Executive summary: Mercor built impressive brand equity as an AI hiring platform, raised $492M, and earned a reported $10B valuation. But in 2026, if you're a CTO trying to hire AI engineers — not data labelers — Mercor's current business model is largely irrelevant to you. Here's the full picture.

Mercor launched with a compelling pitch: a 20-minute AI video interview screens candidates once, drops them into a searchable database, and lets employers query requirements in plain language to get a shortlist with video clips. Clean, fast, modern. Employers type what they need; Mercor's AI scans resumes, portfolios, and interview transcripts and surfaces matches. The engineering job listings still exist. You'll find Software Engineer (Trajectory) roles at $70–$130/hr and Software Engineer (Code QA) positions at $70–$120/hr, with 17 and 112 contractors placed recently, respectively. Those numbers sound meaningful until you understand what "Code QA" means at Mercor: it's largely RLHF data labeling — training AI models by reviewing and rating code outputs. That's not the same as hiring a senior engineer to build your product. The hard truth: Mercor pivoted its core business to serving AI labs — OpenAI, Anthropic, Google DeepMind — who need massive pools of contractors to label data and fine-tune models. Engineering staffing for product companies is a marketing artifact left over from an earlier positioning. The pages still rank in Google. The business has moved on.

This is the most important thing to understand about Mercor in 2026. The company's Expert-as-a-Service (EaaS) model — a cost-plus billing structure for high-volume AI model training work — is where the real revenue lives. Their headline stat of $1.5M paid to contractors per day sounds impressive. But that figure reflects the volume economics of data labeling pipelines, not the placement of senior engineers at product companies. The evidence of strain in that model is significant:

For data labeling at scale, these tradeoffs might be acceptable. For hiring the kind of AI-native engineers who will rebuild your product, they're disqualifying signals.

The sourcing UX is genuinely good. Typing a requirement and getting video clips of candidates answering relevant questions compresses the early sourcing stage meaningfully. For teams that need to move fast on contract work, this is real value. The problem is depth. A 20-minute AI interview captures surface-level communication and self-presentation. It doesn't evaluate system design thinking, AI toolchain fluency, or the architectural judgment that separates a $130/hr engineer from someone billing that rate but delivering $40/hr output.

The single-interview-for-multiple-companies model is genuinely candidate-friendly — in theory. In practice, widespread reports of ghosting after the AI interview suggest the funnel is better at collecting candidates than connecting them with employers. For engineering roles specifically, candidates report long wait times with no feedback, a pattern that reflects the mismatch between Mercor's current business focus and its engineering-facing marketing.

Mercor has no public pricing. This is a deliberate enterprise sales strategy, not an oversight. Third-party analyses estimate the model as either:

For a $100/hr engineering contractor, a 30% markup means you're paying $130/hr while the contractor sees $100. That's a market-rate margin for staffing, but without transparency, you can't benchmark it, and you can't negotiate from data. Compare that to platforms with published fee schedules: you either know what you're paying or you don't. Mercor's opacity is a budgeting problem for engineering leaders who need to model contractor costs across quarters.

Altman's point applies directly here: the infrastructure for AI development — including the human capital layer — needs to scale with precision. You can't scale hiring intelligently when you can't model costs.

For AI lab data labeling roles, Mercor's supply is genuinely large and arguably the best in class. If you need 500 annotators with Python backgrounds who can evaluate code quality for RLHF — Mercor is probably the right call. For product engineering — building features, owning services, shipping AI-native applications — the quality signal is murkier. The same pool of candidates serves both use cases, and the vetting process (a single 20-minute AI interview) wasn't designed to distinguish between a strong data labeler and a strong product engineer. Those are different jobs requiring different evaluation. The Code QA category is the tell: 112 placed recently sounds like traction. But "Code QA" in Mercor's context means evaluating AI-generated code for training purposes — not QA engineering in the traditional software sense. Employers searching for engineering talent need to read every job category carefully.

Employer feedback is thin in public review forums — Mercor operates largely through direct enterprise relationships, not self-serve channels where reviews accumulate. What does surface:

Candidate sentiment, particularly on forums like Reddit, skews negative in 2026 around the data labeling contractor experience — specifically around the pay cut episode and monitoring software requirements. Senior engineers with options are self-selecting away from Mercor's contractor pool.

Nextdev was built for exactly the use case Mercor has drifted away from: hiring AI-native software engineers for product companies.

The structural difference matters: when a platform's primary revenue comes from data labeling pipelines, its incentives around engineering talent diverge from yours. Nextdev's business only works if great engineers get placed at great engineering jobs. That alignment is the whole model. Mercor's 20-minute AI interview is an efficient funnel for volume. Nextdev's vetting evaluates what actually predicts engineering performance in 2026: AI toolchain literacy, architectural judgment under AI-augmented conditions, and the ability to leverage tools like Cursor, Claude, and GitHub Copilot as genuine force multipliers rather than autocomplete.

Use Mercor if:

Look elsewhere if:

Mercor is a well-funded, well-marketed platform that solved a real problem — and then pivoted to a different real problem. That pivot was rational for Mercor's business. It's not rational for yours if you're trying to hire software engineers who will build AI-native products. The $10B valuation reflects Mercor's position in the AI lab supply chain, not its value as an engineering hiring platform. Those are different markets. Confusing them is expensive. The engineering teams that will win in 2026 aren't the ones that hire the most contractors from the same pool feeding OpenAI's data pipelines. They're the ones that hire fewer, better engineers with genuine AI fluency — and they need a platform whose entire operation is organized around finding exactly those people. That's a different product. It's what Nextdev is built to do.