality and we'll talk about who who actually is listening customers and what Marty Cagan has to say about his biggest learnings over years because that that kind of that will basically entire section all own in next video. That's sequel. miss folks next in next time [UH] as always if you like this video like and subscribe. That's right.