Acknowledgements and References
Sources and intellectual debts
The individual adoption stages in Part 3 are adapted from Steve Yegge's "Evolution of the Programmer" model and Daniel Jones's "11 Stages of Agentic Coding."
The Theory-of-Constraints framing in Part 3 applies Eliyahu Goldratt's principle that optimising any step other than the binding constraint reduces throughput. The observation that AI acceleration amplifies downstream dysfunction is a direct application of this principle.
The adoption sequence in Part 4 draws on Bryan Finster's work on delivery improvement sequencing.
The remediation sequence in Part 4 is original to this framework but is consistent with Jason Gorman's observation (Codemanship, "Stuck In A Doom Loop? Drop A Gear") that teams trapped in AI-amplified chaos benefit from reducing agent autonomy before increasing it. Gorman's "AI-Ready Software Developer" series more broadly is a practitioner-layer treatment that this framework's organisational-layer positioning complements rather than duplicates.
The spec-driven development position in Part 5 draws on emerging industry practice documented by GitHub (Spec Kit), Thoughtworks, and AWS (Kiro), reframed within an XP and Continuous Discovery context. The BDD methodology and Example Mapping practice draw on the foundational work of Dan North, Gojko Adzic, and Matt Wynne.
The spec evolution practices in Part 5 draw on analysis by Lilia Abdulina and Vitaly Sharovatov (BeyondQuality) regarding the severed feedback loop in AI-accelerated development, and on Bryan Finster's persistent task memory mechanisms in his agentic-dev-team implementation.
The mutation-testing framing in the "verifying the verifier" section of Part 5 draws on the extensive academic literature beginning with DeMillo, Lipton, and Sayward (1978) and operationalised in contemporary tools; its application as a routine quality gate for AI-authored tests is a specific adaptation for the agentic workflow context. The eval-driven development position extends this into a broader verification discipline informed by the TDAD (Test-Driven Agent Development) paper.
The management philosophy discussion in Part 6 draws on Douglas McGregor's Theory X and Theory Y.
The adoption principles in Part 7 are derived from the reasoning patterns in the Future Friendly manifesto by Brad Frost, Luke Wroblewski, Scott Jenson, and others, reframed for technology adoption contexts.
The team design questions in Part 8 draw on Matthew Skelton and Manuel Pais's Team Topologies framework, including the enabling-team framing applied to the quality coach role. The cognitive load measurement questions reference work by Skelton, Pais, Weis, and Morgadas.
The underwriting-pack artefact specification in Part 8 develops a concept originated by Stuart Winter-Tear in prior conversations on AI governance escalation.
The analysis of tooling commoditisation in Parts 2, 5, and 8 draws on the published architecture of re:cinq's Lore - specifically its MCP-based agent coordination, persistent memory layer, and per-repository governance primitives. References to Lore describe its architectural patterns and do not constitute endorsement.
Steve Yegge's Gas Town and Beads projects informed the multi-agent coordination and persistent task memory discussions in Parts 3 and 5.
Thanks
This framework was shaped by years of working alongside brilliant practitioners at Armakuni. For the conversations, the challenges, and the shared commitment to returning joy and creativity to the world of software: Zenon Hannick, Andrea Laforgia, Tom Oram, Billie Thompson, Ash Sehra, Victoria Tuck, Faye Benfield, Mark Bradley, Ricardo Vargas, Darren Jones, Shane Harger, Ben Dodd, Tim Savage, and all the other brilliant minds at Armakuni, present and past.
Licensed under CC BY-NC-SA 4.0.