Should AI startups prioritize defensibility from day one, or focus on speed and growth first and address defensibility later?

Harry Stebbings · May 24, 2023

Defensibility from day one is bullshit

Scope: stated as his strongly held view

31:43 20VC: Why Being First To Market Does Not Matter, Why You Do Not Have Defensibility on Day 1, How to Analyse Market Size and Present it to Investors, Vitamins vs Painkillers; Do Vitamins Survive Recessions and Good vs Great Messaging with Guy Podjarny @ Sn

Guy Podjarny · May 24, 2023

At the beginning a founder should put 90%+ of attention on product market fit rather than on building powers, theorizing about defensibility only as a set-aside hypothesis, and tweaking priorities toward power-building only once something is working.

You have to build something of value before you can sustain or differentiate it; powers matter for what to do next, not for getting the first thing working.

Scope: priorities should be tweaked, not overhauled, after PMF

32:15 20VC: Why Being First To Market Does Not Matter, Why You Do Not Have Defensibility on Day 1, How to Analyse Market Size and Present it to Investors, Vitamins vs Painkillers; Do Vitamins Survive Recessions and Good vs Great Messaging with Guy Podjarny @ Sn

Harry Stebbings · Oct 16, 2024

No startup has defensibility on day one, so demanding it of a seed-stage company is the wrong test.

Pushing back on a partner who dismissed a seed company with two customers and $400k revenue for lacking defensibility.

Scope: about day-one/seed stage specifically

12:33 20VC: Why Founder Mode is Dangerous & Could Encourage Bad Behaviour | Why Fundraising is a Waste of Time & OKRs are BS | Why Angel Investing is Bad for Founders to Do and the VC Model is on it's Last Legs with Zach Perret @ Plaid

Anton Osika · Aug 18, 2025

AI founders should not worry about defensibility from day one — execute fast and grow faster, and only start thinking about defensibility once you're already up there

AI startups are like chickens shot out of a cannon: new chickens are launched every day, so survival comes from flapping faster than the others

Scope: described as a first level of analysis for how to operate

10:51 20VC: Lovable CEO Anton Osika on $120M in ARR in 7 Months | The Honest Truth About Defensibility and Unit Economics for AI Startups | The State of Foundation Models: Long Grok, Short OpenAI, Why | Replit vs Lovable vs Bolt: What Happens

Sarah Guo · Apr 28, 2023

Defensibility does not exist at the seed stage and investors are wrong to look for it

At seed there is no company yet — you're starting with nothing; what you're actually investing in is trajectory and founders' ability to navigate a market and a thesis

Scope: specifically about seed-stage investing

25:44 20VC: In AI Who Wins? Startups or Incumbents? What Happens to Wealth Inequality? Why Will $10BN+ Companies Only Have 10 People | Why Defensibility in Startups is BS & Speed is Everything? Why Large Groups Worsen Decision-Making with Sarah Guo

Guy Podjarny · May 24, 2023

At the very beginning a startup cannot be defensible, with the possible exception of founders holding very unique knowledge in the field

At the start you have nothing to defend

Scope: exception for unique domain knowledge

31:57 20VC: Why Being First To Market Does Not Matter, Why You Do Not Have Defensibility on Day 1, How to Analyse Market Size and Present it to Investors, Vitamins vs Painkillers; Do Vitamins Survive Recessions and Good vs Great Messaging with Guy Podjarny @ Sn

Harry Stebbings · Jul 31, 2024

Sam Altman's heuristic is a good test for founders: if a 100x improvement in OpenAI's model would excite you, you'll get steamrolled; if it would scare you, you're fine

Scope: endorsed as a heuristic rather than a rule

42:17 20VC: Is More Compute the Answer to Model Performance | Why OpenAI Abandons Products, The Biggest Opportunities They Have Not Taken & Analysing Their Race for AGI | What Companies, AI Labs and Startups Get Wrong About AI with Ethan Mollick

Nabeel Hyatt · Apr 4, 2025

Founders should pick a problem hard enough that today's AI can't solve it perfectly and that they could still be innovating on in seven to ten years; if you can fully solve it by the end of a three-month incubator you'll get lapped

That's the nature of a startup — and choosing a hard problem also buys slack if the models take an extra three months to reach the needed fidelity

Scope: an MVP that hallucinates constantly but gets a little paid traction can be the right starting point

55:40 20VC: Why To Win in AI, Investors Need to Change Their Approach | Why VC is Run by Principals and Associates and is a Broken System | The Bull Case for Anthropic & Whether Deepseek Changes Their Strategy with Nabeel Hyatt @ Spark Capital

Guy Podjarny · May 24, 2023

A company with product-market fit but no power — easy to copy, or reliant purely on operational excellence — is in a risky position, so founders should think about power early even if they don't optimize for it first

Power is what lets you build on and grow the business; without it any competitor or a burst of incumbent operational excellence can take your place, leaving you stuck

30:10 Product market fit without a power is fragile so think about power early even while prioritizing fit first

Harry Stebbings · May 24, 2023 · hedged

Deep AI expertise built over a decade in a place like Stanford's AI lab is a genuinely defensible moat of knowledge that others are unlikely to have.

That depth of accumulated AI knowledge is rare and hard for others to acquire.

32:03 Deep technical expertise built over years is a genuine defensible moat

Harry Stebbings · Jul 18, 2022

Defensibility does not exist on day one; it is built over time through the processes a company runs and the products it builds subsequently

Anyone can have an idea and many people have relevant backgrounds — his own media business only became defensible through accumulated process and product

7:41 Defensibility is built over time through process and product not present on day one

Des Traynor · Jul 18, 2022

Real day-one defensibility does not exist, but rare exceptions exist where beautifully crafted, high-end engineering makes a product take a year or more to merely catch up to

With products like Figma or Zenly, the 'road to the starting line' is so long that a copyist would need a year to two-plus years just to reach parity

8:00 Day one defensibility mostly doesnt exist except for rare highly crafted products with a long catch up road

Des Traynor · Jul 18, 2022

Founders tackling a long road to the starting line should build and acid-test their differentiators first, before the table-stakes commodity work

He already believes competent engineers can solve SMTP, IMAP, spam filtering and rich text editors; what he needs to see is the new-to-market difference early, not after two years of catching up to Gmail with three months of runway left

9:21 Build and acid test differentiators before commodity catchup work when facing a long road to parity

Eiso Kant · Oct 7, 2024

The opportunities where companies like Poolside can exist are at the intersection of three factors: how economically valuable a domain already is, how large the gap between model and human capability is, and whether web-scale data exists to close that gap

The bigger the intelligence gap today, the more data is needed to close it, so data availability determines feasibility while economic value determines whether it's worth doing

22:20 Economic value capability gap and data availability jointly determine opportunity