Lab · Built in the open

Self-directed research,with the receipts

Work I do on my own time, held to the same standard as the paid kind: named baselines, stated uncertainty, and the dead ends published next to the wins.

Beane · Deep research harness

Beane: #1 globally on DeepResearch Bench II

I built Beane, a deep research harness that ranks #1 globally on DeepResearch Bench II and #2 on DeepResearch Bench, ahead of every OpenAI, Google and Perplexity entry on both boards. It reads a hundred sources, checks every quote against its page in code, and writes a cited report.

DeepResearch Bench I and II · GPT-5.5 judged · LangGraph · Private repo, open source intended

DRB II · 60.52 · #1

DRB I · 56.14 · #2

+17.4 over OpenAI o3 Deep Research

See the rankings →︎

Both rankings are on the official DeepResearch Bench leaderboard ↗︎.

Published study · Currency forecasting

Which currency questions are actually answerable?

Ten years of daily prices for the euro, yen and rupee against the dollar. Three questions, each graded against a stated simple guess. Two were dead ends. The third fixes a number companies set every month.

Forecasting · Public data only · Fully reproducible repo

Q1 · No better than guessing

Q2 · 51-57% less error

Q3 · 16-28% less error

Read the study →

Build log

What landed, and when

2026-10-28

New scores for Beane listed on the official DeepResearch Bench leaderboard: #1 on DeepResearch Bench II (60.52) and #2 on DeepResearch Bench (56.14).

2026-10-11

Beane listed on the official DeepResearch Bench leaderboard: #1 on DeepResearch Bench II (56.70) and #4 on DeepResearch Bench (55.37), where the top four are within 0.62 points of each other.

2026-10-06

Beane ran both DeepResearch Bench benchmarks in full and scored them with the official evaluation code: 55.33 on the original and 57.40 on DeepResearch Bench II. Submitted for the official listing. Written up in two parts: the harness, and the benchmarks.

2026-09-18

Beane reached a 63.0 mean on DeepResearch Bench II, up 6.7 on the same tasks - recall carried it, up nine and a half points. Second on the board, 1.4 behind the leader and clear of every deep-research product a frontier lab ships. Pre-submission; every label is in the repo ledger.

2026-08-06

Currency study published: three questions, ten years of data, every answer graded against a stated simple guess. Two dead ends documented alongside the one usable result.

2026-07-31

Researcher agent committed. Plan: LangGraph agent, trace recording from day one, replay player, then live mode behind hard caps.

Want to know when something new lands here? Email [email protected] with the subject "lab".