Labor Market Network Effect

The missing network effect

Brendan Foody frames his company Mercor's strategic endgame around a gap he considers the largest unbuilt marketplace on the internet: every other major marketplace eventually got a transformative network effect, and the labor market never did.¹ "We have these enormous networks of drivers we can order via Uber. We have the network for apartments on Airbnb. Yet the closest thing we have to a network effect in the labor market is LinkedIn, which is almost the worst product I use every day." The underlying gap, in his terms, is the absence of a real system of record for human skill: a reliable, structured answer to what a given person is actually capable of and expert in, both inside a single company and across the wider market. Without that system, human knowledge and ability are deployed badly, since the right person rarely ends up matched to the right task. LinkedIn, in this framing, is a self-reported directory rather than a performance-graded matching engine, and it has scale in node count without a genuine effect on match quality.

Why this is a moat rather than a slogan

Adarsh Hiremath names two compounding effects this gap resolves into once it is closed.² The first is an ordinary two-sided marketplace effect, in which every additional company and every additional candidate deepens the pool available to match against, the same shape as Uber or Airbnb. The second is a job-prediction data flywheel, in which outcome data on who actually performs well, and why, lets a platform surface the right person for a role even when that person does not know it themselves. This second effect is presented as the counterpart, on the labor side, to the argument that software itself carries little durable moat as coding agents push the cost of building software toward zero: if the product itself is easy to copy, the durable advantage in a hiring or human-data business has to be the network and the proprietary performance data behind it, which is exactly what is not copyable.

Why it matters now

The thesis gains urgency from the broader claim that closing the enterprise AI data gap requires deploying the right human experts to manufacture context and build evaluations, and a company cannot deploy human expertise optimally without a working system of record for it. The unbuilt labor-market network effect is therefore not a side project but, in this telling, the infrastructure layer sitting underneath both the market for human training data and the future of work more broadly.

Open questions

Whether a credible skills system of record can be built by a marketplace from scratch, as Mercor is betting, or whether it ultimately requires the data already held by incumbent enterprise human-resources and identity vendors, is unresolved. So is whether the resulting performance-data flywheel can out-defend both existing incumbents and the frontier labs' own incentive to build vetting and matching capability in-house once the approach is well understood.