For frontier AI labs

Original data and environments
from real work.

Professional documents. Complete codebases.
Environments with outcomes you can measure.

Supplying frontier AI labs and leading investment firms.

Names withheld under NDA
  • micro1
  • Mercor
  • Turing
  • n8n
  • Sanctuary Wealth
  • Citadel Securities
  • Frontier AI Lab
  • Top 3 Global Market Maker
  • Largest Open Source Model Builder
gerra / researchWorking 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

SourceUnitCount
ChatMessages38.4M
EmailEmails11.2M
File storageFiles3.6M
JiraTickets312K

This report describes availability and structure. It reports no signal, model or predictive result.

Gerra Research01

Structure, shown on a public Gerra report

table · 3 columns

Rows, column headers and units stay connected.
Read the report
Public Gerra research, re-typeset to show document structure. Professional-work samples are available on request.

01 / Professional work

The work product.
And what went into it.

Original documents from engineering, finance, sales and consulting. Editable sources, rendered pages and the company records that explain the decisions behind them.

  • Native documents, tables and figures
  • Reading order, structure and metadata
  • Human reference outputs and task rubrics
Request a sample Also explore whole-company archives ↗

02 / Software

More than the
final commit.

178 complete codebases. 15,238 commits. The branches, tests, failed approaches and changes that explain how real software gets built.

Explore software histories & tasks

Repository history, up close

One example from the 178-repository collection · 23 commits in this history

Discuss access
9 commits11 Jan
3 commits12 Jan
2 commits13 Jan
4 commits14 Jan
1 commit15 Jan
16 Jan
17 Jan
18 Jan
19 Jan
1 commit20 Jan
3 commits21 Jan
6bd579be
11 Jan 2026, 19:26 PST
commit 4 of 23

Add auction-based secondary trading to SettlementEngine

The first hunk of this change is open below.

Discuss access

Add auction-based secondary trading to SettlementEngine

This update replaces the previous listing mechanism with an RFQ-style auction system for secondary market position trading. A new contract, SettlementEngineAuction, encapsulates auction logic, and SettlementEngine now inherits from it instead of the old listing contract. The change also updates documentation and imports accordingly.

26 files · 5 added · 7 deleted · 14 modified · 3,662 patch lines
contracts/SettlementEngine.sol
@@ -2,14 +2,14 @@ pragma solidity ^0.8.24;  import "./SettlementEngineQuote.sol";-import "./SettlementEngineListing.sol";+import "./SettlementEngineAuction.sol";  /**  * @title SettlementEngine- * @notice Prediction market escrow and secondary trading- * @dev Inherits quote and listing logic, deploys as single contract+ * @notice Prediction market escrow and secondary trading via auctions+ * @dev Inherits quote and auction logic, deploys as single contract  */-contract SettlementEngine is SettlementEngineQuote, SettlementEngineListing {+contract SettlementEngine is SettlementEngineQuote, SettlementEngineAuction {
A de-identified excerpt of the recorded patch.
Browse all 23 commits
  1. 111 Jan 2026, 16:04 PSTAdd example env, gitignore, and README
  2. 211 Jan 2026, 18:30 PSTAdd admin controls, protocol fee, and updated env example
  3. 311 Jan 2026, 18:38 PSTAdd Business Source License 1.1 to repository
  4. 411 Jan 2026, 19:26 PSTAdd auction-based secondary trading to SettlementEngine
  5. 511 Jan 2026, 21:05 PSTRefactor SettlementEngine to use internal balances and token approvals
  6. 611 Jan 2026, 21:26 PSTAdd comprehensive admin authorization tests
  7. 711 Jan 2026, 21:30 PSTAdd RESOLVE prefix to resolution hash
  8. 811 Jan 2026, 21:33 PSTAdd testnet addresses to README and create setup script
  9. 911 Jan 2026, 21:37 PSTUpdate MockERC20 deployment name to reflect SettlementEngine
  10. 1012 Jan 2026, 01:29 PSTUpdate README with comprehensive documentation and new architecture diagram
  11. 1112 Jan 2026, 01:33 PSTAdd misc/ to .gitignore to exclude miscellaneous files
  12. 1212 Jan 2026, 16:30 PSTAdd scripts to distribute TESTUSD and fund test wallets
  13. 1313 Jan 2026, 17:00 PSTSwitch SettlementEngine from auction to bet logic
  14. 1413 Jan 2026, 17:40 PSTUpdate README to reflect major refactor and new terminology
  15. 1514 Jan 2026, 12:34 PSTAdd internal balance transfer with gasless support
  16. 1614 Jan 2026, 12:55 PSTAdd internal balance transfer and gasless support
  17. 1714 Jan 2026, 13:08 PSTUpdate README v2.2: simplify seller signature, add gasless transfer
  18. 1814 Jan 2026, 18:13 PSTAdd script to generate and fund test wallets for SettlementEngine
  19. 1915 Jan 2026, 12:13 PSTUpdate README for v2.3 API changes and gasless signatures
  20. 2020 Jan 2026, 23:47 PSTAdd project guidelines and reorganize contracts with restricted admin
  21. 2121 Jan 2026, 00:00 PSTAdd comprehensive README for SettlementEngine Smart Contracts
  22. 2221 Jan 2026, 12:48 PSTAdd retry logic and error handling to full-deploy script
  23. 2321 Jan 2026, 13:29 PSTAdd wallet setup script with retry logic and deployment ID generation

Also available: repository build and repair, spec-to-repo, security, and browser-graded application tasks.

Request coding data

03 / Agent environments

A task to complete.
A result to verify.

Agents need work that spans tools, decisions and time. Train on business and research environments with known outcomes and the trajectories that reach them.

  • 16,000 graded business instances
  • 24,000 computer-use trajectories
  • Starting state, actions and final-state checks
Request a task sample Why partial credit changes the training signal ↗Explore real business environments ↗Explore computer-use trajectories ↗
Recorded task replay

Research reproduction audit

A captured research-agent run that builds and tests an auditor for a physics-informed model reproduction packet.

Complete recorded action sequence. Tool inputs and outputs are summarized; identifying details are removed.

Captured run · de-identified
Task

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 / 06

Inspect the benchmark packet

bash
01Agent intent
Inspect the benchmark packet
02Tool call
bash

Inspect the provided repository and the public benchmark specification.

03Observed 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 result context2 records
surfacestatus
repositoryinspected
reference PDElocated
Checks6 pending
← / → stepspace play / pauseTask
Open the full research-task replay Explore all recorded agent runs ↗

Also from Gerra

A wider field of data.

Browse every collection

Gerra Research

The judgment
behind the data.

What makes a useful training example? What invalidates an evaluation? We write about the decisions, failure modes and evidence inside the work.

Explore the research

Our research also spans markets, compute and physical AI. Read our prediction-market collaboration with Eagle Alpha, published as a guest contribution at Integrity Research.

Work with Gerra

What capability
are you building?

Start with a collection, inspect a sample, or tell us the task your model needs to learn.

Find the right data team@gerra.com