Has Startup Advice Ever Worked? | OpenAI's Strategy Reset | Delve's Compliance Fraud

I think the startup advisory complex has gotten out of control. There's a lot of people who are pushing narratives and frameworks not because they work or they're helpful, but because that's their personal brand. Part of what I tell a lot of founders that we work with is most of this stuff doesn't apply to you and it's simply noise. And you have two choices. If you have a good filter, you can actually take this stuff in, listen to it, and extract the parts that are relevant to you. But if you don't have a good filter, in some cases the safest thing to do is actually to ignore this stuff.

It's the week of March 23rd. Welcome to the Learning Corner, a Precursor Ventures podcast. Welcome to the Learning Corner by Precursor Ventures. I'm Mia Farnam, and I'm joined by our GP and managing partner, Charles Hudson, where we will be talking through the articles and Goodreads that caught our team's attention over the past week. For our first piece this week, an anonymous source published a lengthy exposé on Delve via Substack. It's a YC-backed compliance automation startup that recently raised a $32 million Series A, accusing the company of systematically fabricating audit evidence generating auditor conclusions on behalf of rubber stamp certification mills and leaving hundreds of startup clients unknowingly exposed to serious legal liability under HIPAA and GDPR. The piece walks through how Delve's AI-native platform is in reality a collection of pre-populated forms and fake integrations where clients click through templated board meeting minutes, fake security incidents, and pre-written risk assessments that never actually happened. What makes this especially damning is the leaked spreadsheet. It's a publicly accessible Google Sheet containing hundreds of draft audit reports showing that the auditor's conclusions were machine generated by Delve before any client even filled in their company information. High-profile Delve clients named in the piece included Lovable, Bland, Cluey, Whisperflow, and even a NASDAQ-traded company, with the broader argument being that the startup compliance space may be rife with this kind of manufactured credibility. To be clear, we don't know the full story here. We didn't invest in Delve. We're not inside the company, and we don't have access to information beyond what this published, but it does raise a real question. When a company is just a concept and a team, how do you actually diligence something like this so early on? Delve raised a $32 million Series A, which is a significant amount of capital for a very young team. Both founders were on Forbes 30 Under 30, and you have to wonder if that kind of raise created pressure that pushed them above their skis before the product was ready. I was with friends earlier this week and we all asked the question of when a company has this large of a public blowup, how much of that could you have hedged from the true earliest of stages. Wanted your perspective on it.

I feel like this happens. I was telling somebody this weekend, this seems to happen more often in regulated industries. And I have a theory that what ends up happening is in regulated industries, the regulations are designed to make you go slow. So if you were to call Drata or Vanta or someone else, they're going to give you an answer about how long it's going to take to get SOC 2. That's got to be rooted in compliance. And I think there's always something attractive where a startup shows up and says it's going to take them 3 months. I can do it for you in 2 weeks. Which ought to beg the question, how? How can you do it so much faster than somebody else? And sometimes the answer is we're cutting corners. Sometimes the answer actually is technology. And in some ways, I think this is kind of like the perfect storm. You have all of these companies that are building these AI apps that are growing really quickly. They're talking to enterprises and enterprises are like, you've got to be SOC 2 compliant. And if you're a 3-person startup, you're like, well, what's SOC 2? Oh, it's a box I got to check to be compliant. This company's telling me it's going to cost me a lot of money and it's going to take a long time. And this new startup is telling me they can do it quicker and cheaper. I'm a startup. I'm going to go with the startup. And the implicit trust is that, well, they must be doing the right thing because I don't think most startup customers know how to navigate the SOC 2 compliance process on their own. If they did, it'd be a much harder job to be Vanta or Drawda. And I do think as investors, whenever I see companies that are growing much faster than somebody else in a regulated space, my immediate question is, how are they doing this? And I think in this case, the answer was probably it's AI. Oh, we're using AI and it's saving all this time. And if you want to believe that story, you can be like, well, I have tons of examples of how AI is taking processes that used to be really long. And if you're an enterprising founder, you could tell a story to investors that says, look, we're using AI to dramatically speed up the compliance process. We're not cutting corners. We're just taking all of the human back and forth slow stuff and compressing it. And we're in an environment where right now I don't think investors get paid to ask hard questions. If you ask hard questions, you get— this sounds a lot like we talked about in 2021, where if you ask hard questions and it's a hot round, that founder might decide to work with someone who's going to ask fewer probing questions. And the people that are investors in that company are good investors. I'm not saying that they didn't do their homework. I was talking to one of our teammates yesterday and she was like, well, what if they'd called customers? I'm like, the customers wouldn't have been able to tell you anything either. They would've just said, hey, like they pitched me on this SOC 2 thing. I got a certificate. I gave it to my, to my customer. My customer accepted it. And instead of having to wait 3 months, I got my SOC 2 in 2 weeks. I'm super happy. That would've been probably what you would've heard from Delve customers.

Like without, it feels like the leaked spreadsheet, it'd be harder to kind of connect the dots on, oh, none of us even had the right accreditations. What makes this really dangerous for founders is that if you are a Delve customer, this could blow up your entire legitimacy as a company. I mean, there are really well-funded folks who are on that customer list there, but there are some folks who maybe are much earlier on that were not mentioned that like something like this could upend everything. It does raise the question for founders. Since exactly what you said, we're flooded with a market of new tools and great backers behind them too, which used to be the legitimacy benchmark. What should founders actually be diligent about when they're evaluating vendors right now, especially in places like this, like compliance, that actually have a true touchpoint on the future of the business?

I think my experience has been startup founders apply critical energy and skepticism to things that they view as mission critical. So if it were your cloud service provider, if it were something that you're like, this is a mission critical business function, you would of course analyze it. I think the issue is a lot of startups are like, well, SOC 2 is this checkmark thing I have to get. It's a binary. I either have it or I don't have it. If I don't have it, I can't sell to this customer type. So let me just go get it. And when it's that kind of purchase, oftentimes people will optimize for cost and speed, all things being equal. And I think it's worth reminding people that some things that don't appear to be business critical can turn out to be business risks. And all vendor relationships are important and you should think about them. And I just think back about like Zenefits and I think about Synapse and all of these other other companies that we've seen that have had these like rough patches. And I think it's hard to come back from these things because in regulated businesses, so much of the currency is trust. And so I don't know, I hope that, you know, I don't wish anything ill of the Delve. I'm sure there were employees of the company who had no idea that this was what's going on. And I mean, the other thing is we had, I know people for whom Delve was a significant customer of theirs and a flagship customer and someone they highlighted as someone who used their product or service. And now they're like, great, our like shining star customer turns out to be not who we thought they were, which nothing to do with the service that they were selling that company.

Looking at it from the vantage point of an investor, because this is what we were talking about with my friends yesterday, they walk through in the piece how most of the integrations that were claimed by the Delve team were fake. The funny part is at, at pre-seed, tons of demos I've been on with like, this isn't built up yet, I'll show you the Figma instead of the actual demo. That I can get comfortable with. Just, I need to eventually see that it works, but I can totally believe when it's pre-launch that not everything is set up. The differences when you go from seed to Series A, where that fell through. And this isn't the first time, like we've seen this so many different times.

And we're back in an environment where we're operating at super high speed. Deals are getting done quickly, companies are growing quickly, and I think it creates this dynamic on both sides to rush. And look, I think it's also worth noting 99% of startups are not doing this. So on average, if you're like, I can trust what a startup is telling me is true, and that they're doing what they said they did in the way that— so on average, trusting the startups that you meet, if they're run by good people, is usually a safe strategy. It's just when it isn't, it's very unsafe.

Our next piece is titled "We Have Learned Nothing," written by Jerry Newman, and it makes a pointed and data-backed argument that decades of startup advice, the lean methodology, customer discovery, growth hacking of it all has produced no measurable improvement in whether startups actually succeed. Jerry tracks seed to Series A conversion rates over time and finds that they have steadily declined regardless of how many companies were being funded or how widely startup best practices were being adopted. His argument is that we are living through a Red Queen dynamic. Once every founder is running the same playbook, the playbook stops conferring any real advantage and everyone ends up just running in place to stay average. The deeper and more uncomfortable conclusion is that startups may simply be too human, too variable, and too chaotic to ever reduce to a reliable framework, and then if we cannot run controlled experiments on them, what we call startup wisdom might be closer to pseudoscience than any of us would like to admit. Jerry makes the argument that we have spent countless hours and billions of dollars on startup frameworks, and at best, the numbers show they have had no real effect on whether startups actually succeed. He raises 3 possibilities for why. The theories are simply wrong. They are too obvious to matter. Or once everyone runs the same playbook, the advantage disappears entirely. So the Red Queen dynamic. Which of those three resonate with you? And do you think that it changes how you advise founders?

I think the answer is all of them. And so I think if you zoom out, there's a lot of things that have changed in the last 20 years. There are a lot more people who start startups than there used to be. Part of it is that the cost of doing so has come down. Part of it is that the allure of startups has increased, whether that's the social network or just general, like, media awareness of startups, it's not surprising to me that increased entry and creation of startups has not meaningfully changed because building a successful startup has always been a low probability event. And so it's not surprising to me that adding in more competition and more entrants has not dramatically changed the average because the average startup is simply average. I think the value I see in things like Lean Launchpad and Lean Startup, you know, we didn't have a framework before. I mean, to be fair, like a lot of the Y Combinator advice is build something that people want. Duh. And so I think in some ways, the fact that we have clear language to describe what startups are supposed to do and what winning looks like is not the same as making it easier to do so. And now what I find is the whole language around product-market fit and Lean Launchpad, it gives you kind of a roadmap to figure out, well, where have we gone wrong? Have we built a product that nobody wants? Have we built a product that a small audience of people wants? Have we built a product that people want but won't pay for? And I think before it was just like startups would talk about, well, we launched a product and nobody likes it. And everything started from first principles. And I think it's good to have a common language now. You know, startups will ask me, we think we found product-market fit and we can argue about whether they found it because we're both talking about kind of the same concept. And I do think before we had this vocabulary, there were a lot of startups I met that were like wasting time because they didn't have a roadmap or a plan. And so I think the contribution of these things is not that I never thought that like, oh, product market fit would help you make startups better on average. It would help you highlight when you were on the wrong track, which is unfortunately the case with many startups. The thing I think is totally right, and this is, I think the hardest thing to tell founders is, oh, I wanna know what such and such company did, 'cause we're gonna copy it. And I'm just like, well, these frameworks or these growth hacking, these things are useful because they're novel and they work because people didn't expect them to work. Like the Brex email pretending to be from the CEO where they'd forward and like, give this person a great— like that worked because people had never seen that before. The first time a vendor sent you donuts or some cool swag, that was impressive because like you'd never seen it before. Startups doing billboards that were laughed at in the beginning and then they started to work. People hadn't— so I'm just sort of like the real answer. I tell people, I was like, look, I could introduce you to this company. They could tell you what they did that worked. It probably won't work as well for you as it did for them. The trick for you is to find out your own novel unique. Strategy. And I think the challenge, and I think Jerry kind of hints at this, is some of these frameworks have run the risk of being reductive. And people think, if I just follow these rules, I will be successful. That is not true. Those frameworks to me are a guidepost for a set of questions you should ask yourself as you're going to make sure you're headed in the right direction, which is very different than like, I think there were people who were like, oh, well, Lean Startup should keep you from building a dumb startup. And I'm always like, well, no one I know has ever built something with the intention for no one to use it.

And, and each framework or belief system lives in its own camp too, and they can sometimes conflict with one another.

Yeah.

And so like folks who come from IDO really build intentionally and really empathize with what your customer's thinking at the end of the day. And I think founders within our own portfolio who have come from IDO do build in that direction. Yeah. I think folks who come outta YC, to your point, build and iterate quickly. And then so it's trying to strike the balance of also what noise you wanna absorb and bring into your own startup. What's the negative impact from too many frameworks being shared?

I think the startup advisory complex is has gotten out of control. And there's a lot of people who are pushing narratives and frameworks, not because they work or they're helpful, but because that's their personal brand. And I think part of what I tell a lot of founders that we work with is most of this stuff doesn't apply to you and it's simply noise. And so you have two choices. If you have a good filter, you can actually take this stuff in, listen to it, and extract the parts that are relevant to you. But if you don't have a good filter, in some cases, the safest thing to do is actually to ignore this stuff and to not engage. I was talking to a founder, they're like, I read this article on agency. We gotta find all these people who have agency. I'm like, well, what does that mean to you? And they're like, what do you mean? I'm like, well, you read an article. What did you take away from it? And how are you going to apply it to your startup? 9/9/6. There's a bunch of things. I'm just like, these are ideas. If you're going to adopt them, you should understand what the beliefs are that underlie this and whether they apply to your company at all. And I get it. There's a lot of— it's funny. It's never, I feel like, been harder to convince founders to like follow their instincts and their gut and what they think makes them special because there's so much advice out there that sounds so prescriptive. Do this. Don't do this, this always works, this never works. And I'm like, that's not been my experience with startups.

Yeah. And sometimes that advice is coming directly from your own investors. So I don't blame you as a founder if you feel like you're getting whiplashed. This is a funny question that I just thought of. I guess I know that you help lecture a course over at Stanford and both of us have gone to business school. Should you be cautious of courses that you enroll in that say Startup Framework, Startups 101? What can you actually extract from those classes, do you think?

I know your class runs more guest speaker-esque, but I think the utility of frameworks in startups is the same as the utility of frameworks in strategy class, in HR, in people management. It should give you a way to think about how to solve the problem. It's probably not going to give you the answer. And I think that's the thing that's alluring is a lot of these frameworks purport to give you the answer. I'm like, what they're really giving you is a roadmap to get to the answer. You have to still walk the path and make all the decisions.

And the ecosystem has evolved. So the game has changed and I think it's always nice to kind of explore what new frameworks or the stories that have worked for other companies, but to rinse repeat for yours just without experiencing it yourself, I think is danger zone.

The flip side though, I will say is if you operate without any frameworks, you have to do everything from scratch. And I will say in defense of MBAs, I think in most cases when MBAs that I work with are presented with the problem, the immediate thought is, is there a framework I've seen that could help me reason my way through this problem? Not answer it, but help me reason my way through and structure like what we're actually trying to solve. And I think That is a very useful thing in startups because you can't solve every problem from scratch.

Well, our last piece is from the Wall Street Journal reporting that OpenAI is undergoing a significant strategy shift, pulling back from its do-everything-all-at-once approach and refocusing the company squarely around coding and enterprise users. OpenAI's CEO of applications told staff in an all-hands meeting that top leaders, including Altman, were actively identifying what to deprioritize, warning the company that It cannot miss this moment because we are distracted by side quests. The pressure is coming largely from Anthropic, which has quietly built a dominant position in the enterprise and coding market while OpenAI was spreading its bets across video generation, hardware, a web browser, and social features. Products like Sora launched to fanfare but saw usage flatline, and internal employees described a period of real organizational confusion around strategic direction and computing resource allocation. It is a fascinating moment to watch in real time, a company that has defined the AI era forced to confront whether being first and being everywhere are actually the same thing. Charles, I know you mentioned internally that this reminded you of Google's more wood behind fewer arrows moment back in 2011 when they shut down a wave of products under similar competitive pressure. Do you think that's the right analogy here? And when you see a company make this kind of public pivot, is it a sign of maturity and discipline, or does it raise questions about what got them here in the first place?

I think almost every company that's really successful like this— Facebook went through this cycle too. You go through a phase where people are like, whatever they release is gonna be amazing. They're a total juggernaut. They're unstoppable. Which if you buy into that narrative as the company, you think we can do anything. We can take it. We can do anything. We're gonna continue to launch stuff 'cause we can do anything. We can't miss. And it does remind me of that. I mean, I worked at Google in part of that era and people would do things like, oh, we're Google, we can do this. I'm like, are you sure? They're like, yeah, why not? We haven't failed. We haven't failed that much yet. Let's go. And then you have a few things that don't work and you start to realize like, well, we're mortal too. And then the bigger question becomes, what could we have advanced that was— and I remember like this for me was a big lightbulb at Google. Once people realized the opportunity cost of putting people on ads, they're like, if we just took that engineer who's working on that weird little labs project and put them on ads quality and they made a 1 bps improvement in ad fill rate, that's tons and tons and tons of money. Or if we put this person on the YouTube search algorithm and it gets 1% better, there were just all of these things at scale where you're like, these little things don't matter. I think it's interesting that this is just, I think, you know, you think about Google. Google operated with 20% time and all of this little tinkering for almost a decade before More Wood Behind Fewer Arrows came out. For OpenAI, it was like 4 years. And to be fair to OpenAI, at the time when they were doing all of this stuff, it felt like they were so far ahead. Anthropic, we weren't talking about them in this way 18 months ago. We weren't talking about Gemini in this way. We weren't talking about Deepseek and KIMI and these open source models in the same way. It really did feel like they're so far ahead. Said, why not? Why not try all of this stuff? Because there was nobody in the rearview. And now that that market that their core market they're in has become much more competitive, I think it makes a lot of sense to say, hey, these things are, while they're interesting, they're not core. I'll be very curious to see what happens with the hardware stuff.

I know. I mean, when that news release came out with Johnny Ive, I was like, whatever they create is going to be in every single household in the next 10 months. And now I'm like, where is that? That was such a great little video they released when they were walking down literally our streets outside our office. And now I'm like, where is the product?

There's a lot of activity. Activity there, they're a couple doors down from us. I see them like working hard. I think at some point, for me, the big question is, you know, OpenAI has an unbelievable global consumer footprint and brand with ChatGPT. And the question is, what's the best use of resources? Is it to lean into that and be the basically front door to AI for consumers? Or is it to try to be the front door for consumers and for businesses? And I think the business use case is so big, it's hard to like cede that to OpenAI and Gemini and just say we're not going to compete there. But I think these are— someone will write a Harvard Business School case study on what happens at OpenAI in the next 2 years around this strategic decision-making framework. And I'll be very curious to see what they decide is worth spending time on.

And again, to what you said earlier, kind of inflection point of deciding when to focus versus kind of tackling surface area. It happens in every single company. It just, the magnitude of capital that OpenAI has and the time that they've been able to kind of exist in more of the work on everything concept is way longer than the average startup that has started at the same time with far less capital. So the fact that they're doing this isn't surprising, nor should I think is a ding on them. It more is a call of like, this is still anyone's game. Everyone should be innovating in the direction that they think they can win in.

And startup news operates in narrative cycles, and for a while OpenAI was the narrative winner and Anthropic was the narrative also-ran cute second company. And you know, tech news loves the narrative shift, then there was a big narrative shift, and now Anthropic's on top. Claude Code and all these things that are shipping. There's nothing to me that means that the narrative can't shift back, and in a year people will say, like, you know, we'll see a picture of Sam Altman in a suit, the return of the focused version of OpenAI, how a year of focus helped them get their mojo back, and now they're back in it again, better than ever. And these things move in narrative cycles, and there will be another narrative cycle for OpenAI.

We are all out of time today. Thank you, Charles, for popping in. Thank you to our listeners and subscribers. Please like and share with friends, and we will talk with all next week. Thank you. Bye.