micro1

What it is

micro1 is a data lab that sells expert human data to AI labs, and it no longer sells anything a talent team can buy. The homepage reads “the data research company to accelerate AI in the real world.” The product line is three items: Realm, reinforcement-learning environments and frontier evaluations across 100+ fields including finance, medical, legal, coding, STEM, vision-language and audio; Cortex, an evaluation layer that diagnoses why agents fail in production and produces targeted expert data against those failure modes; and Robotics, expert-demonstrated data for robotics labs. Every one of them routes through a project-scoping form aimed at frontier labs.

Zara still exists, and if you evaluated micro1 before, this is the sentence that matters. The AI technical interviewer micro1 sold as “GPT-Vetting” is now a component inside Realm — micro1’s own recruiter agent, sourcing and vetting the domain experts who produce its data, alongside Flow (where those specialists build datasets) and Merit (the quality dashboard). Zara interviews for micro1 now. It is not a product pointed at your funnel. The vendor has removed the pages that said otherwise: micro1.ai/zara, /hire, /talent and /pricing all 301-redirect to the homepage as of 2026-09-10.

The money explains the move. TechCrunch reported on 2026-08-20 that micro1 had reached a $500M gross annual run rate, up from $100M eight months earlier, retaining 60-70% after paying its contract experts — a net run rate of $150-200M. Founder Ali Ansari moved into data labeling after noticing that data clients were using his vetting platform to recruit annotators. The last confirmed round is a $35M Series A led by 01 Advisors in September 2025 at a $500M valuation.

Why it shows up in Recruiting stacks

It shouldn’t any more. micro1 keeps surfacing in AI-interviewer shortlists for three reasons, none of them a reason to evaluate it:

• The roundups are 2025-vintage. Vendor-comparison sites still list micro1 with $89 and $149 monthly plans and a “top 1% of applicants” marketplace. Those describe a business micro1 exited. Check the vendor’s own product navigation before you scope a pilot: micro1’s nav is Realm, Cortex, Robotics, Research, Expert Opportunities, Data Partnerships. There is no buyer-side hiring product in it.
• Zara’s name still ranks. micro1 publishes candidate-facing interview-prep material under its “For talent” footer, and third-party guides to “passing the Zara interview” keep the AI-interviewer story alive in search. That content is for people applying to micro1’s expert pool, not for teams buying a screening tool.
• Mercor’s shadow. Mercor made the same move toward AI-lab data work but kept its contractor marketplace, so readers assume micro1 did too. micro1 retired its own.

Pricing reality

Nothing is published, and nothing is priced for a recruiting buyer. micro1.ai/pricing redirects to the homepage; Realm, Cortex and Robotics each end at a “Get in touch” form asking about your project, with no tiers, no seat rates, and no self-serve signup.

The $89/month and $149/month self-serve plans that review sites reported through 2025 — carried by an earlier version of this entry — no longer correspond to anything micro1 offers. Treat any 2025-dated micro1 price you find as an artifact of the retired business. The ~$38/hour engineer rate is in the same category: micro1 still pays contract experts, but that is now its cost of goods, not a rate card you hire against. If you are buying data work rather than recruiting software, expect enterprise scoping against volume and domain mix, negotiated per programme.

Best for

An AI lab or an applied-AI team that needs expert-produced training data, RL environments, or a diagnosis of why its production agents fail — bought as a contracted data programme. That is a different budget line and usually a different buyer from the one this page’s readers hold.

Not for

Anyone in a talent-acquisition role, at any company size. If you arrived looking for an AI technical interviewer, a proctored coding screen, or pre-vetted engineers delivered against a req, micro1 sells none of the three and will not quote you for them. Stop the evaluation and move to the list below.

Versus the alternatives

This section is a migration path, not a comparison — micro1 is not in the set any more.

• HackerRank and CodeSignal — the two largest technical-screening vendors and the default replacement if what you wanted was a defensible engineering assessment with an ATS integration and a decade of validity data behind the scoring. Pick HackerRank above roughly 50 engineering hires a year on library breadth and recruiter familiarity; pick CodeSignal when you want a certified score that transfers across roles and req cycles.
• Apriora — the fastest-growing entrant still selling what Zara used to: an AI that conducts the live interview against your own funnel, across technical and non-technical roles. Pick it when interviewer hours are the binding constraint and you want the conversation automated rather than an assessment scored.
• Karat — trained human interviewers as a service. Pick Karat when the reason you wanted Zara was throughput but your legal or hiring-manager stakeholders will not accept an AI conducting a first round.
• Mercor — the replacement for the marketplace half. Same vet-then-match model, same drift toward AI-lab data work, but the contractor marketplace is still a live product with an API. Read its entry for the same vendor-attention caveat that applied to micro1 before the pivot.

Below roughly 10 engineering hires a year, none of these earn a contract. A structured take-home reviewed by two of your own engineers, run against a written interview loop, costs less than any subscription in this list.

Watch-outs

• A 2025 shortlist will send you to a vendor that will not sell to you. micro1 is the second AI-vetting marketplace to reorient around AI-lab data work, after Mercor, and the first to withdraw the hiring product outright. Guard: before scoping any AI-interviewer pilot, open the vendor’s own product navigation and pricing page. No product page and no pricing page means no product — that check would have caught this one in under a minute.
• An inherited Zara deployment has a data question attached. If your team ran Zara through 2025, candidate recordings, transcripts and integrity scores sit with a vendor that has retired the product. Guard: get a written retention-and-deletion answer and export any scored candidate records you still need for NYC Local Law 144 or Illinois AI Video Interview Act recordkeeping before the contract lapses, not after.
• The pivot pattern repeats, and the numbers say why. micro1’s data business went from a $100M to a $500M gross run rate in eight months (TechCrunch, 2026-08-20); no corporate-hiring product in this category grows at that rate, so the vendor followed the deeper budget. Guard: when you shortlist any AI-vetting marketplace, ask in the first call what share of revenue comes from corporate hiring versus AI-lab data work, and write a termination-for-convenience clause you can actually use — see the recruiting AI-use policy for the rest of the vendor-risk checklist.
• Cortex is not a recruiting purchase, whatever the shared logo suggests. It evaluates agents you run in production, sold as a contracted data programme, and buying it restores no screening capability you lost. Guard: route it to whoever owns your agent stack and score it against agent-evaluation vendors, not against your assessment budget.