Yariv Adan, General Partner, ellipsis venture
There has never been a better time to be an AI engineer. If you combine technical chops with a sense of product design and a keen eye for automation, you might have even built a highly useful app over a weekend hackathon. So, is it time to pitch VCs? Common wisdom says that if you can find a market gap, deliver real value, and ship quickly, you have the recipe for a venture-backed startup. You are likely watching countless peers do exactly that. But before you join the hunt for a billion-dollar unicorn, you have to ask yourself: would you be better off herding donkeys?
and startups are changing. Not incrementally, but fundamentally. Over the past year, we’ve met team after team doing everything right: moving fast, building useful products, targeting real customer pain, delivering real value. And yet, we passed on many of them. Not because the teams were weak, but because the moats that would protect their value have fundamentally eroded.
The most basic rule of venture hasn’t changed: a company needs differentiation and defensible moats to sustain high-margin success at scale. But what counts as a defensible moat has shifted dramatically, with the bar rising to a much higher level. If your business lacks a genuine moat, whether proprietary data or unique expertise that can withstand an army of highly-skilled AI agents, it will inevitably face disruption within the commoditization kill zone.
Two years ago, we coined the term Commoditized Magic to describe the future we saw AI painting. Technology and products are becoming truly magical, unlocking previously impossible capabilities yet they are almost completely commoditized by frontier models. We remain optimistic about the “magic” part: it introduces a massive economic opportunity by unlocking value that was previously inaccessible. But the commoditization risk is real and disruptive, making entire areas uninvestable.
In this piece, we want to unpack that commoditization dynamic: why the unicorn is even harder to hunt in the current landscape. But we also want to suggest that a new creature, or rather, a very familiar one, is about to emerge: herds of donkeys.

Source: Gemini 3
AI is eating software and services, but at the same time, the unit economics of creating value are drastically changing. The cost, expertise, time, and overall resources required to bring a product to market are spiraling down. That changes everything, and commoditization is rushing in from all sides.
The user as builder. There is a new class of apps replacing previously purchased software: the ephemeral app. Whether it’s a simple prompt that creates an artifact, a Claude Code session, or some combination of skills, tools, and plugins users can now build any app they can imagine. Any experienced engineer knows that building even the most complex module for a single, one-time user is trivial; the traditional complexity and expertise kick in only when making it modular, generic, scalable, and maintainable. A single user-builder is a formidable competitor to an entire SaaS company when it comes to building exactly the app she needs at a given moment. This scales to teams as well, and through organizational memory, beyond that.
The explosion of competitors. As coding agents improve and reach the level of professional human engineers at much lower cost and complexity of management the entry barrier to becoming a SaaS company drops dramatically, leading to orders of magnitude more competitors. The result is crowding at every level, and we already see it in our dealflow. Every use case now has numerous startups attacking it, each starting from a small beachhead where they have some unfair advantage, hoping to expand and win the market. But when they raise their heads, they see beachheads all around them, with no clear differentiation. These companies may deliver real value, some may even be profitable but they don’t make sense as venture-backed businesses.
Venture and startups have always been a numbers game of hits and misses. But when the ratios shift by orders of magnitude, with far more companies, solo founders, and tiny teams all enabled by the same tools, the old rules break down. You end up with many more misses than hits, to the point where the VC model itself stops working.
An argument we often hear is that in a world where software is a commodity, it’s all about distribution: move fast, capture those first customers, and you win. Unfortunately, commoditization and AI are rewriting the rules of go-to-market and distribution as well.
First, there is the crowding problem. If you can move quickly, rapidly prototype an MVP, and sign a pilot, all in four weeks with two people, so can your many competitors.
Second, not only does AI unlock ephemeral, hyperpersonalized apps, but integrating traditional software has also become much easier, quicker, and cheaper. Traditional SaaS products arrive generic and require complex, expensive integration projects, a major source of stickiness and first-mover advantage. In the new world, where these integrations can be automated or regenerated on the fly, those moats are rapidly disappearing. As lock-in effects weaken and the customer no longer needs to worry as much about future support and compatibility, they can focus on what they need now, and who does it best, especially in highly commoditized and competitive markets.
As a result, we expect software procurement AI agents to emerge that replace old, human-led methods. These agents could bid and test in real time for required capabilities, threatening to render brand, distribution, and first-mover advantage largely irrelevant. The economics are clear: when switching costs approach zero, loyalty follows.
Finally, Big Tech is moving up the stack and across verticals. Consider how frontier model providers and platform owners, think email, chat, and docs in the enterprise, or mobile, search, and social for consumers, can now build vertical use cases themselves, faster and better than ever. Google adding AI capabilities directly into Workspace, Microsoft embedding Copilot across Office, Apple integrating intelligence into iOS. These giants are moving into territory that once belonged to startups, leveraging distribution advantages that startups simply cannot match. The ability to develop at much higher velocity applies to Big Tech as much as it does to a two-person startup, and Big Tech starts with a billion users.
This is the new reality in the software and services market, as useful intelligence becomes a commodity.
Is this the end of entrepreneurship, is there no path forward for strong small teams who can deliver quick value to underserved markets? Far from it.
There is clearly a massive opportunity for new unicorns, just with a higher bar. That’s the opportunity we’re focused on as a VC. But we also believe that the superpowers and speed of AI have unlocked another avenue for entrepreneurs, one that doesn’t require venture capital at all.
What if, instead of chasing a single elusive unicorn, you used agents and the low cost of development to automate and scale the creation of value-generating businesses? Can a solo founder build a herd of passive-income-generating donkeys at scale?

Source: Gemini 3
Think about what that looks like in practice. You automate ideation and market research to generate, prioritize, and prune a pipeline of ideas. You automate user research and interviews, customer outreach, hypothesis generation, prototyping, experimentation, and analysis. You bootstrap these businesses, run them in parallel, kill the losers, double down on the winners, and adapt as needed.
Imagine a founder running fifteen micro-businesses simultaneously, each serving a narrow niche targeting an underserved market segment they have access to: one automating compliance reports for small European fintech firms, another generating custom training materials for logistics companies, a third managing invoicing workflows for freelance consultants. Most probably even with geographical focus. None of these is a billion-dollar market. None of them will land on a TechCrunch headline. But each generates steady, sustainable revenue, and together they compound into something meaningful. The founder isn’t managing fifteen teams; AI agents handle the build, the iteration, the customer support. The founder’s job is portfolio management: which donkeys to feed, which to retire, which niches to enter next.
This is the inverse of the venture model. Instead of concentrating risk into one massive bet, you distribute it across many smaller ones. Instead of needing a 100x return on a single company, you build a portfolio where the aggregate outcome is what matters. The math is different, the risk profile is different, and critically, it doesn’t require outside capital, which means the founder retains full ownership and control.
We recommend this path to teams we meet who are doing excellent work but operating in spaces where the moat simply isn’t deep enough for a venture-scale outcome. Often very small and efficient, these teams are perfectly positioned to bootstrap rather than raise. The donkey path isn’t a consolation prize. For many founders, it may be the smarter play.
This isn’t a venture-scale play, and that’s precisely the point. It’s a new avenue for entrepreneurs willing to trade the dream of one massive outcome for a portfolio of smaller, sustainable ones, and to use AI to make that portfolio manageable at a scale that was previously impossible.
We believe there is a real opportunity here, and we’ve started exploring the tools to make it work. Stay tuned.