Primary sources for public edition 0.2, 5 October 2026. Preprints, research notes and technical essays retain their source status. A citation is not a claim of independent replication. The review is focused, not exhaustive.
Context and cost
M1
Martin Monperrus, Dark Tokens, June 30, 2026 web essay.
M2
Martin Monperrus, Measuring AI Tokens Locally, June 30, 2026, technical essay.
R1
Thibaud Gloaguen, Niels Mündler, Mark Müller, Veselin Raychev, Martin Vechev, Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?, February 12, 2026 preprint v1.
R2
Jai Lal Lulla, Seyedmoein Mohsenimofidi, Matthias Galster, Jie M. Zhang, Sebastian Baltes, Christoph Treude, On the Impact of AGENTS.md Files on the Efficiency of AI Coding Agents, January 28, 2026 preprint v1.
R3
Prakhar Khatri, Do Context Files Help Coding Agents? A Two-Agent Ablation Study on Real Repositories, July 28, 2026 preprint v1.
R4
Elias Lumer, Faheem Nizar, Akshaya Jangiti, Kevin Frank, Anmol Gulati, Mandar Phadate, Vamse Kumar Subbiah, Don't Break the Cache: An Evaluation of Prompt Caching for Long-Horizon Agentic Tasks, January 9, 2026 preprint v1.
R5
Sayash Kapoor, Benedikt Stroebl, Zachary S. Siegel, Nitya Nadgir, Arvind Narayanan, AI Agents That Matter, July 1, 2024 preprint v1.
R10
Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, Percy Liang, Lost in the Middle: How Language Models Use Long Contexts, TACL 12 (2024), 157–173, DOI 10.1162/tacl_a_00638.
Acceptance and delivery
M6
Martin Monperrus, The Epistemic Agent—An Agent That Must Say Why, September 4, 2026, prototype essay.
M7
Martin Monperrus, Bootstrapping Coding Agents: The Specification Is the Program, arXiv v1 March 18, 2026; publication record names IEEE Software 2026, DOI 10.1109/MS.2026.3687051.
R6
Parker Whitfill, Cheryl Wu, Joel Becker, Nate Rush, Many SWE-bench-Passing PRs Would Not Be Merged into Main, March 10, 2026 METR research note, explicitly less reviewed than research posts.
R7
Joel Becker, Nate Rush, Beth Barnes, David Rein, Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity, July 10, 2025.
R8
METR, We are Changing our Developer Productivity Experiment Design, February 24, 2026 research update.
R11
METR, Measuring AI Ability to Complete Long Software Tasks, March 19, 2025 original post with mutable current dashboard.
C1
Yifan Shi, Wei Zhang and Tianyi Cui. A Programming Paradigm for Spatiotemporal Composability. arXiv:2608.25512v1, 26 August 2026. Cordis lifecycle and system-boundary discussion, §§5.3 and 6.1–6.6. Preprint; no independent proof or implementation validation is claimed.
Personal computing
U1
Dennis M. Ritchie and Ken Thompson. The UNIX Time-Sharing System. Expanded 1978 Bell System Technical Journal version of the 1974 CACM paper; §§5.2 and 6.
U2
Alan Kay and Adele Goldberg. Personal Dynamic Media. Computer 10(3), 31–41, March 1977. Original essay reprinted in The New Media Reader, beginning on p. 393; the preceding editorial introduction is not attributed to the authors.
U3
Andrew J. Ko and colleagues. The State of the Art in End-User Software Engineering. Author-hosted prepublication manuscript, marked ACM Computing Surveys, to appear; introduction and §§2.1–2.3. The filename is not used to infer publication year.
U4
Eelco Dolstra and Andres Löh. NixOS: A Purely Functional Linux Distribution. Submitted manuscript dated 2 April 2008; abstract and §§1–2.
U5
Theodore R. Sumers, Shunyu Yao, Karthik Narasimhan and Thomas L. Griffiths. Cognitive Architectures for Language Agents. arXiv:2309.02427v3, 15 March 2024; §§4.1–4.5.
U6
Kai Mei and colleagues. AIOS: LLM Agent Operating System. arXiv:2403.16971v5, 12 August 2025; COLM 2025. Architecture, §3.1.
U7
systemd project. System and Service Manager. Project documentation, accessed 3 October 2026. Linux-specific service and process lifecycle facilities; detailed semantics require the installed version’s documentation.
U8
Andrew Gallant and the ripgrep contributors. ripgrep README. Project documentation, accessed 3 October 2026. Recursive search and default ignore, hidden-file and binary filtering; no speed comparison is used here.
U9
Git project. Git documentation. Project reference, accessed 3 October 2026. Version-control primitives do not supply authorization or rollback for remote effects.
Provenance and earlier assistance
P1
W3C. PROV-DM: The PROV Data Model. W3C Recommendation, 30 April 2013. Entities, activities, agents and derivation supply established provenance foundations.
P2
Santiago Torres-Arias and colleagues. in-toto: Providing farm-to-table guarantees for bits and bytes. USENIX Security 2019. Software supply-chain integrity; not a substitute for a task's application-level acceptance predicate.
U10
Eric Horvitz. Principles of Mixed-Initiative User Interfaces. CHI 1999. LookOut illustrates automated assistance combined with user control, uncertainty and refinement.
U11
Yezdi Lashkari, Max Metral and Pattie Maes. Collaborative Interface Agents. Author-hosted research account of collaborating learning agents for electronic mail. The personal-computing comparison credits this earlier assistance work; no priority claim or comparative performance result is inferred.
Cited researchers have not thereby reviewed or endorsed these manuscripts. Martin Monperrus did not personally review or endorse them.