References and further reading paths

All sources were consulted on November 1, 2026. Publication dates and data-collection dates are distinguished where relevant. Documentation pages without an explicit date are identified as current documentation. The identifiers [F01] to [F39] are links to the original sources.

17.1. Premise and attribution

• F01 Butler, Bart — “Proton’s AI manifesto for software engineering”. Proton, November 1, 2026. Public source of the manifesto provided by the user. A position and organizational guidance text; not a productivity study.
• F02 Proton — “Bart Butler”. Author's official page, current. Verification of the attribution and the CTO title.

17.2. Changes in work, productivity, and organization

• F03 DeBellis, Storer, Harvey et al. — “DORA 2025 State of AI-assisted Software Development Report”. DORA/Google, 2025. Qualitative research and survey on AI and software development.
• F04 Storer and DeBellis — “Introducing the DORA AI Capabilities Model”. Google Cloud, September 23, 2025. The researchers' presentation of the organizational capabilities associated with outcomes.
• F05 Baolin and Harvey — “Balancing AI tensions: Moving from AI adoption to effective SDLC use”. DORA, March 10, 2026. Thematic analysis of engineers' responses; useful for investigating shifts in effort.
• F06 Lopopolo, Ryan — “Harness engineering: leveraging Codex in an agent-first world”. OpenAI, February 11, 2026. Vendor engineering account of agent-driven execution and environment preparation.
• F07 Rajasekaran, Prithvi — “Harness design for long-running application development”. Anthropic, March 24, 2026. Development experience with planning, generation, and evaluation.
• F08 Anthropic Economics — “Agentic coding and persistent returns to expertise”. June 16, 2026. Observational study of sessions; useful for understanding domain expertise and division of labor.
• F09 Cui, Demirer, Jaffe, Musolff, Peng, and Salz — “The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments with Software Developers”. June 2025. Field experiments presented by the authors at Microsoft Research.
• F10 Becker et al. — “We are Changing our Developer Productivity Experiment Design”. METR, February 24, 2026. Methodological update on selection, measurement, and simultaneous work with agents.
• F11 Forsgren et al. — “The SPACE of Developer Productivity: There’s more to it than you think”. ACM Queue, February 2021. Multidimensional productivity framework; authors' page at Microsoft Research.
• F12 Noda and Tacho — “Applying the DevEx framework”. March 1, 2024. The authors' guidance on application. Complements the original article “DevEx: What Actually Drives Productivity”, by Noda, Storey, Forsgren, and Greiler, published May 3, 2023.
• F13 Houck et al. — “The SPACE of AI: Real-World Lessons on AI’s Impact on Developers”. Microsoft Research, August 2025. Mixed-methods study on experience, productivity, and team adoption.
• F38 Harvey, Nathen — “DORA’s software delivery performance metrics”. DORA, current documentation. Official definitions of the five metrics and guidance on contextualized use.
• F39 Anthropic Engineering — “Redesigning source control at Anthropic”. October 13, 2026, updated October 15, 2026. Account of infrastructure and version control under an exceptional AI-driven workload.

17.3. Training, fundamentals, and judgment

• F14 ACM, IEEE Computer Society, and AAAI — “Generative AI and the Curriculum”, CS2023 section. January 2024. Curricular guidance on changing activities and enduring fundamentals.
• F15 Anthropic — “How AI assistance impacts the formation of coding skills”. January 29, 2026. Write-up of the Shen and Tamkin experiment on immediate learning with AI assistance.
• F16 Lee et al. — “The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers”. CHI, April 2025. Correlational survey on critical effort and confidence.
• F35 Winters, Titus — “What Is Software Engineering?”. In Software Engineering at Google, 2020. Reference on time, scale, and engineering trade-offs.
• F36 Chen and Barolak — “Knowledge Sharing”. In Software Engineering at Google, 2020. Reference on collective learning, documentation, and distribution of expertise.
• F37 Rodden, Hutchinson, and Fu — “Measuring the User Experience on a Large Scale: User-Centered Metrics for Web Applications”. CHI, 2010. Presentation of the HEART framework and user-oriented metrics.

17.4. Context, tools, agents, and evaluation

• F17 Anthropic — “Effective context engineering for AI agents”. September 29, 2025. Practices for organizing and maintaining context during execution.
• F18 Anthropic — “Writing effective tools for agents — with agents”. September 11, 2025. Design and evaluation of tool interfaces.
• F19 Anthropic — “Equipping agents for the real world with Agent Skills”. October 16, 2025, updated December 18, 2025. Organization of reusable instructions, resources, and scripts.
• F20 Anthropic — “Demystifying evals for AI agents”. January 9, 2026. Methodology for evaluating tasks, runs, trajectories, and outcomes.
• F21 Kim et al. — “Towards a Science of Scaling Agent Systems”. arXiv, version 3, April 8, 2026. Experimental preprint on coordination and alignment between architecture and task structure. Citing this specific version avoids mixing numbers from earlier revisions.
• F22 Carlini, Nicholas — “Building a C compiler with a team of parallel Claudes”. Anthropic, February 5, 2026. Capability experiment and account of testing, environment, and division of labor.
• F27 Delimarsky, Den — “Spec-driven development with AI: Get started with a new open source toolkit”. GitHub, September 2, 2025. Methodological proposal and presentation of Spec Kit.

17.5. Security and operational limits

• F23 Anthropic — “How we contain Claude across products”. May 25, 2026. Security account of isolation, access limits, and containment.
• F24 OWASP GenAI Security Project — “OWASP Top 10 for Agentic Applications for 2026”. December 9, 2025. Expert framework. Both the overview page and the downloadable document were consulted.
• F25 Model Context Protocol — “Security Best Practices”. Documentation for version 2026-07-28. Technical guidance on identity, authorization, and integration risks.
• F26 GitHub Docs — “Customizing or disabling the firewall for GitHub Copilot”. Current documentation. Concrete example of the limits and coverage of a network egress control.

17.6. Data, reliability, and operations

• F28 PostgreSQL — “Transaction Isolation”. Documentation for version 16, section 13.2. Isolation semantics and concurrent operations used in the technical example. The reference is explicitly to this version.
• F29 Featonby, Malcolm — “Making retries safe with idempotent APIs”. Amazon Builders' Library, no explicit date. Experience designing retry contracts and identifying intent.
• F30 AWS Prescriptive Guidance — “Transactional outbox pattern”. Current documentation. Boundary between database writes and publication, with considerations about duplicates.
• F31 OpenTelemetry — “Signals”. Current documentation. Reference for traces, metrics, logs, and observability signals.
• F32 Ewaschuk, Rob — “Monitoring Distributed Systems”. In Site Reliability Engineering, Google, 2016. Fundamentals of monitoring and signal interpretation.
• F33 Google — “Canarying Releases”. In The Site Reliability Workbook, 2018. Release automation and gradual rollout with evaluation.
• F34 Google — “Eliminating Toil”. In The Site Reliability Workbook, 2018. Identifying and reducing repetitive operational work.

17.7. Suggested reading paths

To change your own practice: manifesto F01, context F17, specification F27, and evaluation F20. Pair the reading with a real delivery.

To expand the team's capacity: DORA F03 F04, metrics F38, DevEx F12, and knowledge sharing F36. Choose a bottleneck and track its effect.

To exercise technical judgment: curriculum F14, skill formation F15, and sources on fundamentals relevant to your system, such as F28 F29 F32. Solve variants and investigate failures.

To increase autonomy with control: agent security F23 F24, integration F25, and coordination architecture F21. Assess the concrete environment before expanding permissions.