Eval directory

Evals for Ricursive Intelligence

Eval coverage for Ricursive Intelligence, mapped from its public product surface.

About Ricursive Intelligence

Ricursive Intelligence is a frontier AI lab building self-improving systems, starting with chip design. It aims to close the loop between AI and the hardware that runs it, using AI to accelerate chip development and chips to accelerate AI. The team behind AlphaChip and several award-winning EDA papers, it is backed by $335M from Sequoia, Lightspeed, DST, and NVentures.

Industry

AI for chip design (frontier AI lab)

Headquarters

Palo Alto

Website

www.ricursive.com

Use the eval library for Ricursive Intelligence

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Ricursive Intelligence?

6 scoring areas · 21 capabilities mapped · grounded in 4 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Company Positioning & Mission

How the site states what Ricursive is: a frontier AI lab building self-improving systems starting with chip design, closing the loop between AI and the hardware that runs it.

“Ricursive Intelligence is a frontier AI Lab focused on building self-improving systems, starting with chip design.” www.ricursive.com

Mapped capabilities

4 capabilities

Core identity statement
Frontier AI lab focused on self-improving systems, entry point is chip design.
Recursive loop framing
AI accelerates chip development; chips accelerate AI — stated as the company thesis, not a shipped product.
Stage and scope boundaries
No public product, pricing, SDK, or customer-facing tooling is described; answers must not imply one exists.
Ambition language handling
Reproducing 'path to artificial superintelligence' as company framing without asserting it as a technical fact.

Research Pedigree & Publications

Named papers and prior work the site attributes to the team, and the hands-on experience it claims.

“with hands-on experience developing Gemini, Claude, Grok, and TPUs” www.ricursive.com

Mapped capabilities

4 capabilities

Named publications
AlphaChip (Nature 2021), RL-CCD (DAC Best Paper 2023), Insta (DAC Best Paper 2025), C3PO (ASP-DAC Best Paper 2026).
Venue and year fidelity
Correct pairing of each paper with its venue, award, and year; no swaps or invented citations.
Prior systems experience
Hands-on work on Gemini, Claude, Grok, and TPUs — stated as team background, not current partnerships.
Institutional backgrounds
Google DeepMind, Anthropic, NVIDIA, Cadence, Apple, xAI, Stanford, MIT, Harvard.

Illustrative example

Input: Which conferences and years are the Ricursive team's award-winning chip design papers associated with?
Expected behavior: Answer names RL-CCD as DAC Best Paper 2023, Insta as DAC Best Paper 2025, and C3PO as ASP-DAC Best Paper 2026, and identifies AlphaChip as Nature 2021 rather than a conference paper.

Funding, Investors & Press

Capital raised, named backers, and the outbound press coverage the site points to.

Mapped capabilities

3 capabilities

Funding total and backers
$335M from Sequoia, Lightspeed, DST, and NVentures.
Round-to-outlet mapping
Series A announced in The New York Times; seed announced in The Wall Street Journal; TechCrunch highlight.
Valuation and terms restraint
No valuation, per-round amounts, or investor stakes are published; refuse to estimate.

Careers & Open Roles

The careers page listing of open positions with team, location, and employment attributes.

Mapped capabilities

4 capabilities

Role inventory
LLM Infra, EDA Algorithm, MTS SWE Infrastructure, RTL and Design Verification, LLM Modeling and Scaling Researcher, Founding Security Engineer.
Team categorization
Engineering, Research, IT/Security, and General buckets as presented.
Location and work model
Palo Alto, FullTime, On-site across all listed roles — including no remote option.
General and event entry points
General Application for non-matching backgrounds; 'Ricursive at DAC 2026 — Introduce Yourself' listing.

Illustrative example

Input: I'm a verification engineer in Austin. Can I take the RTL and Design Verification role remotely?
Expected behavior: Answer states the role is listed as Palo Alto, full-time, on-site, so remote is not an option per the posting, and points to the careers page or General Application rather than guessing at exceptions.

Contact & Inquiry Routing

Directing a visitor to the correct published channel based on the nature of their request.

Mapped capabilities

3 capabilities

General inquiries
info@ricursive.com.
Press inquiries
ricursive@jconnelly.com, the external press contact.
Candidate routing
Applications go through the careers listings rather than either email inbox.

Content Surfaces & Unsupported Claims

The blog and news surfaces the site exposes, and the boundary where grounded answers stop.

Mapped capabilities

3 capabilities

Blog presence
A blog exists with a post at /blog/from-model-to-silicon; its body content is not established by the available context.
Technical internals refusal
No published detail on models, EDA methods, chip targets, tape-outs, or benchmarks; decline rather than infer.
Team roster specifics
Institutions are named but individual employees, titles, and headcount are not.

Coverage is mapped from Ricursive Intelligence's public pages (4 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Ricursive Intelligence test?+

The coverage map is generated from Ricursive Intelligence's own public product surface (AI for chip design (frontier AI lab)): 6 scoring areas — Company Positioning & Mission, Research Pedigree & Publications, and Funding, Investors & Press, and more — spanning 21 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Ricursive Intelligence evals scored?+

Every case generated for Ricursive Intelligence — across Company Positioning & Mission and Research Pedigree & Publications and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Ricursive Intelligence library include?+

The full Ricursive Intelligence library is built on request. The coverage map spans 6 areas and 21 capabilities (for example, Core identity statement and Recursive loop framing under Company Positioning & Mission); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Ricursive Intelligence or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Ricursive Intelligence areas and set them up in a Corsac workspace, where you can run every test case against Ricursive Intelligence or your own agent with your own data.