We build the data and environments frontier models train on.
We originate documents and codebases from real work at 4,000 companies. Our environments, evals and trajectories run from real business systems to synthetic coding tasks.
Request a sampleCustomers
- micro1
- Mercor
- Turing
- n8n
- Sanctuary Wealth
- Citadel Securities
- [Frontier AI Lab]
- [Top 3 Global Market Maker]
- [Largest Open Source Model Builder]
- [ ] Names withheld under NDA
Professional documents and the company records behind them.
Real work files from engineering, finance, sales and consulting teams, in their original formats. We process hundreds of thousands every week. Each document comes with its rendered pages, its extracted text and a SHA-256 hash for every file.
Data availability report
Operational Telemetry
A cross-tool activity graph from real companies.
Records from 38 tools, resolved across people, teams, documents and tasks.
Selected inventory
| Source | Unit | Count |
|---|---|---|
| Chat | Messages | 38.4M |
| Emails | 11.2M | |
| File storage | Files | 3.6M |
| Jira | Tickets | 312K |
This report describes availability and structure. It reports no signal, model or predictive result.
Extracted text
gerra / research Working draft Data availability report Operational Telemetry A cross-tool activity graph from real companies. Records from 38 tools, resolved across people, teams, documents and tasks. Selected inventory Source Unit Count Chat Messages 38.4M Email Emails 11.2M File storage Files 3.6M Jira Tickets 312K This report describes availability and structure. It reports no signal, model or predictive result. Gerra Research 01
SHA-256 of this text: d44971a3dc546f21…
Codebases with their full history: every branch, test, merge and revert.
Each codebase comes as a Git repository you can clone. We also deliver 96,000 repo tasks, 9,600 spec-to-repo tasks, 16,000 exploit-and-patch tasks, 9,600 app specs and 12,800 web tasks graded in a browser.
- c99b8051Add wallet setup script with retry logic and deployment ID generation
- ace112faAdd retry logic and error handling to full-deploy script
- 876cc1fdAdd comprehensive README for SettlementEngine Smart Contracts
- 8a1bcb20Add project guidelines and reorganize contracts with restricted admin
- 8c7c2e29Update README for v2.3 API changes and gasless signatures
- b15153cfAdd script to generate and fund test wallets for SettlementEngine
13 earlier commits
- 9739cf7dUpdate README v2.2: simplify seller signature, add gasless transfer
- 130a558aAdd internal balance transfer and gasless support
- 5673843bAdd internal balance transfer with gasless support
- 852592efUpdate README to reflect major refactor and new terminology
- ed9069e5Switch SettlementEngine from auction to bet logic
- 94e5e98cAdd scripts to distribute TESTUSD and fund test wallets
- b2ff4a65Add misc/ to .gitignore to exclude miscellaneous files
- 5f83d5adUpdate README with comprehensive documentation and new architecture diagram
- 3ae590c0Update MockERC20 deployment name to reflect SettlementEngine
- d0102f34Add testnet addresses to README and create setup script
- 51ea1a6aAdd RESOLVE prefix to resolution hash
- 30623902Add comprehensive admin authorization tests
- dddbc666Refactor SettlementEngine to use internal balances and token approvals
- 6bd579beAdd auction-based secondary trading to SettlementEngine
Replaces listings with auctions for reselling positions: a seller opens an auction, buyers bid and the seller accepts one. The auction code lives in a new contract, SettlementEngineAuction, which SettlementEngine now builds on instead of the listing contract.
- Commit
- 6bd579be
- Time
- 11 Jan 2026, 19:26 PST
- Files changed
- 26
- Added, deleted, modified
- 5, 7, 14
- Patch lines
- 3,662
26 files changed
contracts/ · 6
- ~SettlementEngine.sol
- +SettlementEngineAuction.sol
- ~SettlementEngineError.sol
- ~SettlementEngineEvent.sol
- −SettlementEngineListing.sol
- ~SettlementEngineStorage.sol
test/ · 20
- +auction/accept-bid.test.js
- +auction/close-auction.test.js
- +auction/open-auction.test.js
- +auction/register-bid.test.js
- ~edge/authorization.test.js
- ~edge/multi-transfer.test.js
- ~edge/position-transfer.test.js
- ~edge/reentrancy.test.js
- ~edge/state-transition.test.js
- ~edge/timestamp.test.js
- ~edge/zero-value.test.js
- −fuzz/listing-price.test.js
- ~fuzz/state-machine.test.js
- ~helper.js
- −listing/buy-position.test.js
- −listing/cancel-listing.test.js
- −listing/edge-case.test.js
- −listing/list-position.test.js
- −listing/secondary-sale-resolution.test.js
- ~token/token-compatibility.test.js
contracts/SettlementEngine.solFirst hunk of commit 4@@ -2,14 +2,14 @@22 pragma solidity ^0.8.24;33 44 import "./SettlementEngineQuote.sol";5-import "./SettlementEngineListing.sol";5+import "./SettlementEngineAuction.sol";66 77 /**88 * @title SettlementEngine9- * @notice Prediction market escrow and secondary trading10- * @dev Inherits quote and listing logic, deploys as single contract9+ * @notice Prediction market escrow and secondary trading via auctions10+ * @dev Inherits quote and auction logic, deploys as single contract1111 */12-contract SettlementEngine is SettlementEngineQuote, SettlementEngineListing {12+contract SettlementEngine is SettlementEngineQuote, SettlementEngineAuction {
3 earlier commits
- f80f62c7Add Business Source License 1.1 to repository
- 08f1dbddAdd admin controls, protocol fee, and updated env example
- 03c46aabAdd example env, gitignore, and README
Graded environments for training and testing agents.
Each environment runs as delivered and comes with recorded agent runs. We also deliver 1,600 ML competition runs, 800 paper replications and 960 post-training runs.
Each back-office task plants one error in real, de-identified accounting records. The agent is graded on finding the wrong entry and on closing the books to the cent.
Research reproduction audit
Captured run, de-identified
Pass
Audit a one-dimensional convection PINN reproduction packet. Recompute the evidence from samples and distinguish a valid reproduction from fabricated metrics, a boundary mismatch, an unsatisfied residual, missing curvature gain, or insufficient final accuracy.
01 / 06Task
Toolbash
- Intent
Inspect the benchmark packet
- Tool call
Inspect the provided repository and the public benchmark specification.
- Observed result
The task defines a one-dimensional convection PINN reproduction with a reference PDE, initial condition, periodic boundary, and analytical solution. The working tree contains the model, PDE, reproduction script, requirements, and rubric materials.
repository inspected benchmark specification located
Recorded state2 records
surface status repository inspected reference PDE located
- Intent
Check the packet format
- Tool call
Inspect the available reproduction packet and rubric files.
- Observed result
The packet format and rubric are present. The auditor can be written against the normative fields without relying on a precomputed result file.
packet format checked rubric materials located
Recorded state2 records
surface status packet fields checked rubric located
- Intent
Write the auditor
- Tool call
Write a normalized audit artifact that recomputes the packet metrics and returns a classification.
- Observed result
The audit artifact is written. Its public role is to recompute evidence from raw packet samples, rather than trust a reported score.
normalized audit artifact written
Recorded state2 records
artifact status audit.py written output contract recomputed metrics and classification
- Intent
Exercise failure paths
- Tool call
Run local checks covering valid, fabricated, boundary, residual, curvature-gain, and final-accuracy cases.
- Observed result
The local checks return the expected classifications for the valid reproduction and each named failure mode, including insufficient final accuracy.
failure-path checks completed outputs summarized
Recorded state2 records
check family status valid reproduction classified failure paths classified
- Intent
Submit the audit run
- Tool call
Submit the completed audit artifact for machine grading.
- Observed result
The captured run reports the auditor as complete and ready for the machine grade.
audit artifact submitted awaiting terminal grade
Recorded state2 records
surface status submission complete grader pending
- Intent
Terminal grade
- Tool call
Evaluate the submitted auditor against six held-out reproduction-audit checks.
- Observed result
The machine grader reports a passing result for all six held-out checks.
verdict: PASS held-out checks: 6 / 6 score: 1.0
Recorded state2 records
grade field value held-out checks 6 / 6 verdict Pass
{
"instance_id": "<withheld>",
"source": {
"operator": "<opaque token>",
"period": "<withheld>"
},
"fault": {
"kind": "<one of the mixed-fault classes>",
"affected_voucher": "<held out from agent>"
},
"oracle": {
"expected_cents": "<withheld>",
"tolerance": 0
},
"scores": ["reconciliation", "diagnosis"],
"deidentification": {
"identities": "stable tokens",
"business": "opaque token",
"amounts_dates_structure": "preserved",
"leak_audit": "passed"
},
"difficulty_gate": {
"frontier_models_tested": 2,
"pass_rate": "0%"
}
}A second recorded run, a training diagnostic, misses its accuracy threshold. Open its replay.
22 products in 7 families
- Professional work & company records3 productsdocuments processed800,000
- Business environments2 productsgraded back-office tasks64,000
- Computer use1 productcomputer-use trajectories96,000
- Code5 productscodebases6,000+
- Research agents1 productML competition runs1,600
- Physical AI3 productsrobot episodes400,000+
- Markets & compute7 productsinvestor messages since 2009426 million