ExpertsFMHow Will People Make Money in the AI Age?
The Startup Ideas Podcast · October 6, 2026
5 Selected Lessons:
heuristic · 2 case · 1 framework · 1 reasoning_pattern · 1
Harness vs. model: the car-engine/car-body analogy
Source
The Startup Ideas Podcast — How Will People Make Money in the AI Age? (October 6, 2026) · at 9:08 ▸
When this applies
when building any product on top of a commoditizing core technology
The takeaway
As the core technology commoditizes, value shifts to the surrounding system (tools, workflows, memory, domain knowledge) built around it, not the core technology itself.
Why it works
The underlying model is a fast-improving but generic engine; wrapping it with industry-specific tools, workflow rules, memory, and checks turns a generic capability into something that understands a specific domain, which is where durable value accrues.
One level up
Durable differentiation comes from the accumulated context and workflow around a commodity engine, not the engine.
Where it stops holding
This stops working when the underlying model itself becomes capable enough to natively absorb the surrounding workflow logic, erasing the need for a separate harness layer.
Detail
approach: Think of the model as the engine (how fast it goes) and the harness as everything else — tools it can access, workflow rules, memory, checks — the body, steering wheel, paint, electrical.
signals:
- started as a wrapper around one model
- added industry tools, workflow rules, memory over time
- became indispensable even as the underlying model improved
rationale: Harnesses are described as 'the new GPT wrapper' — most wrapper startups failed, but the top ~10% evolved into true harnesses with lasting value.
In their words
I think of the harness as the tools it can access, the workflow rules, the memory, the checks, basically everything around the. the engine,
Marketplaces/social networks: moat vs. liquidity tradeoff
Source
The Startup Ideas Podcast — How Will People Make Money in the AI Age? (October 6, 2026) · at 19:31 ▸
When this applies
when evaluating any business model that requires matching two or more sides simultaneously
The takeaway
Weigh a business model's eventual moat strength against the difficulty of bootstrapping the chicken-and-egg problem required to reach it — the latter is more art than science.
Why it works
Marketplaces create strong moats once established because of network effects, but achieving the initial simultaneous demand and supply ('liquidity') is inherently unpredictable and hard to engineer, unlike single-sided businesses.
Where it stops holding
This stops working when one side of the market can be served adequately by existing single-sided tools, removing the need for true two-sided liquidity.
Detail
approach: Evaluate a marketplace idea by separately scoring (a) how strong the eventual moat will be once liquidity is achieved and (b) how achievable building simultaneous supply and demand actually is for this specific market.
signals:
- requires matching two distinct sides (supply and demand)
- early-stage chicken-and-egg problem
- strong network effects once liquidity is reached
rationale: If you can create the marketplace, you do have a true moat — but liquidity is 'more art than science.'
In their words
I feel like marketplaces and social network is more art than science in that way. But I think there's tons of opportunities right now.
Rising cost of digital distribution favors owned audiences
Source
The Startup Ideas Podcast — How Will People Make Money in the AI Age? (October 6, 2026) · at 5:41 ▸
When this applies
when any channel for reaching customers is becoming structurally more expensive over time
The takeaway
When the cost of creating a product crashes but the cost of distribution keeps rising, owning direct audience access becomes the scarce, valuable asset.
Why it works
As more products get built (due to AI lowering creation costs) competition for the same paid channels increases, driving up the price of reaching people even though building a product is now trivial.
One level up
Whenever one part of a value chain gets commoditized, find and own the part that doesn't.
Where it stops holding
This stops working when a niche audience is so small that paid acquisition would actually be cheaper than building organic reach from scratch.
Detail
rule: Prioritize building organic/owned audiences over relying on paid distribution as ad costs keep climbing.
threshold: Facebook ad clicks rose from roughly five cents (circa 2008-09) to several dollars today.
rationale: The price of reaching people has consistently risen even as the cost of building products has collapsed.
In their words
In 2008, nine or whatever, when the Facebook ad platform came out, they were literally charging five cents a click. Now the average click, I don't have the exact data in front of me, but we're talking a dollar, two dollars, even more than that.
Proprietary data as a markup business for AI labs
Source
The Startup Ideas Podcast — How Will People Make Money in the AI Age? (October 6, 2026) · at 5:41 ▸
When this applies
when a new class of buyer (AI labs, or any dominant platform) has an insatiable need for an input they can't easily generate themselves
The takeaway
Identify an input a powerful, resource-rich buyer structurally cannot generate itself, acquire or produce it cheaply, and sell it to them at a markup.
Why it works
Big AI labs lack certain specialized or human-sourced data, so businesses that pay people to produce that data and resell it to labs capture a markup on scarcity.
One level up
Position yourself as a supplier of the one input abundant capital can't easily buy or replicate.
Where it stops holding
This stops working when the data becomes synthetically generatable by the labs themselves, eliminating the need to buy it from outside suppliers.
Detail
rule: Pay people for data the big models don't have, then mark it up to the labs; alternatively, as a solo founder, gather very useful niche information agents/ChatGPT can't get.
threshold: Referenced companies reportedly reaching billions in valuation and hundreds of millions in ARR.
rationale: Data is described as 'the new oil' because it's the scarce input even abundant AI compute/capital cannot manufacture.
In their words
there's all these companies like Mercore, Micro One, who... who are selling data to the big labs and are getting insane valuations, billions of dollars, hundreds of millions of ARR.
Othership: backyard sauna club to commercial membership business
Source
The Startup Ideas Podcast — How Will People Make Money in the AI Age? (October 6, 2026) · at 2:57 ▸
When this applies
when testing any capital-intensive offline business concept before committing to a full location
The takeaway
Validate an experience-based business at minimal cost before scaling into a full, capital-intensive version of it.
Why it works
Starting in a low-cost, informal setting (a backyard) let the founder validate demand and refine the experience before taking on expensive real estate.
One level up
Find the cheapest possible version of your business ('the food truck version') to prove the model before investing in the expensive version.
Where it stops holding
This stops working when the informal MVP can't capture the essential value proposition (e.g. ambiance, scale, or trust) that the real version depends on.
Detail
situation: Founder wanted to create a leveled-up communal sauna experience but offline businesses are expensive to start, especially in high-rent cities like Toronto or Manhattan.
decision: He began by building the sauna experience in his own backyard, inviting friends and community members before committing to a commercial space.
outcome: The concept proved popular enough to scale into paid membership locations in Toronto and New York.
lesson: Prove the experience cheaply before building the expensive full version.
In their words
The first version of Othership was literally in the founder's backyard. He literally made this in his backyard. And obviously that didn't cost too much money.