TT Claude Memory

feedback_recency_vs_maturity

Properties1
descriptionRich iterated three times on this (17-18 April 2026) to refine the shopping heuristic. Must be available NOW (hard filter). Review data is PREFERRED but not required — if materials look right and publisher is excellent, include with the caveat "No reviews but the materials sound perfect and the publisher is excellent." Don't overcorrect on recency or review-strictness.

3 min read

The refined heuristic (after three iterations 17-18 April 2026):

When recommending books for a live purchase decision:

  1. Available NOW — hard filter. Must be shipping today (physical) OR accessible today via subscription (e.g. O’Reilly Learning early-access is “available now” for subscribers). Pre-order, forthcoming, MEAP-without-digital-access are OUT.

  2. Reputable publisher — hard filter. O’Reilly, Manning, Packt, Addison-Wesley, Pearson, Cambridge University Press, Oxford University Press, MIT Press, Morgan Kaufmann, Apress, Springer, Morgan & Claypool, No Starch Press, IOS Press (academic). Not self-published unless clearly excellent provenance.

  3. Materials genuinely match the research question — hard filter. Table of contents or description must show direct relevance. Don’t pad with adjacent books.

  4. Reviews — PREFERRED, not required. Where reviews exist (Amazon / Goodreads / professional), include review signal. Where reviews are thin (especially for books published in the last few weeks / months), the honest caveat is:

    “No reviews yet, but the materials sound perfect and the publisher is excellent.”

    This tells Rich what the uncertainty is without dropping a genuinely useful recent book.

  5. Sweet spot still 2024-early 2025 where possible for maximum maturity + recency + reviews, but don’t exclude fresh 2025-2026 titles just because reviews haven’t accumulated yet.

Three iterations’ mistake pattern:

  • v0.1: defaulted to “2022-2024” floor → missed the cutting edge
  • v0.2: corrected to “2025-2026 priority” → included forthcoming + early-access + no-review
  • v0.3: overcorrected to “2024-early 2025 + 10+ reviews required” → dropped good recent books
  • v0.4 (this memory): balance — available + reputable publisher + materials-match; reviews preferred with honest caveat when absent

The underlying meta-lesson: don’t overcorrect on critique. When Rich says “X is wrong,” the answer is usually calibration back to the right balance, not swinging to the opposite extreme. Each correction should produce an incremental move, not a pendulum swing.

How to apply:

  1. Verify availability today (Amazon UK in-stock OR publisher direct OR subscription access). If unverifiable or future-shipping, exclude.
  2. Verify publisher reputation. If self-published or unknown press, exclude unless clearly excellent (e.g. author is known domain expert).
  3. Verify materials (TOC / description / sample chapter) match the research question. If uncertain, exclude.
  4. Add review data where it exists; add honest caveat where it doesn’t. Do NOT drop otherwise-good books for lack of reviews alone.
  5. State publication month explicitly so Rich can judge maturity.
  6. For O’Reilly Learning subscription-accessible books, flag that separately — these are “available now” at zero marginal cost if Rich has the subscription.

Why this matters. Rich is making real purchase decisions in real time. Over-filtering costs good recommendations; under-filtering costs money on weak books. The caveat pattern (“No reviews, but materials + publisher signal strong”) gives Rich the information to decide, rather than deciding for him.

Related feedback memories:

  • feedback_premature_option_siloing.md
  • feedback_lau_pa_sat_self_marking.md
  • feedback_sequential_vs_parallel_revisions.md

Graph View

Backlinks