A Response to Sam Lessin’s 2025 Update
Gil Dibner

Last week, Sam Lessin published a 2025 update to his seminal 2023 piece on the end of the factory model of venture capital. The update is the first 20 slides of a 104-page slide deck that is available in a docsend link available on Lessin’s twitter account here. It’s absolutely worth accessing the slide deck and reading through it. Lessin is easily one of the most forward-thinking venture capitalists writing today. Much of our thinking at Angular has been along similar lines.

I’d like to respond to a few of Lessin’s points, because I think the conversation that he is driving should be the central conversation in the venture eco-system today.

• From factory-line to regatta. I completely agree with Lessin on both parts of this argument. From 2013 to 2023, a “factory-line” model of venture capital emerged, with (primarily SaaS) startups seeming to graduate semi-automatically from round to ever-larger round, at ever-higher valuations, led by ever-larger VC funds. This assembly line has ground to a halt, replaced by what Lessin calls the “regatta” model. Like sailing boats in a regatta, startups jockey to position themselves at the starting line at the optimal time and speed - and getting there can take varying amounts of time and capital. Inception (preseed/seed) funds need to gear up to be able to support their portfolio through the twists and turns that lead up to that Series A starting line.

• Emergence of clear swimlanes. Lessin phrases it differently than we have, but we have argued that five VC swimlanes have emerged. We agree with Lessin that many preseed/seed funds are “subscale” given the capital requirements and uncertainty of the path to Series A. We agree with Lessin that massive multi-stage giants ($1B+ in fund size) are both having strange effects on downstream financing and are playing a completely different game. Dan Gray, another very sharp observer of the VC market, has referred to the megafunds as “venture banks” to capture their divergent behavior. While all these changes have impacted the traditional mainstream venture pathway, we believe a new class of “New Series A” firms is emerging to fill the gap.

• Provocative overstatements. In the course of his deck, Lessin makes a set of statements, almost in passing, that are very provocative: software is no longer investible, SaaS is worthless, inbound is dead, etc. I think the right way to understand these statements is how my partner David reacted to them: they are not necessarily true and not designed to be taken literally - but they do reflect things that are getting said by investors and founders. There is a tremendous amount of angst within the technology world about the value of software (in a vibe coding world) and about the validity of the SaaS model. Lessin is acknowledging the reality of these fears. While I share the questions that these statements raise, I do not subscribe to the extreme version of the opinion. Software and SaaS remain important investment areas - but they are much harder than before.

• Fragmentation and hyperspecialization. Lessin argues that a single unified “venture market” has been shattered into many smaller markets, each with their own rules and participants. Where Lessin gets prescriptive, however, is where I think the argument becomes less persuasive or incomplete, if still generally true. It’s probably right that specialization is good thing and that VCs should have a plan for how (1) source relevant investments, (2) find the right follow-on investors for their portfolio, and (3) deliver exactly the return profile (IRR, DPI, etc) that their specific LPs are expecting. I take issue, however, with the idea that these objectives are newly important (they have always been important) or that a VC can plausibly mastermind the alignment of all these objectives (if a VC can do it once or twice in a portfolio that would be great).

• AI at seed has been a washout. I largely agree that AI has been a bit of a head fake for most seed investors in most cases. While the excitement and promise of AI have allowed a lot of investments to get made, I’m growing increasingly skeptical about the defensibility of many of these businesses. We’re investing very selectively in vertical AI applications and enabling AI infrastructure where we think there are potential barriers to entry, but we’re being careful to ensure that our thesis is based on more than AI alone. Broadly speaking, AI appears to be commoditizing the value of software faster than it is generating it (a phenomenon that is not unusual in the history of technology) - so we are looking for specific exceptions to that dynamic.

The end of the anomaly. As I read Lessin’s extremely thoughtful, engaging, and almost heretical deck, my main conclusion is that we are at the end of an anomalous period. Perhaps venture is not fundamentally changing. Perhaps it is returning to what it has typically been: a very uncertain game with no rules. Lessin’s description of the last 10-15 years as a “factory model of venture capital” is absolutely accurate, but that period was anomalous and not representative. The factory model period was anomalous because follow-on financing was nearly automatic at every stage, expectations and multiples were irrationally high, and nearly every startup was afforded the benefit of the doubt that they would somehow be able to build a platform of infinite value in a new category. It was anomalous precisely because many of us became convinced that maybe rules and playbooks and “SaaS napkins” could lead us to extraordinary returns. And perhaps they could - for a time. But that time is now solidly over - and venture has returned to its chaotic unscripted essence.

I am old enough to have been trained by a generation of venture capitalists who built their careers before the factory model emerged, before a long list of hundreds of emerging managers competed for every SaaS deal, and before Softbank and Tiger Global taught the rest of the megafunds how to scale up beyond any rational size. If you compare VC in 2025 (as described by Lessin) with VC in 2005 (when I started), the similarities seem to outnumber the differences. In previous eras, the most interesting venture investments were always breaking the mold. This was true of Google, AirBnB, Uber, Github, Airtable. The most interesting companies we are looking at are in a set of one. There is no obvious follow-on funder. There is no obvious exit market. There is no playbook. There is just a brilliant and hungry founding team with the potential to build a real business against a constantly shifting technological backdrop that simultaneously enables the next innovation while commoditizing the innovations of the past.

Engineering Success in a Technical Startup
For technical founders, extraordinary success is highly correlated with how obsessive a founding team is about developing an extraordinarily efficient GTM machine.

What if an AI Was Your Best Customer?
Gil explores how AI agents autonomously selecting and consuming tools can reshape the software industry.

Seven Signals
An evolving VC investment decision framework.

Five Emerging VC Swimlanes
The venture market is evolving into five distinct swim lanes, each with its own rules, characteristics, and key players.

ENTERPRISE/TECH NEWS

Farsight. Palantir Technologies has emerged as the leading stock on the S&P 500, with a 49.1% increase as of April 25, following a 340.5% gain in 2024. In contrast, the "Magnificent Seven" tech giants - Meta Platforms, Microsoft, Apple, Amazon, Nvidia, Alphabet, and Tesla - have underperformed in 2025, with all seeing declines. Tesla leads the drop with a 29.4% year-to-date loss.

Testimony. Kevin Systrom, co-founder of Instagram, made headlines with his testimony in a federal antitrust trial against Meta Platforms Inc. Systrom alleged that Meta CEO Mark Zuckerberg deliberately underinvested in Instagram following its 2012 acquisition, fearing that the app's rapid growth posed a threat to Facebook's dominance.

(Dis)Union. 300 of the 2000 staff in Google’s DeepMind team in London have sought to join the Communication Workers Union. The FT reports that the unusual move was driven by discontent over recent national security contracts. “Unionisation remains relatively rare across the tech sector, which has long resisted attempts to organise its workforce. But there has been growing activity in recent years, including at Amazon and Apple.”

HOW TO STARTUP

Vibe Coding. Supabase continues to ride the wave of ‘vibe coding’, announcing a $200M Series D fundraise led by Accel at a $2B valuation. The open-source backend-as-a-service platform is now used by over 1.7M developers.

Interpretability. Dario Amodei, founder of Anthropic, writes in his blog that while AI capabilities are rapidly advancing, our ability to interpret these systems lags behind, posing significant risks. “Many of the risks and worries associated with generative AI are ultimately consequences of this opacity, and would be much easier to address if the models were interpretable.”

HOW TO VENTURE

Chat. Ben Smith writes about Marc Andressen’s prolific group chat activity in his piece about ‘The Group Chats That Changed America’. “People during 2020 felt that there was a monoculture on social media, and if they didn’t agree with something, group chats became a safe space to debate that, share that, build consensus, feel that you’re not alone,” said Erik Torenberg, an entrepreneur who was the first employee of the tech community hub Product Hunt.” Erik Torenburg, who was behind a number of these group chats recently joined a16z, and Andressen addressed directly the outcome of these conversations in a piece about Erik joining the firm: “Over the last few years, Erik has been at the intellectual frontier, curating some of the most important conversations in technology and big ideas. On a personal level, I appreciate how he’s opened my mind on how people can change. Some day, the full story of group chats will be written, and Erik will have played a valuable role in facilitating the vibe shift.”

Spin In. Intel has decided not to proceed with the previously announced spinout of its corporate venture arm, Intel Capital. This reversal raises questions about the future direction and autonomy of one of Silicon Valley's most active VC entities, which has invested over $20B in more than 1,800 startups.

Reco announced that it has raised $25M in fresh funding, bringing its total capital raised to $55M. The round was led by Insight Partners, alongside Zeev Ventures, boldstart ventures, Angular Ventures, and Redseed.

LightSolver’s CTO Chene Tradonsky explores the promising developments and existing challenges of light-based computing.

Aquant announced two new features to address the unique demands of field work: AI in offline mode and call assist.

Aquant
Implementation Specialist (Boston)

Paradime
Product Engineer - Backend Python (Remote)

Reco
Sales Engineer (US Remote)

Tensorleap
Algorithm Developer (Tel Aviv)