Synthesis
Software is not a business
Sam Lessin's thesis that, once AI makes code cheap to write and copy, software is a commodity on-ramp to a business rather than a defensible business in itself.
This is mostly Sam Lessin’s thesis, repeated across at least a dozen episodes from January to October 2026. His line: software is not a business, it is consumable like media, ‘code is just content’ and tends to zero, so nobody will build a company on something this easy to copy . What he says to build instead is a compounding asset (data, trust, a marketplace, a ‘dark network’ or a brand) with software used as a cheap or free on-ramp . He extends the argument to agent ‘harnesses’, which he calls copyable markdown files . Other hosts largely echo it: Jessica says software companies are ‘scared’ and that durable businesses need something unique people need plus a business model ; Dave relays that new models can open compiled binaries . Brit is the main counterweight, predicting many services keep stickiness because normal users want ‘good enough’ . Read it as one host’s strong, repeated opinion plus a few supporting voices, not as independent expert consensus. Sam is a Slow Ventures investor and runs software-heavy companies, so the view is not neutral.
Key points
- Sam: software is no longer a differentiator; it is consumable like media and will be copied instantly — 5 episodes. (Sam · SaaS Companies Beware: AI Is T (Jan 2026) ▶ 18:24)
- Use software as an on-ramp and give it away; build a compounding data, trust, network or marketplace asset — 4 episodes. (Sam · Apple, OpenAI & Why Nobody Mak (Feb 2026) ▶ 14:33)
- Zero implementation cost means no lock-in: Sam can switch providers at will and cannot see who makes money — 1 episode. (Sam · OpenAI’s $852 Billion Valuatio (Apr 2026) ▶ 29:38)
- Agent harnesses and bot businesses are singled out as copyable and heading to zero margin — 2 episodes. (Sam · Why SpaceX’s $1.75T IPO Makes (Jun 2026) ▶ 30:11)
- Silicon Valley’s historical edge (knowing how to write software) erodes once everyone has it — 2 episodes. (Sam · Apple, OpenAI & Why Nobody Mak (Feb 2026) ▶ 33:03)
- Even closed source is reported to be fading, as new models can reconstruct compiled binaries (secondhand) — 2 episodes. (Dave · Amir Efrati (May 2026) ▶ 11:12)
- Generating software now competes with buying it from a vendor — 1 episode. (Guillermo Rauch · Guillermo Rauch (Apr 2026) ▶ 18:48)
Where guests agree
- Among the hosts, writing software alone is no longer a moat; the open question is what replaces it (Sam · SaaS Companies Beware: AI Is T (Jan 2026) ▶ 18:24)
Where guests disagree
Do software products lose their stickiness completely? ⚠️
- Sam: zero lock-in, he can switch local, cloud or China and cannot see how anyone makes money (Sam · OpenAI’s $852 Billion Valuatio (Apr 2026) ▶ 39:18)
- Brit predicts many services will keep their stickiness because normal users want good enough and do not want to change (Brit · OpenAI’s $852 Billion Valuatio (Apr 2026) ▶ 30:22)
Frameworks
- Software as an on-ramp, not the product — Sam’s view of what to build when software is commodity: write software cheaply or give it away, and use it to compound an asset users depend on. (Sam · Apple, OpenAI & Why Nobody Mak (Feb 2026) ▶ 14:32)
- Software: cranked out and nearly free
- Durable compounding data asset
- Trust asset
- Marketplace asset where users must keep using you
Stories & examples
- Sam re-implements Granola: Annoyed with Granola’s limits, Sam built his own transcription and summarisation tool; when his first API limited him, Claude pointed him to AssemblyAI, in which he invested 100k in the first round. Lesson: With no implementation cost, switching to a better service is instant, so nobody earns real money. (Sam · OpenAI’s $852 Billion Valuatio (Apr 2026) ▶ 28:29)
- Advice to a successful software founder: Sam spoke to a software founder who had made money and was tinkering with apps; they concluded software is no longer a differentiator and that the founder should build personal brand, trust and distribution. Lesson: If building is cheap, the durable assets are credibility and distribution. (Sam · Founders Beware: Illiquid Valu (Jan 2026) ▶ 25:08)
- Sam says Constellation Software, a roll-up of SaaS companies and his favourite stock, has fallen sharply since May 2025 and is a clean illustration of the SaaS story. (Sam · SaaS Companies Beware: AI Is T (Jan 2026) ▶ 21:24)
Numbers
- Constellation Software share price: from a 5,000 peak to about 3,000 (Sam, stumbling over a 16,000% growth figure): 3000 dollars per share (implied) (host’s framing, approx.) (Sam · SaaS Companies Beware: AI Is T (Jan 2026) ▶ 21:52)
Caveats & warnings
- Jess does not name the software companies or examples she refers to. (Jess · SaaS Companies Beware: AI Is T (Jan 2026) ▶ 18:51)
- Jess says she does not know how to sift for durable companies and may be overthinking. (Jess · Founders Beware: Illiquid Valu (Jan 2026) ▶ 29:52)
Also covered here: Speed and copying
The cleanest statement is guest Ben Broca’s ‘the only moat is speed’, with model companies holding a head start of about three months by knowing the next model first ; his co-guest Henrik Werdelin adds relationship capital as a second moat that compounds with speed . Sam’s version is the cost side: following is much cheaper than trail-blazing, so businesses others can clone with fewer tokens see margins collapse, and with zero implementation cost he switches services freely ; he cites a GitHub competitor needing three people . Dave agrees ideas will be copied and worries platform owners will bundle features, as Apple might with Granola’s category . Jessica says copycats have always existed . Brit predicts stickiness persists for normal users . Sam’s own Granola replacement involved an AssemblyAI investment he mentions, a small stake. The topic is a consequence of the commodity argument rather than an independent finding.
Key points
- When everything can be copied quickly, speed is the only moat; model labs hold a head start by seeing new models early — 1 episode. (Ben Broca · Henrik Werdelin & Ben Broca (Mar 2026) ▶ 44:30)
- Following is far cheaper than blazing the trail; clonable businesses see margins collapse — 3 episodes. (Sam · Apple, OpenAI & Why Nobody Mak (Feb 2026) ▶ 17:59)
- Zero implementation cost means no lock-in: users re-implement or switch instantly — 2 episodes. (Sam · OpenAI’s $852 Billion Valuatio (Apr 2026) ▶ 29:38)
- Generative apps can displace incumbents fast (J Mail), and legacy monopolies may be exposed — 2 episodes. (Guillermo Rauch · Guillermo Rauch (Apr 2026) ▶ 20:12) (Sam · Apple, OpenAI & Why Nobody Mak (Feb 2026) ▶ 33:40)
- Platform owners can bundle and out-privacy startups, so startups’ features get absorbed — 1 episode. (Dave · iPhone Duo, Meta Muse & Why AI (Sep 2026) ▶ 13:52)
- Copying is expected but not fatal: durable businesses need something unique, and normal users resist change — 3 episodes. (Jess · Founders Beware: Illiquid Valu (Jan 2026) ▶ 29:38)
Where guests disagree
Does copyability destroy stickiness? ⚠️
- No one earns real money because switching is instant (Sam) (Sam · OpenAI’s $852 Billion Valuatio (Apr 2026) ▶ 29:38)
- Many services keep stickiness since normal users want good enough (Brit); durable unique businesses always faced copycats (Jessica) (Brit · OpenAI’s $852 Billion Valuatio (Apr 2026) ▶ 30:22)
Stories & examples
- Sam re-implements Granola: Annoyed with Granola’s limits, Sam built his own transcription and summarisation tool; when his first API limited him, Claude pointed him to AssemblyAI, in which he invested 100k in the first round. Lesson: With no implementation cost, switching to a better service is instant, so nobody earns real money. (Sam · OpenAI’s $852 Billion Valuatio (Apr 2026) ▶ 28:29)
- J Mail: A group of people built a better UI for the Epstein files, which became the Wikipedia of the case. Lesson: Generative apps can take over fast. (Guillermo Rauch · Guillermo Rauch (Apr 2026) ▶ 20:12)
- Apple Watch audio intelligence vs Granola: The watch now transcribes conversations all day, privately and without audio; Dave worries about Granola because Apple has effectively built it into the watch. Lesson: Platform owners can absorb a startup’s feature at higher privacy. (Dave · iPhone Duo, Meta Muse & Why AI (Sep 2026) ▶ 13:52)
- EHR monopolies: Cerner, Epic and McKesson sit on shaky monopolies because competing needed too much engineering time; AI might blow them up, though their ecosystems can also move faster. Lesson: Hard-to-replace legacy software may be the first to be disrupted. (Sam · Apple, OpenAI & Why Nobody Mak (Feb 2026) ▶ 33:40)
Numbers
- New-model head start for model companies: 3 months (self-reported, approx.) (Ben Broca · Henrik Werdelin & Ben Broca (Mar 2026) ▶ 44:38)
Caveats & warnings
- Ben says AI lets everything be copied much faster, so the only moat is speed; model companies hold a speed advantage by knowing new models about three months early. (Ben Broca · Henrik Werdelin & Ben Broca (Mar 2026) ▶ 44:24)
In their words
Following is much cheaper than blazing the trail, so businesses that others can clone with fewer tokens see margins collapse. — Sam · Apple, OpenAI & Why Nobody Mak (Feb 2026) ▶ 17:59
How much to trust this
- Evidence: 17 episodes, 3 independent speakers (the four hosts and unidentified host speech count as one voice, the show; guests counted individually).
Blueprints that use this
Sources
- (01)
SaaS Companies Beware: AI Is The New UI (Anthropic's Claude Code and Cowork)
18:24 into “SaaS Companies Beware: AI Is The New UI (Anthropic's Claude Code and Cowork)”
- (02)
Stripe Is Trying To Buy PayPal For $150b | OpenAI Cyber Incident, Frontier Model Economics, Peptides
- (03)
iPhone Duo, Meta Muse & Why AI Finally Scared Dave Morin
13:52 into “iPhone Duo, Meta Muse & Why AI Finally Scared Dave Morin”
- (04)
Founders Beware: Illiquid Valuations vs. Real‑World Taxes
25:08 into “Founders Beware: Illiquid Valuations vs. Real‑World Taxes”
- (05)
Apple, OpenAI & Why Nobody Makes Money in AI | The SaaS Apocalypse
14:32 into “Apple, OpenAI & Why Nobody Makes Money in AI | The SaaS Apocalypse”
- (06)
OpenAI's $852 Billion Valuation Makes Zero Sense (ClawCon Tokyo, Anthropic's Leak, Goodbye Internet)
- (07)
Agentic AI Will Break the Internet (ClawdBot, OpenAI & the AI Demand Delusion)
15:16 into “Agentic AI Will Break the Internet (ClawdBot, OpenAI & the AI Demand Delusion)”
- (08)
Normal People Don't Want AI Agents (And Silicon Valley Can't Accept It)
11:12 into “Normal People Don't Want AI Agents (And Silicon Valley Can't Accept It)”
- (09)
Fire Sam Altman, The End of Software Engineers, and Why AI Is All Narrative
16:24 into “Fire Sam Altman, The End of Software Engineers, and Why AI Is All Narrative”
- (10)
Stripe Buys OpenRouter, OpenAI vs. Anthropic & the AI Backlash
40:41 into “Stripe Buys OpenRouter, OpenAI vs. Anthropic & the AI Backlash”
- (11)
SpaceX Hit $2.5 Trillion. Now Watch What Elon Buys Next.
36:14 into “SpaceX Hit $2.5 Trillion. Now Watch What Elon Buys Next.”
- (12)
Google I/O Just Killed SEO? Agentic Search, Universal Cart, and the New Web
43:54 into “Google I/O Just Killed SEO? Agentic Search, Universal Cart, and the New Web”
- (13)
Why SpaceX's $1.75T IPO Makes No Financial Sense (And Doesn't Have To)
30:11 into “Why SpaceX's $1.75T IPO Makes No Financial Sense (And Doesn't Have To)”
- (14)
Vercel CEO: 70% of Our Traffic Is Now AI Agents "Nobody Was Prepared" | Anthropic, OpenClaw, OpenAI
- (15)
Anthropic’s Bet on Coding Is Working (OpenAI Shopping Pivot, A16Z’s Top 50 List, $1B Tennis Channel)
- (16)
The New Startup Stack: One Founder + Agents | Henrik Werdelin (Audos) and Ben Broca (Polsia)