Skip to content

RECOMMENDATIONS.md

Latest commit

 

History

History
61 lines (49 loc) · 2.56 KB

RECOMMENDATIONS.md

File metadata and controls

61 lines (49 loc) · 2.56 KB
Raw

Recommendations

What to work on next, in priority order. Claude maintains this file and rewrites it when something new occurs to it. Last updated 2026-09-03, after the whole-book adversarial review was applied: 252 findings, 213 applied (one agent per chapter, each finding reproduced before its fix), 1 declined, 38 author-level calls queued in adversarial_undecided.md. Its record is archive/~adversarial_review.md; make verify green. Before that, the whole-book exercise pass was applied: 45 chapters, ~320 exercises performed cold, 75 findings in 30 chapters, 73 applied and 2 declined. Its record is archive/~exercise_review.md. It followed the whole-book correctness sweep: all 47 chapters checked, 56 false claims fixed (commits 7583395d through e2e5ba75, plus e4a4e6cd applying the review file), make verify green. The sweep's record and its reasoning live in archive/~correctness_review.md.

[[Note: find the GoF CD and upload it so Claude can verify against it]]

Chapter Edit Steps

  1. make rewrite on the chapter
  2. Hand-edit the chapter
  3. Run /bruce-edit-capture on the chapter

The correctness review that used to be step 4 is done book-wide and is now a gate for the part of it a machine can decide (make self-reference). What the gate cannot decide still needs a reader, so keep a correctness pass in any chapter that gets substantial new prose.

The recommendations

  1. Write the chapter 01 AI anecdote. The last entry in adversarial_undecided.md: one concrete case where book-level Python judgment steered an AI to a better solution, closing the AI section's now-final sentence with evidence. Only you can write it. Archive the file when it lands. The other 37 queue entries were approved and performed 2026-09-03; make verify green.

  2. Defer the whole-book readability sweep to last. Every rewrite and review round adds new prose, and new prose is where AI tells appear. One sweep after chapter-level work settles mops it all up; running it earlier audits text that will change.

Open items the sweep left for you

Two attributions to GoF Design Patterns were rewritten in e4a4e6cd from what looked like Java rather than GoF, on reasoning rather than on the book itself. Chapter 30's Observer no longer carries a changed flag, and chapter 29 now names the pluggable and two-way adapters. Both are worth a check against your copy.

Suggested sequence

  1. Item 1 (the chapter 01 anecdote, the queue's last entry).
  2. readability sweep
  3. One green make sweep.
  4. 'Chapter Edit Steps' through the remaining chapters.