Rubric AI
Startup building expert-verified rubrics and benchmarks so AI labs can check model answers.
On the record
Checkpoint: Deadline for first paying regulated-vertical customer or verified anchor logos in one vertical
The call
Rubric AI has a real benchmark and a $500K pre-seed check, but no named customer yet, sitting between self-grading labs and billion-dollar data rivals.
Rubric AI helps companies and labs grade AI outputs using expert-written rubrics and human reviewers. The problem is that larger competitors like Mercor, Surge AI, and Handshake already own this space and have way more money and customers, while at the same time, the labs Rubric sells to are building their own grading tools from scratch to avoid paying anyone. The one thing that could save Rubric is if it can become the trusted, regulation-approved grading company for industries like healthcare or finance where companies legally need human experts to sign off on AI decisions, but it has not yet shipped this or proven customers want it.
Why
01
AI pressure: Frontier labs' own rubric-automation research (Think-with-Rubrics, DeepVerifier, OpenAI HealthBench). The labs buying Rubric AI's rubrics are also publishing research that lets models grade themselves without human experts. Could AI do the core job? Yes, likely.
02
Bigger players: Mercor, Surge AI, Handshake (via Cleanlab acquisition). Mercor and Surge already run billion-dollar expert-data and RL-environment businesses that could add regulated verticals anytime. Could a giant copy it? Yes, easily.
03
What protects it. Deep, credentialed clinical expert review (like WHBench) that a self-grading model or general annotation vendor hasn't replicated yet.
What would change this call
Mercor, Surge AI, or Handshake announces a healthcare/regulated-vertical expert-curation practice. Would show incumbents closing the one niche Rubric AI could defend before it lands a customer.
Rubric AI names a paying deployer customer (hospital, pharma, health plan) with confirmed ARR by Q4 2026. Would show it can convert its WHBench research edge into real revenue, not just publicity.
A frontier lab publishes or ships automated rubric/self-verification tools (successor to HealthBench) matching WHBench's clinical accuracy. Would show labs can replace Rubric AI's core deliverable without human experts.
Rubric AI names a paying deployer customer (hospital, pharma, health plan) with confirmed ARR by Q4 2026. Would show it can convert its WHBench research edge into real revenue, not just publicity.
A frontier lab publishes or ships automated rubric/self-verification tools (successor to HealthBench) matching WHBench's clinical accuracy. Would show labs can replace Rubric AI's core deliverable without human experts.
Key numbers
- Mercor annualized revenue passed $2B
- Surge AI revenue over $1B with ~50,000 experts
- WHBench: top model scored only 72.1% on women's health
Named risks
- No independently confirmed paying customers or revenue as of March 30, 2026.
- Core deliverable of human-curated rubrics is a target of frontier-lab automation research (Think-with-Rubrics, 2026).
- Competes against billion-dollar incumbents (Mercor, Surge AI, Handshake) with no structural barrier to entry.
- Very small team of 1-10 employees limits execution capacity at pre-seed stage.
- Public entity confusion with unrelated Rubrik Inc (NYSE:RBRK) and localization firm Rubric.com.
Rivals, as read in July 2026
| Rival | Direction | Evidence |
|---|---|---|
| Mercor | losing to | Mercor annualized revenue grew to over $2B by June 2026 versus Rubric AI's $500K raised |
| Surge AI | losing to | Surge AI reported over $1B revenue in 2024 with about 12 frontier lab clients, dwarfing Rubric AI |
| Handshake | gaining on | Acquired Cleanlab in January 2026 to enter evals and RL environments, reaching $1.1B revenue by April 2026 |
| Scale AI | contesting | Scale's June 2025 Meta deal triggered customer exodus that could favor smaller entrants like Rubric AI |
| OpenAI (HealthBench) | contesting | OpenAI shipped its own physician-curated rubric health benchmark in 2025, overlapping with Rubric AI's WHBench |
Questions people ask
- Is Rubric AI profitable in 2026?
- No. Rubric AI is pre-seed with only a $500K convertible note raised as of March 30, 2026, per CB Insights. No independently confirmed revenue, ARR, or customer count exists for the company.
- Could AI replace what Rubric AI does?
- Possibly. Rubric AI sells human-expert-curated rubrics and reasoning traces, but frontier labs are researching self-generated rubrics and LLM-as-judge verification (Think-with-Rubrics research, mid-2026), and OpenAI shipped its own rubric-based HealthBench in 2025.
- Who is Rubric AI's biggest competitor?
- Mercor, Surge AI, and Handshake. Mercor's annualized revenue hit over $2B by June 2026 at a $10B valuation, Surge AI reportedly topped $1B revenue in 2024, and Handshake reached about $1.1B by April 2026 after acquiring Cleanlab in January 2026.
- Who owns Rubric AI?
- Rubric AI is a private, founder-owned startup backed by Y Combinator as part of its Winter 2026 batch. Co-founder Pragya Saboo previously worked in product at Asana, Oscar Health, and Apella, per its March 11, 2026 YC launch page.
- Does Rubric AI have any named customers?
- No named customer logos appear in any public source as of March 30, 2026. The company claims to be trusted by frontier AI labs, but this is a vendor claim by Rubric AI itself and unverified by third parties.
- What does Rubric AI actually build?
- Rubric AI builds reasoning and verification infrastructure: expert-verified reasoning traces for AI post-training, a runtime guidance engine that checks model outputs against verified traces, and regulated-industry data licensing, targeting healthcare, life sciences, and finance.
Timeline
- 2026-07-24Mercor annualized revenue reported over $2B, per Sacra-linked coverage
- 2026-06-09Fondo blog covers Rubric AI public launch (dated to March 18, 2026)
- 2026-05-01Third-party analysis frames Rubric AI's goal as landing 3-4 anchor customers
- 2026-04-21Handshake reported at $1.1B/year revenue after Cleanlab acquisition
- 2026-03-30CB Insights snapshot: Rubric AI $500K convertible note, no named customers
- 2026-03-11Rubric AI publicly launches on Y Combinator W26 Demo Day
- 2026-03-10Rubric AI publishes WHBench, testing 22 LLMs on women's health scenarios
- 2026-03Mercor suffers data breach, pausing its Meta contract
- 2026-01Handshake acquires Cleanlab to expand into evaluations and RL environments
- 2025-06-19Meta's $14.3B Scale AI stake triggers customer exodus to rivals like Mercor
Revision history
27 July 2026
First assessed. Losing its protection. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources
- Launch YC: Rubric AI (2026-03-11)
- Products, Competitors, Financials, Employees, Headquarters Locations (2026-03-30)
- Domain-Specific Data Infrastructure for AI
- Complete Analysis
- Evaluating Frontier LLMs with Expert-in-the-Loop Validation on Women's Health Topics (2026-03-10)
- Mercor versus Scale and Surge (2026-02-16)
- The AI Training Data Sector Spawns a Near-$100 Billion Valuation Business (2026-07-24)
- Mercor -> History
- Meta's $14.3B stake triggers Scale AI customer exodus (2025-06-19)
- $1.1B/year Indeed for data labelers (2026-04-21)
- Evaluating Large Language Models Towards Improved Human Health
- Think-with-Rubrics: From External Evaluator to Internal Reasoning Guidance