rekursiv.ai
Two-person YC S26 startup building autonomous AI agents that do scientific research.
On the record
Checkpoint: Deadline for rekursiv.ai to name a paying customer or exclusive data partnership
The call
A tiny team with strong builder credentials is selling an orchestration layer that Google, Edison, and Sakana already ship at scale, with no customer or data lock announced yet.
Rekursiv.ai is a two-person team making software that helps researchers use advanced AI models to run experiments and solve scientific problems. The challenge: Google already launched a competing product in May 2026 and deployed it to government labs, while other companies like Sakana AI and Edison Scientific have raised far more money and have real customers paying to use their tools. Because Rekursiv uses the same public AI models as everyone else (from OpenAI, Google, and Anthropic) and has no special data or locked-in customers, there is nothing stopping Google, OpenAI, or Anthropic from building the same orchestration layer themselves, and they have strong reasons to do it first. The one thing that could change this picture: if Rekursiv lands a paying customer, signs an exclusive deal with a data partner, or focuses on one specific industry instead of trying to be a general platform within six months after launching from Y Combinator (around January 2027).
Why
01
AI pressure: Frontier LLMs from Anthropic, Google, and OpenAI. These labs can fold agent-orchestration straight into their models, the exact layer rekursiv.ai sells today. Could AI do the core job? Yes, likely.
02
Bigger players: Google DeepMind Co-Scientist. Co-Scientist reached general availability and is already running in all 17 US DOE national labs as of May 2026. Could a giant copy it? Yes, easily.
03
What protects it. Founders built TensorFlow Probability, VideoPoet, and Dream Machine, giving them speed, but no data, contract, or lock-in blocks copying.
What would change this call
Rekursiv.ai announces a named paying customer or revenue by January 2027. Would show it can convert its tech pedigree into real commercial traction against funded rivals.
Rekursiv.ai signs an exclusive data license in a specific science vertical (materials, catalysis, therapeutics). Would show a pivot from unbarred horizontal platform to a defensible niche.
Google expands Co-Scientist beyond DOE labs or Edison adds pharma partners beyond Incyte. Would show incumbents extending distribution lead while rekursiv.ai has none.
Rekursiv.ai signs an exclusive data license in a specific science vertical (materials, catalysis, therapeutics). Would show a pivot from unbarred horizontal platform to a defensible niche.
Google expands Co-Scientist beyond DOE labs or Edison adds pharma partners beyond Incyte. Would show incumbents extending distribution lead while rekursiv.ai has none.
Key numbers
- Rekursiv.ai runs on a 2-person team
- Edison Scientific: $70M seed, $200 per run, live at Incyte
- Autoscience Institute: $14M seed led by General Catalyst
Named risks
- No proprietary data, regulatory moat, or lock-in; core framework sagent is open-sourced and replicable.
- No disclosed funding beyond presumed YC batch investment while rivals raised $70M to $335M+.
- Category already occupied by better-capitalized, shipped competitors including Google, Sakana, Edison.
- Headline benchmark claims (ARC-AGI, Sudoku) are self-reported, vendor-authored, not peer-reviewed.
- Brand confusion with much larger similarly-named firms Ricursive Intelligence and Recursive Superintelligence.
Rivals, as read in July 2026
| Rival | Direction | Evidence |
|---|---|---|
| Google DeepMind Co-Scientist | losing to | Reached GA and deployed to all 17 US DOE national labs by May 2026 |
| Sakana AI | losing to | $200M+ raised, $1B+ valuation, Nature-published AI Scientist methodology, March 2026 |
| Edison Scientific | losing to | $70M seed, live paying Incyte pharma partnership at $200 per run as of May 30, 2026 |
| Autoscience Institute | contesting | $14M General Catalyst seed, Carl/Mira agents in customer production ML models, March 19, 2026 |
Questions people ask
- Is rekursiv.ai profitable in 2026?
- No. As of July 25, 2026, rekursiv.ai has disclosed no revenue, customers, or commercial traction. The company is a 2-person Y Combinator Summer 2026 batch startup that has published technical benchmark claims but no business metrics.
- Could AI replace rekursiv.ai itself?
- Ironically, rekursiv.ai's whole product is AI agents automating the ML researcher's job: hypothesis generation, experiment design, and iteration. As of July 22, 2026 its own roadmap depends on scaling agent fleets, meaning it is the disruptor, not obviously a target, though it uses no proprietary tech frontier labs lack.
- Who is rekursiv.ai's biggest competitor?
- Google DeepMind's Co-Scientist reached general availability in May 2026 and was deployed to all 17 US DOE national labs, giving it far greater distribution than rekursiv.ai's 2-person team. Sakana AI ($200M+ raised) and Edison Scientific ($70M seed, live Incyte deal) also compete directly.
- Who owns rekursiv.ai?
- rekursiv.ai is privately held with no parent company, founded in 2026 and part of Y Combinator's Summer 2026 (S26) batch, based in San Francisco. No outside funding round or valuation has been publicly disclosed beyond standard YC batch terms as of July 25, 2026.
- Is rekursiv.ai the same company as Ricursive Intelligence?
- No. Ricursive Intelligence is a separate, much larger AI chip-design startup that raised $335M at a $4B valuation by February 16, 2026, founded by ex-Google Brain researchers Anna Goldie and Azalia Mirhoseini. rekursiv.ai is an unrelated 2-person YC S26 startup building AI research agents.
- Has rekursiv.ai raised any venture funding?
- No confirmed funding round has been publicly reported for rekursiv.ai as of July 25, 2026, beyond a presumed but unconfirmed standard Y Combinator batch investment (typically around $500K). No valuation has been disclosed.
Timeline
- 2026-07-25YC directory lists rekursiv.ai team size at 2 employees, no job listings posted
- 2026-07-22Rekursiv.ai publicly launches on YC, solicits outside benchmark problems, claims ARC-AGI and Sudoku results
- 2026-06-12Rekursiv.ai open-sources sagent, a multi-provider self-mutating coding agent CLI, on GitHub
- 2026-05-04rekursiv.ai website live describing Cockpit for Autonomous Research platform
- 2026-03-01Rekursiv.ai blog claims small AI Scientist team produced ML paper for $500 of compute
- 2026rekursiv.ai founded and accepted into Y Combinator Summer 2026 (S26) batch
Revision history
27 July 2026
First assessed. Little protection left. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources
- Autoresearch Tools Compared (2026-06-10)
- Towards Fully Automated Open-Ended Scientific Discovery (2024-08-12)
- Agentic AI in Pharma R&D: Incyte & Kosmos AI Scientist (2026-05-30)
- General Catalyst backs California's Autoscience in $14M round (2026-03-19)
- Machine Learning Startups funded by Y Combinator (YC) 2026 (2026-07-25)
- rekursiv.ai: Scale AI scientists whose own breakthroughs accelerate the next.
- Launch YC: rekursiv.ai: AI scientists that invent new knowledge (2026-07-22)
- rekursiv.ai, Autonomously Create Knowledge (2026-05-04)
- rekursiv-ai/sagent: Multi-provider, hot self-mutating coding LLM CLI (2026-06-12)
- rekursiv.ai Funding
- How Ricursive Intelligence raised $335M at a $4B valuation in 4 months (2026-02-16)
- What happens when AI starts building itself? (2026-05-14)