% ============================================================================ % UQO Working Paper Template — Economics / Finance % ============================================================================ \documentclass[12pt,letterpaper]{article} % --- Encoding & Language --- \usepackage[utf8]{inputenc} \usepackage[T1]{fontenc} \DeclareUnicodeCharacter{2212}{--} \DeclareUnicodeCharacter{2192}{$\rightarrow$} \DeclareUnicodeCharacter{2265}{$\geq$} \DeclareUnicodeCharacter{2264}{$\leq$} \DeclareUnicodeCharacter{00D7}{$\times$} \DeclareUnicodeCharacter{2026}{\ldots} \DeclareUnicodeCharacter{2248}{$\approx$} \DeclareUnicodeCharacter{03B2}{$\beta$} \DeclareUnicodeCharacter{03C4}{$\tau$} \DeclareUnicodeCharacter{0394}{$\Delta$} \DeclareUnicodeCharacter{03B1}{$\alpha$} \DeclareUnicodeCharacter{03C1}{$\rho$} \DeclareUnicodeCharacter{03C3}{$\sigma$} \DeclareUnicodeCharacter{0398}{$\Theta$} \DeclareUnicodeCharacter{2208}{$\in$} \DeclareUnicodeCharacter{2227}{$\wedge$} \DeclareUnicodeCharacter{00B2}{$^2$} \DeclareUnicodeCharacter{2713}{yes} \DeclareUnicodeCharacter{2714}{yes} \DeclareUnicodeCharacter{2011}{-} \usepackage[english]{babel} % --- Page Layout --- \usepackage[ letterpaper, top=1in, bottom=1in, left=1in, right=1in ]{geometry} % --- Typography --- \usepackage{mathptmx} % Times-like font (text + math) \usepackage{microtype} % Improved typographic refinements \usepackage{setspace} \onehalfspacing % 1.5 line spacing % --- Math --- \usepackage{amsmath,amssymb,amsthm} % --- Tables --- \usepackage{booktabs} \usepackage{tabularx} \usepackage{multirow} \usepackage{threeparttable} \usepackage{siunitx} \sisetup{ input-symbols = {()}, table-format = -1.3, parse-numbers = false } % --- Figures --- \usepackage{graphicx} \usepackage{subcaption} \usepackage{algorithm} \usepackage{algpseudocode} \usepackage{longtable} \usepackage{array} \usepackage{adjustbox} \usepackage{fvextra} \usepackage{float} % --- Colors & Links --- \usepackage[dvipsnames]{xcolor} \usepackage[ colorlinks = true, linkcolor = NavyBlue, citecolor = NavyBlue, urlcolor = NavyBlue, breaklinks = true ]{hyperref} % --- Bibliography --- \usepackage[authoryear,round,semicolon]{natbib} \bibliographystyle{aer} % --- Captions --- \usepackage[ font = small, labelfont = bf, labelsep = period, justification = justified, singlelinecheck = false ]{caption} % --- Headers & Footers --- \usepackage{fancyhdr} \pagestyle{fancy} \fancyhf{} \renewcommand{\headrulewidth}{0pt} \fancyfoot[C]{\thepage} % --- Section Formatting --- \usepackage{titlesec} \titleformat{\section}{\large\bfseries}{\thesection.}{0.5em}{} \titleformat{\subsection}{\normalsize\bfseries}{\thesubsection.}{0.5em}{} \titleformat{\subsubsection}{\normalsize\itshape}{\thesubsubsection.}{0.5em}{} % --- Appendix Support --- \usepackage[toc,page]{appendix} % --- Misc --- \usepackage{enumitem} \setlist{nosep,leftmargin=*} % ============================================================================ % METADATA — Edit these fields for each working paper % ============================================================================ \newcommand{\WPnumber}{29} \newcommand{\WPtitle}{Protocol-Guided Autonomous Research with Claude Code} \newcommand{\WPsubtitle}{Scientific Verification, Reproducibility, and the Cost of Delegation: A Controlled Study} \newcommand{\WPdate}{October 2026} \newcommand{\WPversion}{1.0} \newcommand{\WPabstract}{% Instruction files that steer coding agents through research tasks range from one sentence to several thousand words of phases, gates and integrity rules, but whether that structure changes what the agent delivers has not been measured with the agent held fixed. We ran \nMainRuns{} live Claude Code runs (\nMainSessions{} sessions; client 2.1.314, model claude-opus-5-5) on \nConfTasks{} bounded scientific tasks in six families, \nRepeats{} fresh sessions per task under each of four regimes: an open instruction; a structured protocol with phases, ledgers, stopping and recovery rules; the protocol plus a controller-enforced executable gate; and the same plus a separate review-and-repair session. Executors ran in bare mode on isolated hosts; deterministic hidden oracles, validated on seeded artefacts, scored every run on a separate host; the design, tasks and analysis were frozen with hashes before the confirmatory runs. Intervention-free verified completion was \pA{} under the open instruction and \pD{} under the full method; the paired task-level difference is \dDA{} (task-cluster bootstrap 95\% interval [\dDAlo, \dDAhi]), the mechanism contrasts are equally null, and critical integrity violations were rare in every arm. The protocol, gate and reviewer bought full compliance with the deliverable contract at two to three times the inference cost per run. Every non-success of the frozen scoring was, on inspection, a defect of our oracles rather than of the executors (an undefined feasibility label, a generator artefact that twelve sessions correctly excluded, a rounding tolerance, a regular-expression false positive); under a corrected rubric every arm completes every task, and the null is unchanged. Arm D's review-repair loop, though available, was never triggered. Interruption-and-resumption, stale-input, duplicate-job and missing-fixture experiments, a length-matched control and three held-out end-to-end mini-projects complete the evidence, together with an instrumented account of how this paper was itself produced under a mandate of the same kind. For bounded tasks with known answers, the marginal value of protocol structure lies in auditability, not correctness, and the scarce resource is a trustworthy oracle.% }\newcommand{\WPkeywords}{autonomous research agents; Claude Code; research protocols; executable verification; reproducibility; agent evaluation; delegation} \newcommand{\WPjel}{C18, C63, C88, O33} % --- Author --- \newcommand{\WPauthor}{Simon-Pierre Boucher} \newcommand{\WPaffiliation}{% Département des sciences administratives\\ Université du Québec en Outaouais% } \newcommand{\WPemail}{simon-pierre.boucher@uqo.ca} \newcommand{\WPaddress}{% Gatineau -- Pavillon Alexandre-Taché\\ 283, boulevard Alexandre-Taché\\ Gatineau, Québec, Canada J9A 1L8% } % ============================================================================ % DOCUMENT % ============================================================================ \input{tables/numbers} \providecommand{\nRepeats}{3} \begin{document} % --- Title Page --- \input{sections/titlepage} % --- Main Body --- \input{sections/introduction} \input{sections/literature} \input{sections/framework} \input{sections/corpus} \input{sections/implementation} \input{sections/design} \input{sections/evaluation} \input{sections/results} \input{sections/recovery} \input{sections/transfer} \input{sections/reflexive} \input{sections/limitations} \input{sections/conclusion} \input{sections/disclosure} % --- References --- \newpage \bibliography{references} % --- Appendix --- \input{appendices/appendix} \end{document}