Notes from the work
Written from things actually built and decisions actually made.
Reports and points of view
2 pieces
- Code Is Cheap, Trust Isn’tState of Software Engineering 2026: what changed when AI started writing software, what the evidence says, and what leaders should build next.
- Where the Gain GoesA free 80-page research report on why AI productivity gains show up per developer and thin out before they reach production.
On this site
2 pieces
- 002The AI Value SystemWhy machine capability is not enterprise value. A short video on Business AI, Human AI, and deciding which work should exist at all.Video · 4½ min
- 001The AI productivity gain is real. It is not reaching main.Why the AI productivity studies disagree, where the difference goes, and what the report does to make its own claims checkable.
Published elsewhere
5 pieces
- Verification debt, and why Agent Karma writes no codeAI writes more of the code each year; the share anyone verifies does not keep pace. Framed through dharma and karma, and the tool built for it.
- From Automation to Autonomy: What Agentic AI Means for Enterprise LeadersTwenty minutes with Ankit Bose on Agents 3.0: agent ecosystems that adapt rather than automate, and the transparency, governance and human oversight they demand.
- What really stops enterprises from moving AI into productionWhy data readiness, process clarity, and organisational alignment matter more than tools, and where governance stays the weak link.
- My Ex-love Python!Balancing sustainability against generative AI, and Spring AI as a Java-friendly alternative to the Python default.
- People and AI: the good, bad and uglyWhere AI adoption goes wrong when governance is missing, and the GRAM test: Governance, Right skills, Adaptive design, Meaningful data.