You Know What AI Agents Are Great At? ![rw-book-cover](https://pbs.twimg.com/profile_images/1825702427640705024/-5Et3Ydl.jpg) 🔗 **Source:** [You Know What AI Agents Are Great At?](https://twitter.com/austinginder/status/2098948269241978995/?rw_tt_thread=True) > [!note]+ Document Notes > tags: [ai-assisted-development, wordpress-plugins, fable-5-1, abandoned-projects, wp-shipyard] > author: Austin Ginder > source: X / Twitter > reference: https://twitter.com/austinginder/status/2098948269241978995/?rw_tt_thread=True > date created: 2026-09-19 16:27 > > > [!info] Generated analysis — not the author's words. Sections below are machine-written > > from the captured source. Only text inside `> [!quote]` blocks is verbatim from the author. > > Do not quote or attribute any other line to the named author. > > <!-- deep-read: light-touch --> > > # 🏷️ Title: You know what AI agents are great at? Finishing half started WordPress plugins from 3 years ago > <!-- deep-read:sections start --> ## 📝 Summary - Ginder reports finishing a three-year-dormant WordPress plugin in about an hour using Fable 5.1. - The result, WP Shipyard, is now at v1-release quality by his assessment. - Matters because the claimed win is on resurrection of abandoned work, not greenfield building — a different and more testable use case. - Bottom line: one developer's single data point, with a named model, a named timeframe, and a shipped artifact. ## 📖 Main Content - Half-finished projects are unusually good agent territory: the architecture is already decided, so the remaining work is mostly completion rather than design. - 'Good enough for a v1 release' is the author's own bar, not an external quality measure. - Fable 5.1 and roughly one hour are specific enough to attempt a comparable test. - Unverified: nothing here says what was left to do, so the hour is not comparable across projects. ## 🗒️ Notes - wpshipyard.com is the artifact; it can be inspected directly. - Ginder is the same author whose Elementor Pro RCE writeup informs the agent-audit guardrails in Courtney's CLAUDE.md — a known-credible source on agent workflows. - Directly applicable: Courtney has dormant plugin repos in ~/Documents that match this shape. > [!quote] > Finishing half started WordPress plugins from 3 years ago. In like an hour with Fable 5.1, WP Shipyard is now good enough for a v1 release ## 📚 References - [WP Shipyard](https://wpshipyard.com) ## 🎯 Conclusion - The transferable claim is about project shape, not model capability: decided architecture plus unfinished implementation is where agents pay off fastest. - Next step: pick one dormant repo and run the same experiment with a fixed time box. ## 🔄 Refactor Appendix — You know what AI agents are great at? Finishing half started WordPress plugins from 3 years ago ### 📚 Literature Atoms (H3) #### [H1] Agents finish dormant projects faster than they start new ones > [!quote] > AI agents are great at... Finishing half started WordPress plugins from 3 years ago **Context & use**: - Post opening, Sep 13 2026 - The architecture already exists, so the remaining work is completion **Tags**: #xref #articleseed <!-- atomicity: 1 idea ✓ --> #### [H2] Ginder shipped WP Shipyard v1 in about an hour > [!quote] > In like an hour with Fable 5.1, WP Shipyard is now good enough for a v1 release **Context & use**: - Same post - Named model and timeframe make the claim reproducible **Tags**: #xref <!-- atomicity: 1 idea ✓ --> ### 🧠 Permanent Candidates (H4) #### [P1] Decided architecture is what makes agent work cheap **Claim** (one sentence, Courtney's voice): An agent's speed advantage is largest where the design decisions are already made and only implementation remains. **Evidence pointer**: [[#H1]] **Scope & conditions**: Holds for abandoned projects with committed structure; fails where the original design was the thing that stalled the project. **Links** (≥1 with annotation): - [[Agent-task-selection]] — because extends - [[Dormant-repo-revival]] — because applies-in personal projects <!-- atomicity: 1 idea ✓ --> ### ✏️ Fleeting Queue - [F1] Which dormant repos in ~/Documents have settled architecture and unfinished implementation? #fleeting #inbox ### → Things 3 (max 2 actions) - [ ] Time-box one dormant plugin revival and record the actual hours #things <!-- deep-read:sections end --> 📄 **Full text:** [[Full Document Contents/Tweets/You Know What AI Agents Are Great At.md|You Know What AI Agents Are Great At? by Austin Ginder]] ## Highlights - In like an hour with Fable 5.1, WP Shipyard is now good enough for a v1 release: ([View Highlight](https://read.readwise.io/read/01m2zdsrgt8388qtw1m5v3a4zm))