1# WP42 — The Autonomous White-Collar Worker23Measuring the economic frontier of Claude Code, computer-using agents and autonomous knowledge work. UQO Working Paper Series No. 424(Department of Administrative Sciences, Université du Québec en Outaouais). Author: Simon-Pierre Boucher. Built 2026-11-02 →.56**Research question.** What share of economically valuable white-collar deliverables can an autonomous Claude Code worker complete to a7professionally acceptable standard, at what cost, with what reliability and supervision needs, and how much of it is explained by the8level of computer access (file-only → shell → macOS layer → browser/office/portal → self-repair)?910## Headline results (pilot, 150 runs, claude-opus-5-5, 2026-11-03)11Scenario-weighted acceptance (deterministic completion ∧ no critical violation ∧ blind quality ≥ 70): A0 0.00 · A1 0.87 · A2 0.90 · A3 0.90 · A4 0.83.12Shell premium A1−A0 = +0.87 [0.67, 1.00], p = 0.004. Preregistered primary contrast A3−A1 = +0.03 [0.00, 0.10], p = 1.0 (precise null, ceiling).13Cost per accepted deliverable 0.87–1.11 USD (Sonnet 5.5: 0.40 USD at 0.85; Fable 5.1: 3.30 USD at 0.85). Exploratory recovery experiment:14after an induced interruption at 120 s, one bounded recovery session raises acceptance from 0.05 to 0.85 (20/20 self-check passes).15Human arms H0/H1 designed, not executed. Total API spend 261.89 USD. Monograph: `release/wp42_uqo.pdf` (189 pp.); journal version: `release/wp42_uqo_journal_version.pdf` (41 pp.).1617## Layout18| Path | Content |19|---|---|20| `CLAUDE.md` | the research mandate (verbatim) |21| `RESEARCH_PLAN.md`, `PROJECT_STATUS.md` | operational design, phase status |22| `protocol/` | preregistration, amendments, ethics (human arms, not executed), power analysis |23| `literature/` | 110 verified references (`references.bib`), structured review, verification log |24| `benchmark/` | authoring spec, taxonomy, public scenario bundles (`scenarios/<TASK_ID>/`), shared evaluator helpers, public self-check |25| `src/wp42/` | controller: store, launcher, arms, bundle, portal, scheduler, evaluate, judge, metrics, export, dashboard, watchdog, CLI |26| `data/raw/` | SQLite store + immutable raw streams per run (git-ignored) · `data/validated/` Parquet/CSV datasets |27| `analysis/` | econometrics, robustness, figures, tables (all generated by scripts) |28| `manuscript/` | LaTeX monograph · `journal_version/` short version |29| `dashboard/` | private monitoring page (`python -m wp42.dashboard --serve 8642`) |30| `tests/` | unit tests (`.venv/bin/python -m pytest -q tests`) |3132Hidden oracles (ground truth, deterministic evaluators, rubrics, validation reports) live **outside** the repository in `~/wp42-oracles/`33and are never copied to executor nodes. The API key lives in `~/Desktop/Cluster/secrets/wp42/.env` (controller) and `<node>:~/wp42/.env`.3435## Running the laboratory36```bash37cd wp42 && export PYTHONPATH=src38.venv/bin/python -m wp42.cli sync-tools # portal package → nodes39.venv/bin/python -m wp42.cli assign --tasks ACC-01 FIN-01 ... --arms A0 A1 A2 A3 A4 --reps 2 --seed 4242 --stage pilot40.venv/bin/python -m wp42.cli run --parallel 12 # executes, collects, evaluates, judges (envelope enforced)41.venv/bin/python -m wp42.cli status | ledger | export --tag pilot42.venv/bin/python -m wp42.cli kill-all # emergency shutdown43.venv/bin/python -m wp42.dashboard --serve 864244```45Workers run as `claude --bare -p` on M3U96a/M3U96b/M4M64a/M4M64b (Claude Code 2.1.314) in `~/wp42/runs/<run_id>/`, with the arm's tool46flags, 4 USD / 1,800 s / 150-turn caps, no Agent tool and a depth guard. See `docs/PHASE_0_DISCOVERY.md` for the measured environment.47