feat(writings): Fourteen Hours with Opus 4.7 — field report on first production session#103
Merged
Merged
Conversation
A field report on my first real production session with Opus 4.7: fourteen hours from discovery of a schema limitation in the oddkit challenge refactor to production deployment, across five pull requests, with an hour and forty minutes of broken production in the middle. Core argument: the producing side of AI-augmented work got cheaper and the reviewing side got more expensive — the same asymmetric cost shift ODD has been built around since day one. Opus 4.7 is a SOTA iteration, not a new model. The capability moved, the structure didn't, and the discipline travels because there is nothing new under the sun. Notable: this draft is the first essay committed using the canonical frontmatter schema per canon/meta/frontmatter-schema (simple identifiers unquoted, booleans/integers/dates as native YAML types, inline tag array). Existing writings/ articles use a quote-everything corpus convention that disagrees with the schema. If the renderer chokes on this essay's frontmatter, that's signal the schema needs to adapt to match corpus; if it renders cleanly, that's signal the corpus should be migrated toward the schema over time. Derives from: - odd/ledger/journal/2026-04-17-pr100-final-resolution.md (not yet committed; lives on klappy/oddkit chore/handoff-journal-pr100 branch) - odd/ledger/journal/2026-04-17-pr100-rage-quit-handoff.md (same branch) Flagged for review: - Schema vs corpus frontmatter convention gap (see body text) - Summary/TL;DR heading pattern for public essays — not in current writing canon but requested by author; may need canon update to codify
…discernment A second-read observation: SOTA models are in an uncomfortable middle state where asking questions feels pedantic and not asking feels presumptuous. The real shape is the boss-presentation: detect the conflict, propose a resolution with rationale, surface the decision needed. The next essay will unpack that shift from pedantic vs overconfident to present or don't come in. Also names the training-signal corollary: operators who push back on pedantic questions without rewarding well-formed presentations train models to skip the presentation layer entirely. The bosses make the bosses.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
New field report: Fourteen Hours with Opus 4.7
A field report on my first real production session with Opus 4.7 — the oddkit E0008 challenge refactor that shipped across PRs klappy/oddkit#100–#104 on 2026-04-17.
What the essay argues
Source material
odd/ledger/journal/2026-04-17-pr100-final-resolution.md— end-to-end DOLCHE journal (currently onklappy/oddkitchore/handoff-journal-pr100branch, not yet merged)odd/ledger/journal/2026-04-17-pr100-rage-quit-handoff.md— same branchGovernance compliance
canon/meta/frontmatter-schema(all 8 universal required fields + all essay-required fields, native YAML types)canon/meta/writing-canon(title + compressed blockquote + Summary section with descriptive subtitle + header scan passes)canon/constraints/ai-voice-cliches: negation-parallelism cluster from earlier drafts removed; em dash density reducedTwo governance gaps flagged for follow-up
Schema vs corpus disagreement.
canon/meta/frontmatter-schemasays simple identifiers unquoted, booleans/integers/dates as native types. Every existing file inwritings/quotes everything and uses block-list tags. Schema doc literally says "when this document and another disagree, this document wins." This essay follows the schema. If the renderer chokes, that's signal the schema needs to adapt to match corpus. If it renders cleanly, the corpus should migrate toward the schema over time.Summary/TL;DRheading pattern. Writing canon currently specifies## Summary — [descriptive subtitle]as the stable extraction pattern. For public essays the author requested## Summary/TL;DR — [descriptive subtitle]to serve both the oddkit extraction key role and the technical-reader TL;DR convention simultaneously. If this merges, consider a small addition to the writing canon codifying the public-essay variant.Preview URL
Will appear in check comments once the Cloudflare Pages preview builds. Please test-render the frontmatter specifically — that's the load-bearing uncertainty.
Note
Low Risk
Adds a new markdown essay with YAML frontmatter; no runtime or infrastructure code changes, so risk is limited to potential content/frontmatter parsing/rendering issues.
Overview
Adds a new public essay,
writings/fourteen-hours-with-opus-4-7.md, including full YAML frontmatter metadata (uri/slug/date/tags/derivation) and a long-form field report on a 14-hour Opus 4.7 production session.No other files are modified; this PR is a content-only addition that may affect site rendering/extraction only insofar as the new frontmatter and headings are consumed by existing tooling.
Reviewed by Cursor Bugbot for commit 79b52c2. Bugbot is set up for automated code reviews on this repo. Configure here.