What determines whether a software company's revenue and technical advantage endure?

Kyle Harrison · Oct 21, 2022

Businesses that aggregate a critical corpus of data and make it accessible and actionable become a new foundational layer that others build on top of, which is why Pave's compensation data is compelling

Ramp did this with expenses and receipts, Toast on restaurant operations, Persona with identity; Pave's compensation database gives a real-time lens on the tech labor market and that data is very hard to assemble

44:08 20VC: Why 75% of Active Investors Will Disappear in the Next Few Years, The Death of "So So" Venture Firms is Coming, The Rise of Blackstone of Venture Firms and What That Does To Venture Returns, How the World of LPs is Broken and more with Kyle Harrison

Phin Barnes · Oct 2, 2023

The most durable investment lesson is to lean into products that become the system of record for an important job to be done, where the product itself drives distribution and deepens user lock-in over time

His early investments were products he thought should exist based on product empathy; working with Notion taught him that software platforms beat applications and that products which grow with the user create attachment and lock-in

40:37 20VC: The Services Model of Venture Capital is Broken, The Best Founders Do Need Help, The Most Important Signals to Assess When Meeting Founders & Why Kids Bring Less Happiness and More Joy with Phin Barnes @ TheGP

Scott Farquhar · Oct 9, 2023

The real AI value will come from combining multiple unique datasets and tying workflows across a large surface area, not from single-model features

Once data and workflows can be connected end to end — e.g. from an error log back to the code author to an auto-deployed fix — that's where the magic happens

Scope: believes Atlassian is uniquely positioned for these workflows

23:56 20VC: Atlassian Co-Founder Scott Farquhar on The Biggest Lessons Scaling Atlassian to $50BN Market Cap; The Four Roles of the CEO, The Funding Round That Net Accel $6BN, The Regrets of Omission and Commission & The Honeymoon Cut Short

Gustav Söderström · Dec 20, 2023 · hedged

Lots of high-fidelity proprietary user data and great user understanding will remain important in the long term, even as models get very powerful

To ask powerful models good questions you need a lot of data about the user; the MedPrompt result showed GPT-4 beating a medically fine-tuned model, but only by embedding a lot of data into the prompt

Scope: 'the bet I would make'; long term

22:47 20VC Roundtable: Spotify, Adobe & Linkedin CPOs on How AI Changes The Future of Product, Why AI is Now the Product, How TikTok Changed Product, Why Cost is the Biggest Barrier to LLM Usage & Why Incumbents Can Adopt AI Faster Than Any Prior Innovation Cyc

Saam Motamedi · Jul 15, 2024

Market-defining SaaS companies win by building deep, sticky workflow for a specific end user, holding pricing power, and reaching distribution before incumbents can copy the innovation

This is the pattern shared by the companies that became iconic, like HubSpot and Figma, while superficial point solutions flatlined

11:38 20VC: Why We Are in a Bubble & Now is Frothier Than 2021 | Why $1M ARR is a BS Milestone for Series A | Why Seed Pricing is Rational & Large Seed Rounds Have Less Risk | Why Many AI Apps Have BS Revenue & Are Not Sustainable with Saam Motamedi @ Greylock

Harry Stebbings · Oct 28, 2024

Choco's proprietary distributor relationships and its data on products, financials and invoicing are durable moats that a better product alone cannot overcome

A competitor could come with better technology but would lack the historical relationships and the data; Facebook open-sources Llama precisely because its value lies in data, not code

6:31 20VC: Why SaaS is Dead | Why AI First Companies Will Win | We are in the Middle of a Cold War for AI Talent | Why Europe is F******* and We Need to Stop Whining with Daniel Khachab, Co-Founder @ Choco

Jason Wilk · Apr 21, 2025

There are genuine economies of scale in this business, most powerfully in underwriting data

The extra-cash product has been issued 130m times, so the system sees ever more repayment behaviour, driving better loss rates and higher credit limits per user; and more customers on the platform lowers per-unit costs with service providers and networks

37:55 20VC: Do Rich Founders Make Better Founders | The Best Performing Fund Would Only Back YC Founders on Their Second Time | Why SPACs Will Come Back | Why Short Sellers Should Be Banned | Is Trump Better for Business than Biden with Jason Wilk @ Dave

Jason Wilk · Apr 21, 2025

Dave's AI cash-flow underwriting dataset will be its main long-term competitive advantage over other neobanks as it expands into new forms of credit

The company is still early in monetization with only checking and the extra cash overdraft product, so the underwriting data gives a big head start in any new lending product

45:44 20VC: Do Rich Founders Make Better Founders | The Best Performing Fund Would Only Back YC Founders on Their Second Time | Why SPACs Will Come Back | Why Short Sellers Should Be Banned | Is Trump Better for Business than Biden with Jason Wilk @ Dave

Anton Osika · Aug 18, 2025

Maximum defensibility comes not from brand but from being a platform that accrues so much ongoing value for the user that they don't want to leave

If the product generates value for you automatically every day on the platform, leaving is unattractive; Lovable is moving from technical co-founder to general co-founder handling admin and finance operations

10:02 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

Amit Bendov · Sep 12, 2025

The competitive battle in applied AI is over two things: proprietary data (context) and being woven into users' workflows

Anyone can load a couple of call transcripts into a frontier lab model and get an answer, but it won't be a good answer and the user won't know it's bad — the real answer requires capturing everything: emails, Slack back channels, and the right information

22:25 20VC: Why AI SDRs are BS and Do Not Work | How to Use AI in Your Sales Team and Process to Win Today | What Skills Do All New Reps Need to Have in an AI First World with Amit Bendov, CEO @ Gong

Alex Bouaziz · Oct 22, 2025

Building and owning your own proprietary knowledge base and layering AI on top of it is one of the best investments a company can make

Owning the data and setting your own parameters means full control and no hallucination, and it lets Deel surface the right answer for very specific global employment, benefits and payroll cases almost instantly — something competitors can't do

Scope: based on Deel's own two-year build of 20,000 articles and 70,000 changing data points

34:17 20VC: Deel CEO Alex Bouaziz on Raising $300M+ at a $17BN Valuation | Deel vs Rippling: WTF is Going On | Management Lessons from Ben Horowitz and Nik Storonsky | Deel's M&A Playbook: Lessons from 13 Acquisitions: What Works & What Doesn't

Oren Zeev · Feb 2, 2026

Simple software products are genuinely at risk from AI, but operationally complex, distribution-heavy, integration-heavy, regulated, data-rich businesses are hard to disrupt

Anyone can now write the technology quickly, but technology is only 5% of such a business — distribution, integrations, licenses and especially data matter more, and incumbents own the data

Scope: stated as a general framework, not specifically about Navan

9:08 20VC: 50% of Funds Will Go Out of Business | Why Growth Expectations Today are BS and Will Not Last | Why Oren Zeev Takes $0 Management Fees But 30% Carry | Why GPs Should Not Tell LPs Their Strategy

Gokul Rajaram · Mar 16, 2026

The test for an early-stage software company is whether its data asset genuinely improves with every interaction and how deeply it is embedded in the workflow versus being a lightweight layer the system of record could replicate.

Those are the only two durable moats available, so underwriting reduces to whether the data compounds and whether the workflow embedding is real.

23:27 20VC: The 8 Moats of Enduring Software Companies: How to Analyse for Durability and Defensibility in a World of AI | Why Dropouts are "AI Maxing" the World & Remote Early-Stage Companies are Dying with Gokul Rajaram

Andrew Dudum · Apr 4, 2026

Hims & Hers will be able to define the gold standard of preventative care because its patient volume generates a knowledge base larger than any health system's

Treating 10–20,000 patients a day exceeds what the largest health systems treat, building unmatched knowledge of demographics, risk factors, biomarkers, treatment efficacy and side effects

Scope: as the company continues to grow

36:16 20VC: Hims & Hers: $4.3BN Market Cap on $2.3BN of Revenue: The Comeback | Why Being Public is 10x Better | The Death of the "Strategy" Hire | Why Performance Marketing is Worse than Brand Marketing with Andrew Dudum

Danny Rimer · Jun 17, 2024 · hedged

A number of long-public companies currently reinventing themselves are likely to be worth hundreds of billions

Draws on his experience as an equity analyst

Scope: no specific companies named

21:39 20VC: Index's Danny Rimer on Investing Lessons from Hits like Figma, Discord and Etsy to Missing Snapchat, Airbnb, Facebook & Spotify | Why Valuation is a Trap and Market Sizing, Signalling and Sector/Geo-Specific Funds are all Noise

Danny Rimer · Jun 17, 2024

King's success shows that an exceptional team grinding at excellence can survive repeated market collapses and reinvention

King's market collapsed three times — skill-based games on web, then mobile, then Facebook mobile channel — and the team kept reinventing until it worked

38:22 20VC: Index's Danny Rimer on Investing Lessons from Hits like Figma, Discord and Etsy to Missing Snapchat, Airbnb, Facebook & Spotify | Why Valuation is a Trap and Market Sizing, Signalling and Sector/Geo-Specific Funds are all Noise

Mike Cannon-Brookes · Oct 13, 2025

Surviving multiple technology decades requires continuous creation and self-destruction, not defending an existing good product

Companies like Microsoft, Adobe and Intuit navigated DOS, client-server, Windows, internet and mobile by always creating; it is very hard to survive decades with just one thing you protect

Scope: about multi-decade technology companies

44:47 20VC: Atlassian CEO on Why Everything is Overvalued & Are We in an AI Bubble | Do Margins Matter & Does Defensibility Exist in an AI World | Is Per Seat Pricing Dead & The Future of Vibe Coding with Mike Cannon-Brookes

Lucas Swisher · Feb 23, 2026

What investors should chase is not revenue growth but a company's talent density and demonstrated ability to reinvent itself across multiple S-curves

Reinvention is extremely hard, and Databricks under Ali Ghodsi has done it repeatedly — from an ELT data transformation layer, to training and running models, to the center of all enterprise data

11:00 20VC: Inside Coatue's $70BN Machine: Why Price Matters Least | Why Mega Markets are the Most Important | How to Assess Durability of Revenue and Margins in AI with Lucas Swisher

Lucas Swisher · Feb 23, 2026

Canva is a genuine platform company because it has repeatedly hopped S-curves and TAMs into a dozen fast-growing products and leaned into AI unusually early

It went from physical yearbooks to online to SaaS to a multi-product suite, and Cliff called him about integrating AI pre-ChatGPT

44:31 20VC: Inside Coatue's $70BN Machine: Why Price Matters Least | Why Mega Markets are the Most Important | How to Assess Durability of Revenue and Margins in AI with Lucas Swisher

Eric Vishria · Sep 25, 2024

Foundational models are the fastest depreciating asset in human history

Scope: largely true, in his assessment

30:03 20VC: Benchmark's Eric Vishria on Where is the Value in AI: Chips, Models or Apps | Why Nvidia Will Not Be The Only Game in Town | The Commoditisation of Foundation Models | Which AI Apps Have Sustaining Value vs Hype and Short Term Revenue

Varun Mohan · Jun 2, 2025

In the short term it is unlikely that any single model provider runs away with a valuable category such that no one else can compete for years

Six months out someone else may find a new, more efficient technique; that requires believing one company monopolizes both all capital and all great ideas

Scope: short term only; more likely to hold the more valuable the category

49:50 20VC: Windsurf Founder on Will Model Companies Own the App Layer | Why Moats Do Not Exist in a World of AI | Why the Notion of Single Person $BN Companies is BS | Lovable vs Bolt & Cursor vs Windsurf: How Does it All End with Varun Mohan

Martin Casado · Jul 28, 2025

Perceptions of model monopoly are distorted by recency: every major model launch triggers assumptions of permanent dominance, and that has consistently not played out

Model releases are episodic — the Ghibli image launch at OpenAI made everyone think image generation was settled forever, and the excitement passed

Scope: specifically flags this conversation happening right after the Claude 4 launch

6:50 20VC: a16z's Martin Casado on Anthropic vs OpenAI: Where Value Accrues | Cursor vs Replit vs Lovable: Who Wins and Who Loses | The One Sin in AI Investing | Why Open Source is a National Security Risk with China

Harry Stebbings · Feb 23, 2026

The durability of software revenue is more questionable and transient than ever because technology superiority now flips between model providers so quickly

Model leadership rotates rapidly between Gemini, Claude and OpenAI, so no technical advantage holds

10:34 20VC: Inside Coatue's $70BN Machine: Why Price Matters Least | Why Mega Markets are the Most Important | How to Assess Durability of Revenue and Margins in AI with Lucas Swisher

Emad Mostaque · May 17, 2023

As building becomes trivial, distribution, data, relationships and product quality become the things that matter — the technology itself is not the point

Building was already getting easier before AI; the unchanging things are customer satisfaction and delivering value, and people get distracted by technology — as with the decentralization talk at a crypto-x-AI event

41:43 20VC: Why the AI Bubble Will Be Bigger Than The Dot Com Bubble, Why AI Will Have a Bigger Impact Than COVID, Why No Models Used Today Will Be Used in a Year, Why All Models are Biased and How AI Kills Traditional Media with Emad Mostaque, Founder & CEO @

Mitchell Green · Mar 28, 2025

Most software companies are not solving complicated, rocket-science technical problems; what determines outcomes is sales, distribution and go-to-market.

Most software problems being solved are not rocket science; the hard part is commercial execution.

10:55 20VC: Why Traditional VC is Broken: How VCs Learned Nothing from 2021 | Why LPs are More Important than Founders & Advice to Emerging Managers | Bull Case for Bytedance & Why TikTok's Ban Doesn't Matter with Mitchell Green, Lead Edge Capital

Richard Socher · Apr 18, 2025

AI startups are going back to startup basics — branding, marketing, sales and distribution decide winners as the underlying technology gets commoditized

Like the era of a hundred photo-sharing apps where only Instagram won: sharing photos was a commodity capability but subtle product details made the difference

46:28 20VC: Foundation Models: Who Wins & Who Loses | How Economies and Labour Markets Need to Change in a World of AI | China vs the US in an AI Race: What You Need to Know | Rich Socher, Founder @ You.com

George Sivulka · Jan 22, 2025

Salesforce is not going to be disrupted by the next generation of AI

Its moat is habitual stickiness and a network effect in human behavior, not technology — the cost of changing human beings' habits is too high, even though a model like Claude could technically build a CRM

Scope: concedes the technology itself is easily replicated

59:32 20VC: Why All AI Companies Are Under-Valued | The Future of Foundation Models: Scaling Laws, Generalised vs Specialised, Commoditised? | From Unable to Afford Rent to Raising $130M From Index and Peter Thiel with George Sivulka @ Hebbia

Harry Stebbings · Oct 13, 2025

High switching costs are a core trait of a great business — Salesforce retains customers despite a poor product purely because switching is a nightmare

Drawing on Hamilton Helmer's Seven Powers, which lists switching costs as one of the seven durable business advantages

36:31 20VC: Atlassian CEO on Why Everything is Overvalued & Are We in an AI Bubble | Do Margins Matter & Does Defensibility Exist in an AI World | Is Per Seat Pricing Dead & The Future of Vibe Coding with Mike Cannon-Brookes

Tom Blomfield · May 13, 2024

Durable value in application-layer AI comes from deeply embedding in a specific industry — its regulation, tooling, language and workflows — not from the AI itself

Most application-layer AI is 80–90% traditional software with 10% AI; the work is figuring out how Procore or a Salesforce CRM or an Oracle database works in that industry, and OpenAI won't do that deep integration industry by industry

43:19 20VC: Behind the Scenes at Y Combinator: The Interview Process | What the Best & Worst Do in the Program | Do the Best All Raise Pre-Demo Day & YC's Fundraising Advice to Startups | Why the Value is in Application Layer AI with Tom Blomfield

Zach Perret · Oct 16, 2024

'Grinder problems' — brutally tedious, unglamorous integration-style work — are the most attractive problems to build a company around, because willingness to grind produces a product no one else is willing or able to build.

Plaid had to hand-build and self-heal ~12,000 bank screen-scraper integrations; no competitor was crazy enough to do that work, so the grind itself became the product advantage. It's not about the intellectual brilliance of the strategy, it's about ability and desire to do the work.

Scope: people inside the company usually resist because the path is long and may not work

11:22 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

Kieran Flanagan · Jul 11, 2025

Marketing and go-to-market become far more important than before because software itself is no longer differentiated

The cost of code has collapsed and everyone is building more product, so differentiation has to come from how you market and go to market

27:59 20Growth: The Death of Growth Teams? | How Hubspot Use AI to Triple Email Conversion | The Future of AI SEO | Why Prompt Engineering is the New Coding | What Every CMO Needs to Know About AI in 2025

Harry Stebbings · Dec 1, 2025

99% of code next year will be written by AI, meaning technology itself is no longer the moat

28:02 20VC: Scale, Surge, Turing, Mercor: Who Wins & Who Loses in Data Labelling | Is Revenue in Data Labelling Real or GMV? | Why 99% of Knowledge Work Will Go and What Happens Then? | Why SaaS is Dead in a World of AI with Jonathan Siddharth @ Turing

Daniel Khachab · Oct 28, 2024

A significant part of a software company's moat is gone, though not all of it

Years of accumulated technology-building — five, ten, twenty years — is part of many companies' moat, and that specific part can vanish very quickly; other moats remain and should be the focus

Scope: not all of the moat is gone; other moats remain and must be prioritised

7:16 20VC: Why SaaS is Dead | Why AI First Companies Will Win | We are in the Middle of a Cold War for AI Talent | Why Europe is F******* and We Need to Stop Whining with Daniel Khachab, Co-Founder @ Choco

Andrew Ng · Nov 17, 2025

Moats are a function of the industry rather than of AI as a technology, and the software moat specifically has weakened while marketplace and brand moats remain

AI doesn't supply a moat answer for most businesses; ten years of accumulated software used to be hard to replicate and no longer is, but two-sided marketplaces, brand and reputational effects still create defensibility

38:31 20VC: Andrew NG on The Biggest Bottlenecks in AI | How LLMs Can Be Used as a Geopolitical Weapon | Do Margins Matter in a World of AI? | Is Defensibility Dead in a World of AI? | Will AI Deliver Masa Son's Predictions of 5% GDP Growth?

Brad Lightcap · Apr 15, 2024

The best test of whether an AI company is durable is whether a 100x improvement in the underlying model excites them; the companies clamoring for the next model are the ones with a clear path to benefiting from better intelligence.

OpenAI can tell the difference empirically — some companies constantly ask when the next model ships, others never raise it.

22:27 20VC: OpenAI's Sam Altman and Brad Lightcap on The Future of Foundation Models: Will They Be Commoditised | How to Solve the Problem of Compute | Open vs Closed: Which Dominates and Why | Which Companies and Verticals Will Be Steamrolled by OpenAI

Brendan Foody · Sep 15, 2025

The most important question an AI company can ask is whether dramatically better models in one to two years would improve or worsen its business

It is the clearest test of whether the business is durable and well positioned for the future

Scope: credits the framing to Sam Altman

53:31 20VC: Mercor: From $1M to $500M in 17 Months: The Fastest Growing Company in the World | How to Think About Margins and Revenue Sustainability in AI | Why Evaluation Benchmarks in AI are BS Today with Brendan Foody

Harry Stebbings · Aug 4, 2023

He would happily disclose his own business's weakness because a competitor would still need eight years of media-company building to exploit it — the moat to get there is too big

Knowing the weakness is worthless without the time and accumulated build required to act on it

Scope: self-described as possibly arrogant; specific to his media business

33:52 20VC: Why "Hire Great People and Get Out of the Way" is Total BS, Why Your Upbringing Can Make You a Worse Leader & A Bentley, Two Nissan Cubes and Becoming One of Macedonia's Largest Employers; The Story of Slice with Ilir Sela

Micha Kaufman · Jun 9, 2025

A head start is what lets a company build brand, community and hard-won knowledge, so collapsing time to clone forces founders to rely far more on luck

Fiverr's nine-month head start was exactly the window used to seed the brand and build a movement; if time to clone is ten seconds you never get that window and all the stars have to align for you

Scope: acknowledges luck always plays a large role in success, but the required portion varies

18:34 20VC: Fiverr CEO: 'If You're Not Adapting to AI, F* You. You're Done | Why "Time to Copy" is the Most Important Metric in Startups Today | Why 99% of AI Companies Today Will Die | Why Governments Will Take Control of AI with Micha Kauffman

Jason Citron · Nov 25, 2024

The way to build a durable company in gaming is to own an established distribution advantage, which pointed to a group-chat app focused on gaming that helps developers bring their content to life

Content alone isn't durable; distribution is what persists over time

Scope: his view as of 2012, stated as the founding thesis

7:47 20VC: Discord's Jason Citron on Why Everything We are Taught About Hiring & Management is BS | Do Richer Founders & Gamer Founders Make Better Founders? | Never Before Told Moments Behind Scaling to 200M Users

Gokul Rajaram · Mar 16, 2026 · hedged

Klaviyo's defensibility hinges on how tight and exclusive its distribution relationship with Shopify is — a preferred-product arrangement analogous to Google's default deal with Apple would make that part of the business hard to displace.

Proprietary exclusive distribution is a real moat even when the product itself becomes easy to build.

Scope: depends on how strongly Shopify actually promotes them; only protects that part of the business

16:07 20VC: The 8 Moats of Enduring Software Companies: How to Analyse for Durability and Defensibility in a World of AI | Why Dropouts are "AI Maxing" the World & Remote Early-Stage Companies are Dying with Gokul Rajaram

Glen Coates · Mar 8, 2023

Combining Shop's buyer base with the Shopify core platform is becoming exponentially more valuable than either alone, and competing with Shopify in commerce would be terrifying because rivals are pure platforms

Shop's aggregator side has reached critical mass, and a great platform joined at the hip to a buyer aggregation engine is very hard to overcome

37:04 20Product: Shopify's VP Product on Why the Founder is Always the Head of Product, What Makes Truly Special Product Managers, Why The Majority of Product Managers Need to Change, Why Top-Down Decision-Making in Product is Good & How Shopify Will Be Bigger

Glen Coates · Mar 8, 2023

Shopify's product strength is unstoppable because a great merchant experience and an improving developer platform are joined to an aggregator buyer pool that is becoming the most powerful buyer population on the internet

It is the combination of platform success and aggregator success, not either alone, that creates the advantage

42:36 20Product: Shopify's VP Product on Why the Founder is Always the Head of Product, What Makes Truly Special Product Managers, Why The Majority of Product Managers Need to Change, Why Top-Down Decision-Making in Product is Good & How Shopify Will Be Bigger

Mati Staniszewski · Sep 8, 2025

Research alone eventually commoditizes and stops conferring advantage, so an AI company must build product alongside research

The advantage deliverable from research is eventually not enough on its own

12:22 20VC: ElevenLabs Hits $200M ARR: The Untold Story of Europe's Fastest Growing AI Startup | The Real Cost of AI from Talent to Data Centres | How US VCs are in a Different League to Europeans | The Future of Foundation Models with Mati Staniszewski

Mati Staniszewski · Sep 8, 2025

Research leadership only buys a one-to-three-year head start over competitors, so it must be paired with product experience and an ecosystem of distribution and brand to be a durable differentiator.

Research advantage decays, but the head start is long enough to build a phenomenal product experience in parallel.

42:45 20VC: ElevenLabs Hits $200M ARR: The Untold Story of Europe's Fastest Growing AI Startup | The Real Cost of AI from Talent to Data Centres | How US VCs are in a Different League to Europeans | The Future of Foundation Models with Mati Staniszewski

Mike Cannon-Brookes · Oct 13, 2025

The right measure of Atlassian's long-term success is that it is still competing, creating and attracting great talent — a company living on as a 'zombie' rather than a vibrant competitor is not worth wanting

At some stage Atlassian has to live on without him or Scott, so what matters is whether it remains a vibrant place delivering value and solving problems

58:11 Durability is measured by vibrancy not survival as a zombie

Alex Bouaziz · Feb 1, 2023 · hedged

International hiring will be one of the big plays of 2023, and Deel can grow through a layoff cycle on the back of it plus tool consolidation

The shift from growth-at-all-costs to profitability means companies can't sustain $500k salaries and tens of millions in monthly burn, and since key markets haven't repriced, hiring internationally is a good way to cut cost while getting the best talent

27:37 Regulatory and country specific complexity creates a durable moat underestimated by new entrants

Harry Stebbings · Jul 11, 2025

Software is not being commoditized; AI will produce a lot more bad software, but truly great software remains differentiated

Speed of design updates and product depth still separate great products

28:34 Great software remains differentiated despite ai commoditizing bad software

Kieran Flanagan · Jul 11, 2025 · hedged

Great software being non-commoditized is true today, but in two to three years it may not be

People said the same about ChatGPT never being good enough to replace the blue links, and that skepticism proved wrong

28:47 Non commoditization of great software is temporary two to three year horizon

Kieran Flanagan · Jul 11, 2025

Platforms with strong data and ecosystems can still build defensible software, but single point solutions are in a challenging position

Strength comes from platform, data and ecosystem; however true platforms are rare and most started as point solutions, so the line between defensible and not is very hard to draw

29:10 Platform with data and ecosystem remains defensible point solutions struggle

Kieran Flanagan · Jul 11, 2025

The core uniqueness of these companies is training data, so services businesses have a structural advantage — what they really provide is the data to train AI

Every company is effectively training AI, and services businesses accumulate far more data to train on

67:04 Services businesses differentiate through accumulated training data

Anton Osika · Aug 18, 2025

Many companies will offer what Lovable offers today, so defensibility comes down to team execution and continually offering much more — better UX and a better value proposition

In the long term competition reduces to execution, not to product features that others will copy

17:36 Continuous execution and ux improvement not features is the only durable moat

Everett Randle · Nov 10, 2025

The moat in AI is still fundamentally technology rather than distribution and data, but it is specifically a talent-scarcity technology moat

building an exceptional AI product is far harder than building a SaaS product — it requires different people and nuanced pipeline design, and very few people know how to build tastefully on top of these models, which is why AI researchers command Lebron-scale contracts

32:21 Talent scarcity technology moat not distribution or data

Harry Stebbings · Aug 25, 2025 · hedged

Defensibility in AI has essentially disappeared and commoditization is pervasive, making it very hard to know where to invest.

Impressive demos are superseded within weeks by new companies doing the same thing, so the 'wow' is transient.

5:56 Rapid copying has eliminated durable defensibility in ai

Brendan Foody · Sep 15, 2025

The raw API business is not a good business and OpenAI should focus more on model customization

APIs have low switching costs and little pricing power, whereas customization is a more defensible opportunity

53:00 Customization not raw api access is the defensible model business

Alex Rampell · Jan 12, 2026

Displacement of dominant software categories historically took years — VisiCalc took about five years to fall from 100% market share toward 50% after Lotus 1-2-3 arrived

VisiCalc took ~5 years to fall from 100% to 50% share and Lotus ~15 years to go to near zero; today cloud, mobile and near-infinite compute mean a marginally better product reaches a billion people overnight

27:28 Distribution speed collapses the time to displace a leader

Immad Akhund · May 12, 2025

AI startups have not lost defensibility permanently; once the current period of churn settles, the same moats as classic SaaS (strong brand, consolidated market position, multi-product suites, enterprise relationships) will apply.

We're in the 'flashlight/fart apps' phase where everyone tries everything and no one knows anything, but the winner consolidates a position, keeps reinvesting in product, goes multi-product, and builds enterprise connections, which is sticky in its own way.

25:45 Current ai churn is temporary classic saas moats reassert once market consolidates

Daniel Dines · Dec 18, 2024

Automating a single task is not the hard part; the differentiator is orchestrating thousands of automations — deploying, monitoring, analytics, and access control over what agents can run and access

UiPath's existing ability to orchestrate robots at that scale is the key differentiator it can extend to agents

15:24 Orchestrating and managing many automations at scale not single task automation is the durable differentiator

Shyam Sankar · Jan 17, 2024

Margin in government software comes from continuously productizing delivery — building infrastructure that manages software in production automatically — not from lock-in or rent collection

Air-gapped, clearance-controlled environments mean that without an automated deployment layer you end up staffing 20 people per environment and can never reach scale

15:07 Productizing delivery into automated infrastructure not lock in creates durable margin in constrained environments

Harry Stebbings · Jun 9, 2025

Fiverr's value lies in its brand, distribution channels and supply-side trust rather than in its front end or payment system

the enterprise value comes from the externalities and surrounding products, not the cloneable surface

18:11 Brand distribution and supply side trust are the moat not the front end or payments

Harry Stebbings · Jun 9, 2025 · hedged

Defensibility has shifted from time-to-clone to the mid-stage of the spectrum, roughly $1M–$10M ARR, where a company collects, cleanses and utilizes data from high-quality customers

time to clone is now nearly nil so no defensibility remains there

21:04 Defensibility shifted to mid stage data collection and utilization not time to clone

Micha Kaufman · Jun 9, 2025

Most once-great companies died because they could not sustain transformational waves, and the interval between waves is now much shorter than the seven-to-eight-year cycles of the past

Like surfing, you must start rowing early to catch each wave; the waves now come closer together

36:22 Shortening intervals between transformational waves make sustained relevance harder

Des Traynor · Nov 15, 2023

The test for whether an incumbent is disruptable is whether you'd reuse any of their code if you rebuilt the product today: if none, it's go time for a startup; if you'd want to borrow their infrastructure, you're just a thin layer on a surviving incumbent.

A Mailchimp challenger with only an AI composer has to rebuild sender reputation, deliverability at 500m mails a day, attribution, reporting, partnerships and widgets, while Mailchimp only has to build the AI composer — you'd have to believe you can build their whole stack faster than they can build your one feature.

25:13 Test disruptability by asking whether youd reuse the incumbents code if rebuilding today

Phil Carter · Sep 20, 2024

A successful consumer subscription business must be built on a core value promise that is both unique/differentiated and enduring

If the value promise looks like a dozen competitors you can't charge a high enough price or retain subscribers; if it isn't enduring you get very high churn

39:27 Unique and enduring value promise is required for subscription pricing power and retention

Peter Singlehurst · Mar 19, 2025 · hedged

The most enduring competitive advantages usually lie not in a particular product but in business strategy and in the culture and character of the founders and organisation, which makes them harder for AI to destroy

A product-level advantage ('my widget is better than yours') could plausibly be beaten by AI, whereas Bending Spoons' advantage is its M&A and shared-services strategy plus its culture, which erodes only slowly

22:02 Founder culture and ma strategy not product features form the durable moat ai cannot easily destroy

Jean-Denis Greze · Jul 21, 2023

There is a hierarchy of competitive advantage: distribution is the highest-confidence way to keep growing at scale, while brand and network effects are more powerful but rare

Ability to win new customers and sell more to existing ones is what lets most companies compound; many successful companies have no meaningful network effects

36:05 Distribution is the most reliable moat brand and network effects are rarer but more powerful

Sarah Tavel · May 6, 2024

B2B software has rarely had network effects; its durable advantage comes from economies of scale — nailing a specific use case, scaling go-to-market, then layering features to charge more

Charging more yields positive net revenue retention and a go-to-market that can scale with improving efficiency, which is how most outstanding B2B value has been built

29:03 B2b software moat comes from economies of scale not network effects

Glen Coates · Mar 8, 2023

Shopify's fundamental weakness is that it does not control any part of the top of the shopping funnel, leaving it permanently downstream of social, news and ad platforms

Top-of-funnel demand originates on someone else's property, so without owning any of it you always depend on another party; direct access to your audience is what makes you resilient

44:06 Lack of top of funnel control leaves platforms permanently dependent on other channels

Harry Stebbings · Jun 2, 2025

Moats do exist once a startup sits inside a very large company: distribution, brand and resources like chip-buying at unparalleled scale make the seven powers increasingly prominent

Two companies starting at the same line are symmetric, but one absorbed into an OpenAI or Anthropic inherits distribution, brand and purchasing power no independent startup can match

23:42 Absorption into a large company grants real distribution brand and purchasing moats

Varun Mohan · Jun 2, 2025 · hedged

A large wall-to-wall enterprise business creates real switching costs that support valuation, even if not yet Salesforce-class lock-in

Being deployed across very large Fortune 500 companies and working properly for large teams carries inherent switching costs for the business

28:06 Wall to wall enterprise deployment creates real though not salesforce level switching costs

Varun Mohan · Jun 2, 2025

Retention in AI coding comes from differentiated capabilities that take months of deep technical work, such as training your own agentic model, not from switching friction — and the same speed of change that made Windsurf known can unseat it

Most products they build don't work; the frontier-comparable agentic model that now processes hundreds of billions of tokens of code a day took many months and was built by learning from user behaviour; nobody had heard of Windsurf ten to twelve months ago, so the reverse can happen

29:00 Deep technical differentiation not switching friction drives ai coding retention and remains fragile to the same speed that built it

Gokul Rajaram · Mar 16, 2026

Software durability can be assessed against eight moats — proprietary data, workflow, regulatory, distribution, ecosystem, network, physical infrastructure, and scale — deliberately excluding brand.

Not all software companies are equal; a structured scoring of structural advantages separates durable businesses from those that can be replicated with cheap code.

11:20 Eight structural moats not brand decide durability

Gokul Rajaram · Mar 16, 2026

Pure software companies no longer have a scale moat — only hyperscalers with their own data centers do — so Salesforce scores about three, similar to Atlassian.

Software used to be a scale game because producing lots of software made further production cheaper; now everyone can produce software as cheaply as anyone else, so scale only counts where there is physical infrastructure.

18:49 Scale is no longer a software moat only physical infrastructure

Danny Rimer · Jun 17, 2024 · hedged

The membership of the most valuable company cohort will keep changing, especially at the hundreds-of-billions level, and today's languishing software companies are languishing because of execution and product-market-fit problems rather than structural ones

The set used to be five companies and is now seven; NVIDIA was not always an obvious trillion-dollar name

21:03 Weak performers fail on execution and pmf not structural decline

Tomer Cohen · Dec 20, 2023

The durable differentiation is building your own integration layer with the models — a dispatcher/router that serves the right data — rather than pushing all information into the prompt

Prompt-stuffing is only an interim solution; a first-serve integration lets you conflate data at the tier level and mask it from the prompt, and LinkedIn's advantage is its data, nodes and graph

25:48 Building a proprietary model integration and routing layer not prompt stuffing is the durable moat

Jake Saper · Mar 10, 2025

Most smaller software companies at $5-100M ARR funded in the last five to seven years will not make it through the AI transition

They face an existential decision about how much to invest in agents, whether customers will adopt them, and whether it's complementary to their core product

16:12 Most small mid arr saas companies funded recently wont survive ai transition