The Twenty Minute VCMercor CEO on Why Application Layer Companies Have No Defensibility & Token Spend Exceeds Salaries
At a glance
WHAT IT’S REALLY ABOUT
Mercor CEO argues models win, apps commoditize, compute spend explodes fast
- Foody says application-layer software built on top of frontier models will be hard to defend because models are rapidly absorbing product functionality and can increasingly recreate SaaS end-to-end.
- He argues infrastructure vendors (data, compute, evaluation systems, and scaled expert networks) can build compounding moats via network effects, data inventory, and long R&D cycles.
- Mercor claims strong growth and profitability, describing a post-incident security response, rapid ARR expansion, and a vertically integrated workflow that makes its revenue more than simple GMV.
- He predicts agent adoption will drive token/compute spend upward (often surpassing salaries) due to Jevons-paradox-like dynamics where better/cheaper models increase total usage.
- On labor and policy, he expects significant near-term displacement but long-run job creation, with new roles centered on training/managing agents and codifying tacit organizational knowledge, alongside advocacy to reduce income taxes for lower earners.
IDEAS WORTH REMEMBERING
5 ideasAssume many application-layer moats will erode as models absorb features.
Foody argues “the model is the product,” and as models get better at building complete apps (not just PRs), software wrappers and workflow logic become easier to replicate, undermining defensibility.
Infrastructure moats are more likely to compound than app moats.
He points to network effects in talent/data aggregation, data inventory, and long R&D cycles (e.g., compute/chips) as sources of sustainable pricing power relative to thin software layers.
Forward-deployed implementation is a stronger wedge than pure sales or UI.
Even if GTM is strong, customers may copy features; differentiation shifts to post-sales work: encoding tacit knowledge, training agents, and operating bespoke deployments that are harder to reproduce quickly.
Token spend can rise even when unit costs fall.
Invoking Jevons paradox, he says capability improvements unlock many more workflows, increasing total inference consumption; Mercor claims internal agent token spend already exceeds employee headcount cost.
Evals will become the enterprise “system of record” for agents and models.
Mercor uses workflow-specific evals to choose models on a price/performance frontier and to enable hot-swapping providers; he expects large enterprises to adopt similar eval-driven governance to reduce model-layer switching costs.
WORDS WORTH SAVING
5 quotesBuilding defensibility in the software layer on top of the models is going to be incredibly difficult.
— Brendan Foody
I feel like throughout the lifetime of the business, I have been through a lot of very stressful moments. That was definitely stressful, but it definitely w- wasn't close to the most stressful one.
— Brendan Foody
Maybe I would say two things. First is that I think over the last two years, everyone has increasingly realized that the model is the product.
— Brendan Foody
2025 was the year of how do you get a model to make a PR in a code base? And 2026 is the year of how do you get the model to clone Slack end to end?
— Brendan Foody
Right now, we're spending more on tokens for our internal agents than we are on employee headcount.
— Brendan Foody
High quality AI-generated summary created from speaker-labeled transcript.
