Editorial · The Post-Project World
27 July 2026 · 00:20:11
The critical repairs had passed. The book had not. In the third and final episode documenting the reconstruction of a seven hundred and eighty-eight page FX options reference, Luigi Pascal Rondanini follows the work beyond the first approved gate: hundreds of remaining findings, an independent review allowed to fail after weeks of successful work, damaged Git worktrees, crashed sessions, a migration from Kimi to Cursor and Grok, missing mathematics inside a professional-looking PDF, and a technically correct implementation rejected because one warning statistic could not be reproduced. The direct model and tooling cost was roughly seven hundred US dollars. A conservative conventional baseline for technical editing, proofreading and formatting alone would have been roughly fourteen and a half to twenty-two thousand dollars, before specialist quantitative review, fact-checking and project management. The lesson is not that AI made technical publishing cheap. It made verification affordable enough to repeat. In this episode: Why the remaining work stopped being editorial and became operational A gate that could still say no: the package that failed its independent review, and what happened to the failure When the winning model became too slow and too expensive: moving from conversation-centred to repository-centred work The failures nobody writes papers about: locked directories, orphaned worktrees, commands run from the wrong folder Newest is not the same as approved: why provenance outranks freshness Twelve defects in a finished-looking book, including missing Garman-Kohlhagen mathematics that compiled cleanly The seventy-six to seventy-six story: a correct repair rejected over one unreproducible number The economics: what seven hundred dollars bought, what it did not, and why "check it again" became a realistic instruction Read the full field note: https://orbaos.com/field-notes/the-last-mile-was-the-whole-project Episode One: https://orbaos.com/field-notes/the-agent-is-not-the-control-plane Episode Two: https://orbaos.com/field-notes/from-audit-to-evidence The book, FX Options (Pricing, Hedging, Exotics, Risk and Institutional Practice): https://rondanini.com/fx-options The Post-Project World is hosted by Luigi Pascal Rondanini, founder of Rondanini Publishing and OrbaOS, with thirty-eight years in FX markets. The series explores how AI coordination, governance and persistent infrastructure are replacing the traditional project as the unit of work. Keywords: AI governance, agentic AI, AI agents, AI economics, cost of AI, AI verification, AI audit, deterministic tools, multi-agent systems, control plane, Git workflow, Cursor, Grok, Kimi, independent review, LLM evaluation, AI orchestration, OrbaOS, post-project world, FX options, technical publishing, AI in finance, machine-mediated work.
24 July 2026 · 00:13:05
An audit identifies defects. It does not prove that the repairs are correct. In Episode One I set three AI systems against my forthcoming FX options book and let the strongest one audit it. That audit produced one thousa…
View episode
21 July 2026 · 00:17:30
I gave three advanced AI systems the same task: audit the formulas, calculations and worked examples in my forthcoming professional book on FX options. Claude, Codex and Kimi. Seven hundred and sixty-four typeset pages, …
View episode
16 July 2026 · 00:14:40
Every regulated institution has rules. Most cannot prove that their systems enforce those rules consistently. A payment threshold approved by a committee in 2019 is implemented in Python by one team, Java by another, and…
View episode