A version of this continued in the 90s with the rise of the internet. Marc Andreessen observed the trend in his famous 2011 “Why Software is Eating the World” essay. Peter Thiel made his famously dour proclamation: “we wanted flying cars, instead we got 140 characters.” Software enabled new technological products, though notional in their physical presence, to nonetheless be built without much working capital.

During this period, venture had matured from the days of Arthur Rock and become a respected niche in the world of private equity. Despite the influx of new investors, returns nonetheless stayed strong. As Andreessen’s own investing style matured through this period, he focused on understanding software companies as opposed to “constantly questioning their valuations.” It worked, for reasons to be discussed. He and many others made a killing.

In recent years, his eponymous venture firm and the valley as a whole have directed considerable capital towards a new niche, hard tech. With it, they’ve brought three legacy practices from software investing: relative valuation insensitivity, a ten year exit window, and a willingness to double-down.1

These practices work in software because the marginal cost is zero, iteration is fast, and distribution scales exponentially. The same cannot be said for hard tech. We can observe this in outcomes; outside of biopharmaceuticals, hard tech startups founded in the last quarter century have largely floundered. A few eVTOL and self-driving firms (Archer, Joby, Cruise, and Zoox for instance) have IPO’d or been acquired; none have ever been profitable and some haven’t even generated revenue to this day. In fact, if we exclude companies founded by Elon Musk (SpaceX and Tesla) and the biopharmaceutical industry, we arrive at a simple yet terrifying fact: beyond them, there has been one profitable hard-tech company in founded in the US in the last 20 years.2 That company is Pure Storage.3

Sure, to a certain extent, profitability doesn’t matter to venture investors.4 Archer’s IPO could be considered a successful exit. Yet the appetite for unprofitable flashy businesses is bound to wane without a certain expectation of eventual free cash flow. I am not bearish on hard tech per se; I am however extremely bearish on venture returns from investing in hard tech.

As such, this is not to say that Helion, Varda, Skydio, and a multitude of others won’t succeed. But in a business dependent on power law dynamics, it’s difficult to understand how Series A and beyond investors intend to return their fund from 1-2 hitters in the absence of a fundamental change in market dynamics. Barring government intervention, I don’t see that coming.

Regardless of this, I’ve written several angel checks into hard-tech businesses in the last 18 months. I’ve done this for three reasons:5

1. I came in at or near inception

2. My time horizon is longer than a typical fund window

3. The companies used proven technology for new applications

These points seek to address the aforementioned software investing playbook that venture has brought to hard tech. By coming in at a lower valuation on a longer time horizon with a simpler product, upside materializes quicker and more acutely.

You can just do things, even in hard tech. But be wary of an old playbook being applied to a different landscape.

The heuristic applies to all early-stage technology companies (barring, perhaps, biotech and a few notable exceptions). Their primary advancement is seldom discovery of a new technology per se. Rather, they combine existing technologies and repurpose them in a profound way. Garg discusses how companies are often limited by this factor, and it impinges on or enables their success. We can consider this “the spectrum of accessible ideas” in any given moment.

The extent to which an idea relies on a discovery (or challenges the absence of mass adoption of the underlying technology, e.g. location based apps before phone GPS) is indicative of its accessibility.1 To put it in concrete terms: consider Kiko versus SpaceX circa 2005. The former—which acknowledged in its YC application that “Google might crush us like ants” (it later did)—was building an online calendar application.2 The latter was attempting to build low-cost rockets. One idea was considerably more accessible than the other.

The spectrum of accessible ideas is, primo, dependent on—as Howard Stark once put it—the technology of the time. Secondo, and to a far diminished extent, it hinges on what is legal. As such, build something extralegal.

Illegality is a spectrum, and doing something outright illegal (à la Ross Ulbricht) is not the goal (nor the meaning of this heuristic). Yet a better understanding of the range of it clarifies an otherwise blurred line. By that token we can conclude that building on the edge is an edge. Expanding your spectrum of accessible ideas naturally reduces your competition.

To actualize this, consider three categories:

1. Provocative

   1. Clearview AI scraped billions of images for facial recognition matching. Though not outright illegal per se, its product is unpopular and has resulted in numerous legal challenges.

   2. Juul launched flavored vapes in sleek packaging that proved immensely popular with teens, later prompting considerable backlash.

2. Extralegal

   1. Uber launched unlicensed taxi services.

   2. Airbnb launched a short term rental marketplace before cities explicitly allowed them.

   3. PayPal launched peer-to-peer payments before the rules surrounding internet money were written and later obtained banking licenses.

3. Unlawful

   1. Silk Road and AlphaBay sold drugs and arms on the dark web.

   2. Napster allowed for music pirating.

   3. Pirate Bay allowed users to access pirated films.

One category here clearly outperforms the others: extralegal. By building in a frontier space where regulations are still unclear, the founders gained an edge. VCs are willing to accompany founders in such efforts. To a certain extent, this also appears at large founder-operated companies willing to take considerable risk as well; Meta, like Open AI (allegedly), trained Llama on LibGen.

Peter Thiel has noted that he had a very brief window in which to build PayPal, one in which online peer-to-peer payments regulation had not caught up. To that end, part of this is timing, as Garg well notes. Another is being willing to just do things. In other words, take reasonable risk to expand the spectrum of accessible ideas.

It's pretty difficult to break the efficient frontier given physical constraints. In other words, you can only give expand these clusters so much. Even if Project Stargate commits billions to it, without proper government investment (in the form of immense formal allocation in a spending bill), American companies are constrained by cluster size (which is in turn limited by GPU production capacity and energy). Hence why Altman et al have made immense investments in the latter (à la Helion).

So the major labs have turned to test-time compute scaling for o1 and other models. They only have so much compute available to them, at least for the next five years.2 All of this takes time.

To that end, time represents the critical constraint in another respect. OpenAI “personality hire” Aidan Mclaughlin rightly notes (quoting the METR report) that while the model can complete longer tasks at an exponential rate and "think.” They're nonetheless “pretty evidently slowing” in my opinion vis-à-vis actual performance with respect to time. Inference scaling has led to exponential improvement in the aggregate, in the short run.

My favorite quote on this is from OpenAI researcher Noam Brown, who said “it turned out that having a bot think for just 20 seconds in a hand of poker got the same boosting performance as scaling up the model by 100,000x and training it for 100,000 times longer.” Because inference scaling was relatively new to the public at the time, it was treated as a pretty novel breakthrough. It no doubt is.

But his quote also affirms that we're beginning to reach the physical limits of compute, such that the major labs have turned to these ‘alternative methods’ as opposed to just expanding their clusters. To that end, you can only make the models think for so long. Time is a zero-sum resource just like energy and compute. Meaning we are likely hitting a certain wall for now, even if it’s not the ‘formal’ scaling wall.

As Ilya Sutskever said, “the 2010s were the age of scaling, now we're back in the age of wonder and discovery once again. Everyone is looking for the next thing.” So there are basically two things that will happen from here:

1. We will find the “next thing” (akin to inference scaling) in the next year and model performance will continue to improve exponentially.

2. This research will take a little longer, and will coincide with cluster expansion in the next 5-10 years. So we’ll see wild performance improvement, but not in the short term.

I’m inclined to take the latter view. So does Dwarkesh Patel, who recently wrote:

While this makes me bearish on transformative Al in the next few years, it makes me especially bullish on Al over the next decades. When we do solve continuous learning, we'll see a huge discontinuity in the value of the models. Even if there isn't a software only singularity (with models rapidly building smarter and smarter successor systems), we might still see something that looks like a broadly deployed intelligence explosion. Als will be getting broadly deployed through the economy, doing different jobs and learning while doing them in the way humans can. But unlike humans, these models can amalgamate their learnings across all their copies. So one Al is basically learning how to do every single job in the world. An Al that is capable of online learning might functionally become a superintelligence quite rapidly without any further algorithmic progress.

Perhaps Zuckerburg’s Superintelligence Labs or Ilya’s SSI will crack this sooner. For now however, AI superintelligence progress is constrained by what limits us all: time.

Many people following Cluely on the timeline have a limited idea of what they do. John Coogan and Jordi Hays on TBPN, major Cluely boosters who have interviewed Roy three times now (including today for his Series A announcement) declared themselves “above” using the app. Instead, they consigned their intern Tyler to use it. Live on the air, viewers watched him test the product ahead of Roy’s appearance on the show.

Once he had it set up, his answers—when posed with questions like “how much has the Chinese government invested in Shenzhen”—were punctuated with pauses and “let me think about that for a moment.” In other words, there’s considerable latency in the product, which anyone whose used it knows about. While answers pulled up ambiently is valuable (as Coogan noted), for most, it’s not very useful at this point.

Now there’s no doubt that Roy’s ability to capture attention in the form of TechCrunch articles and The Social Network inspired launch videos is unparalleled. By the end of the day, in the eyes of most viewers, I fully expect this announcement to overshadow Murati’s Thinking Machines Lab’s gargantuan $2B seed raise. That is real talent. It inevitably spurs the question: how does attention become a company?

That “attention” might be better termed fame. X user “@annaarthoe”, in a post highlighted by TBPN, posited the following this morning:

“the average 21 year old silicon valley founder is no longer driven by greed (ellison) or idealism (jobs), they're driven by fame. its too gauche to become an influencer, so being a startup founder is the best proxy to status”

It’s a great (and tragic) take, relevant to hundreds of young founders “oneshotted by tiktok edits of the social network or performative “how I got into YC” videos.”1 This is largely how I viewed Roy before he launched Cluely as well. Ex post, for better or worse, the company and its viral founder may grow beyond this. That is my question for today.

An alternative lens of Larry Ellison frames him as someone with immense conspicuous consumption needs, but not motivated by greed per se. Rather, he is driven by glory; he wants to win and be recognized. Of Ellison:

“The argument was his true love, the idea itself just someone he dated. The subject he liked best was himself. He was forever telling people “how wonderful he was, how smart he was, and how rich he was going to be,” Feigin said.”

Ellison cultivated a hard-driving sales culture where product came second. Often, he would promise features and not deliver, but by locking customers into multi-year contracts with sticky enterprise software, Oracle would still get paid.2 A distribution first approach works well in B2B SaaS sales owing to this. It does not, however, work well in consumer software, where switching costs are nonexistent and multi-year contracts seldom exist.3 There, an excellent product is all but essential.4

Roy declares himself to be driven by glory; as he put it today on TBPN, “the only two things I really care about in my life is [are] working on something that I find interesting and getting the work seen by people.” Yet thus far, he and Cluely haven’t found glory, they’ve cultivated fame. Fame is being known, while glory is being known for great achievements.

As I noted earlier this week, for Cluely, such fame’s de facto function has been as a highly successful capital raising tool, not as a sales platform. Despite early revenue generation, they still lack an insanely great product. To that end, if they remain focused on fame and distribution, I find it unlikely they will succeed. Conversely, if they have pioneered leveraging fame as a tool for capital raising with the explicit end goal of product development, that’s a different story.

It seems a16z’s thesis is exactly within the latter bound. As a16z’s Bryan Kim argued today on TBPN:

“The calculated risk here is that Roy can convert this awareness into people clamoring to work at the company that are highly high and exceptional—great people to build products—and then use that to continue to iterate on innovation on the product format that is already amazing. So that’s sort of the bet if you will.”

However, culture eats strategy for breakfast. Despite later commitments to product development, Roy demonstrated this when declared today “whatever you thought—the viral content, you thought this was cool—bro we’re doing 10x this!” I still remain skeptical of the extent to which it can he can shift the culture to focusing on products. The product development we see in the next six months should offer an answer.

There are only two reasons you should ever have these cards:

1. You travel a lot

2. You own a high volume low margin business

These people know who they are, so it’s not worth discussing why it makes sense for them. If you work a typical job, you do not check either of these boxes.1 Yet millennials and Gen-Z with low six figure jobs cosplaying as the coastal elite (“People”) sign up for these cards en masse. They do so for three reasons:

One, they want to feel prestigious. This is the probably the most detestable reason. Besides being indicative of the worst kind of people, the bonus points benefits (5x on flights & hotels) are exactly the kind of discrete purchases that you do not need to flash a card for. When someone puts a Platinum Amex down to pay for dinner, it not only conveys their superfluousness and superficiality, but it also screams “I’m financially stupid.”

Two, they want the airline benefits. These are considerable. Yet with four weeks of vacation a year, that will never be taken advantage of. The lounges suck. The food is bad and the lines are long. CLEAR is crowded There’s a reason no one worth their salt is in these lounges: they’re flying out of Teterboro, they’re in the business class lounge, or they’re young and smart, so they go straight to the gate and read a good book.

Three, they think the aggregate benefits are worth it. This is the pitiful trap at the top of Dunning-Kruger’s “Mount Stupid" where smart, budget conscious young folks often find themselves. Sure, with the airline fee credit, hotel credit, Uber credit, digital entertainment credit, a points vacation every two years, and a once-a-year fee-waived foot massage from a yak in Tibet (if you book through their app and remember to pet) you can make up for it. Spend that time doing something else. There are far more interesting ways to waste the precious hours in the day, and much better ways to save money.

The problem with luxury goods is that their value is inversely correlated with their TAM. In other words, the larger the market for a given product, the less prestigious that product is. Arnault and Pinault know this well, and take great steps to prevent prestige dilution. Acquisitions provide an auxiliary lever through which to defend against this dynamic and maintain their products as Veblen goods.

JPMC and AMEX lack such tact and position. They also suffer from the tragedy of the prestige commons; while unit economics improve with volume, brand equity and the very (service based) businesses they’re running—lounges, upgrades, etc.—diminish in quality. As such, the increasing popularity of these cards all but assures they’ll only get worse.

Opt for cash, debit, and if you insist on a credit card, find a simple and cheap one on NerdWallet.2 But most of all, worry about something else.

Obviously, there are the Alexandr Wangs, Brian Cheskys, and Mark Zuckerburgs of the world: visionaries whose first companies—within 5-10 years of founding—are among the most valuable in the world. Then, there are those who parlay a first company into a second, and sometimes, a second into a third and so on and so forth. Take Elon Musk, Peter Thiel, the Collison brothers, and Jack Dorsey for instance.

Often, some of these individuals are secretly a member of the third camp: those who failed first. Known canonical members of it include Sam Altman, Delian Asparouhov, and even Travis Kalanick (if you take him at his word). By failure, I mean everything between an exit well below the last markup (e.g. Loopt) and an outright cessation of operations (e.g. Scour). This is a slight derivation of the dichotomy pointed out by X user Luke Metro, who proposed:

“There are many blog posts written on how to have a successful tech career or how to start a successful company. But there’s almost nothing on how to leverage your clout into a big payday at the next thing, while still ostensibly remaining employed. Big gap in the market here.”

I aim, naturally, to begin to fill a version of that gap herein. We can address this through the three aforementioned examples and a few more:

• Altman “walked away from Loopt shaken and sad” after an “acquihire” well below their last markup.

• Asparouhov’s Nightingale reached ramen profitability but failed to achieve breakout success and sold for “a small amount to one of our clients.”

• Kalanick’s Scour, which he claimed to be a co-founder of, was sued into oblivion and sold for parts.

• Justin Kan’s first project, YC funded Kiko Calendar, sold the rights to its software on eBay after shutting down.

• Apoorva Mehta famously launched twenty failed startups before founding Instacart.

Of course, the single most important thing is the generic advice to never give up and keep trying. Stay in the arena, continue learning, and you’ll eventually win. David Senra of Founders Podcast captured this well in an anecdote about Rafael Nadal and Larry Ellison excerpted from Guthrie’s The Billionaire and the Mechanic:

Then Larry abruptly stopped himself. “Forget everything I just said. The answer is simple. I never give up.” Earlier, Nadal had said something that made a deep impression on Larry. When asked if he loved winning, Nadal shook his head and replied, “No, I love the fight. If you fight hard the winning will come.” Larry loved the fight.

The obvious can often be the most insightful. How do you win? You keep trying.

Yet there’s a second—perhaps more sagacious and actionable—insight to be drawn: network building during and after their failure enables a subsequent rise. Altman came out of Loopt firmly enmeshed in the YC network; so did Kan. Similarly, Mehta caught his first break there as well. Asparouhov found a penchant for early stage investing, gaining prominence at Khosla and Founders Fund. While Kalanick tried again and scraped along for several years, after a modest exit from Red Swoosh, he too moved to San Francisco and started angel investing before founding Uber. Similarly, though not a failure per se, following Jason Calacanis’ own modest exit, he went to work for Sequoia, laying the groundwork for much of what he is known for today.2

One should, as always, avoid conflating correlation with causation in ascribing success to these choices. Yet the aforementioned path wherein failed founders take the time to build networks during or following a failure affirmatively set them up for future success. In the process, many also found a mentor; Altman found Graham and Asparouhov found Rabois.

In other words, if you’re intent on building but not yet ready, make sure you’re laying the groundwork. Find a VC to work for or a mentor learn from, or better yet, both. Sometimes, à la Jcal, you might just stumble into being the first check into Uber.

A mentee of Keith Rabois, himself a mentee of Peter Thiel, Delian has experienced an exponential rise over the last decade under Rabois’ mentorship, first as his CoS at Khosla, and later under him at Founders Fund.1 Now, Delian’s gone to strike out on his own with Varda. Teaming up with former SpaceX engineer Will Bruey, the two produce a potent combination of engineering prowess and, perhaps more importantly, networks with deep pockets that can de-risk the capital hurdles needed to successfully build an ambitious space drugs company. It is this latter point, networks, that I want to discuss today.

Of course, Delian is among the less prominent and youngest millennials in the Thiel orbit. Peter has an immense knack for not only discovering but more importantly cultivating young talent. To name a few, he’s found or accelerated JD Vance, Sam Altman, Mark Zuckerburg, Joe Lonsdale, Palmer Luckey, and even—to an extent and right at the millennial cutoff—Garry Tan to superstar status at the national or Valley level. Other millennials in his circles, including Blake Masters, Jacob Helberg, Trae Stephens, Stacey Ferreira, and Brian Schimpf might be considered largely separate, though no doubt successful.

Delian obviously lacks superstar Thiel millennial status; such celebration is all but exclusively reserved for those with hundreds of millions liquid or national level elected office. Yet I remain hesitant to lump him with the latter group of individuals. In many ways, he’s carving out a unique niche as a superb venture investor and driven entrepreneur. In his half-decade collaboration and support of Rabois, he’s tacitly avoided the pressures of Thiel mentorship.

What pressures, you might ask? I’m reminded of one paragraph from Max Chafkin’s 2021 unauthorized biography of Thiel, The Contrarian:

“On the day the company announced it was shutting down, he invited Andregg over to his house, where he mercilessly crushed him in a game of chess, and then used it to teach a business lesson. He pointed out that several turns before he’d gotten Andregg into checkmate, the younger man had more than a dozen possible moves. “When you’re deciding what to do next, you should be very careful about making moves, because you’ll end up in a situation where you’re constrained,””

Why haven’t we heard of William Andregg? Why hasn’t he reemerged after over a decade? I would argue it’s because Thiel mentorship in its modern form is positively exacting. Thiel—the one that has emerged following the creation of Founders Fund & Palantir, and thus well after he began mentoring Tan, Lonsdale, and Zuckerburg—places immense pressure on young founders. Many falter. Distance can help insulate from such pressure, while also enabling mentees to ascertain his lessons that now amount to a bible for entrepreneurs. Delian, as a protégé of Rabois, has such distance. Sam Altman did too.

Altman actively railed against Thiel’s Stanford Review while an undergraduate student, and presumably held some animosity towards him during his early years in the Valley. He served as an éminence grise to the Collison brothers as Stripe raised their Series A, fending off the circling hawks of Sequoia and tellingly, Founders Fund. As Patrick put it:

“We were these bright-eyed and bushy-tailed naifs from Ireland dealing with Peter Thiel and Michael Moritz, who were to us kind of deities. And Sam, as usual, was completely unintimidated and uncowed by any of them. And his counsel was really very helpful.”

Altman obviously later became quite close to Thiel; he would sit right next to him at (Peter’s husband) Matt Danzeisen’s birthday in 2023. Yet I posit that his early distance, like Delian’s, has allowed him to learn and grow without being smothered.2 Delian has been able to fail freely at Nightingale, Teespring, and in his early angel investments (Eaze, Naytev).3 It set him up with the knowledge and networks he needed for a meteoric rise at Varda. His recent markups on investments in Sword, Eight Sleep, Ramp, and Senra (just to name a few) further reflect such a trajectory.

In other words, Delian has just enough distance from Thiel to access his network without being overwhelmed by it. Regardless of what anyone thinks of it, his support of Trump in 2016 was incredibly prescient, echoed almost exclusively by Thiel himself. Unlike Andregg, Delian has stayed in the game long enough. He is evidence of the fact that you can just do things, if you stay in the arena. With his footing now firmly established, I expect Delian to become among the most prominent figures in the Valley in coming years. Much more on him to come.

It brings in the age old question of distribution versus product. In Cluely’s case, its product consists (in the simplest terms) of wrapping a foundation model with workflow augmentation software (B2B/C AI SaaS). In Rainmaker’s case, from my understanding, they’re providing B2B/G cloud seeding services, akin to what the NCM does in Dubai. Neither are doing anything revolutionary per se, a fact openly acknowledged by the founders. IIRC, Lee called distribution the only moat, and Doricko is frequently placing himself in a position where he must remind lawmakers that his technology is tried and tested.

I don’t intend to over-intellectualize what in some sense could be construed as young founders having fun on the internet. But their explicit ambition invites questions. To that end, if these founders implicitly admit to building a generic technology, why the lack of focus on product? Sure, you could say they are also focusing on product. But every second that Lee spends at the Cluely YC AI Startup School afterparty is one that he spends not building a better product. Same goes for Doricko’s airtime. Further, one could argue that the narrow audience they reach with their five and six figure follower counts on X has limited use in the grand scheme of mass market customer acquisition.

Conventional wisdom (à la Altman, Musk, Jobs, etc.) posits that building a great product solves the need for distribution. As Buchheit wrote, “if your product is Great, it doesn’t need to be Good.” Obviously, startups are built on unconventional takes. But running awry of warnings from these titans strikes the discerning viewer as a questionable choice. Nonetheless, Cluely and Rainmaker have thus far experienced absolutely meteoric success. Meteoric success that is, in terms of raising.

Indeed, Cluely announced a $5.3M seed round in May, led by Susa and Abstract. Rainmaker announced a sizable $25M Series A led by Lowercarbon during the same period. Roy’s now been rumored to be raising a $12M Series A at $120M post led by a16z, though some say he’s holding out (for the three commas). These young founders have captured an audience, one not composed of end users, but rather, venture capitalists.

It is this question of audience that I believe is so critical; Lee and Doricko publicly treat their internet presence as a distribution network. Practically, it operates more effectively as a capital allocation channel. Their actual end-users—a SaaS sales rep in Boston or a municipality in the Central Valley—will seldom be found watching TBPN or scrolling on X. Whether they know it or not (I think they know it), they are advertising their product and vision to VCs, who are evidently buying it.

Is this just conjecture? Certainly! But as Lee described it in the midst of his rumored raise, “im much more dilution sensitive than i am cash sensitive.” In other words, his audience has made him cash insensitive. Perhaps more telling, as he put it yesterday, “distribution is a solved problem for us.”1 Lee is smart enough to know that distribution is a never-ending battle, even if you enlisted some chronically online zoomers in your fight. What might be solved, however, is his short term capital needs. The question of whether such attention and the accompanying capital can be used to convert good products to great remains to be seen. While past performance is no indicator of future performance, I’m optimistic.