In a recent Jukebox Podcast from WP Tavern, Nathan Wrigley spoke with Damon Cook about WP Trend Watcher, an open-source project Damon built to help busy WordPress professionals keep up with the flood of news, proposals, releases, and community updates.
Why build WP Trend Watcher
Damon has been a WordPress developer and contributor for more than a decade. As a freelancer and contractor he needs to stay ahead of changes in Core, plugins, themes, and the broader ecosystem so he can advise clients, prepare for new features and avoid being surprised when something lands. The project began as a selfish tool to reduce the time spent scanning dozens of feeds while still preserving human judgement and verification.
What WP Trend Watcher does
The tool aggregates a curated set of RSS feeds across the WordPress ecosystem, runs AI-powered summaries on new items, and presents a compact weekly briefing: a numbered, bullet-style list of items with short one-line summaries and links. Crucially, the workflow keeps a human in the loop: Damon reviews the AI summaries, clicks through sources, corrects or augments notes, and adds a final human commentary before publishing the report.
This approach aims to cut the dead time of manually finding interesting links while avoiding blind trust in AI-generated text. The AI surfaces candidates, and the human decides what’s accurate and worth attention.
AI, prompts and local models
Damon designed the system to be prompt-driven and to run on local models where possible. Each feed item is passed through an LLM with a prompt tuned to produce summaries aimed at developers and agency owners — focusing on factual, actionable points rather than popularity. The tool also attempts to detect duplicates, weak language, or missing sections so it won’t summarize the same story multiple times.
To experiment with model quality and cost, Damon runs models locally using LM Studio on his Mac. His setup is constrained by available memory (he mentions an 18GB-capable machine), so he has tried several compact models (examples include Qwen-style and other emerging local models). The project records which model was used for each weekly run so outputs can be compared and results made more deterministic over time.
Human-in-the-loop and transparency
Damon is explicit about resisting full automation. He runs the report, audits the AI output, and writes human notes — both as fact-checking and as a way to internalize the topics. He publishes the reports publicly and is transparent about how much time he spent reviewing each week, inviting readers to judge whether the human review was sufficient.
He admits a bit of cognitive dissonance — feeling “cheap” for leaning on AI — but the public, open-source nature of the project is part of the answer: others can inspect, fork, add sources, or modify the workflow. The project keeps receipts (past reports) that could eventually be used for longitudinal trend analysis.
Publishing and workflow
WP Trend Watcher is built in TypeScript, hosted on GitHub, and the published briefings are served via GitHub Pages. Damon runs the summarization process locally on a weekly cadence, reviews and edits the output, then publishes the finalized report. Because of the required human step, the process isn’t fully automated — intentionally so.
Filtering and what gets included
The tool filters a lot of noise: it checks for duplicate coverage across sources, looks for weak or non-factual language, and biases prompts to prioritize developer- and agency-relevant facts. Not everything harvested by the feeds makes the final briefing. Damon can see what was skipped and iteratively tune sources and prompts to improve relevance.
Results so far
Published editions are compact: roughly a dozen to twenty items with short, linkable summaries. Damon reports that the summaries are generally accurate when prompted appropriately, but he still clicks through many items to validate the AI output and to add any missed nuance in the human summary.
Why it matters beyond time saved
Beyond saving time, AP-style briefings create a reusable corpus: as more weekly reports accumulate, they form a historical record that can be reanalyzed to detect longer-term trends. Damon envisions future features where AI analyzes past reports to infer trends, explain why certain ideas gained traction, or surface the chain of events behind Core decisions. The same pattern could work for other industries — finance, product development, niche communities — not just WordPress.
Open source and getting involved
Damon’s codebase is public. He encourages others to fork it, add sources, try different local models, or build additional analytics on top of the weekly briefings. He also shares the reports on social platforms and his blog, inviting feedback and source suggestions.
Final thoughts
WP Trend Watcher is an example of a pragmatic, human-centered use of AI: use models to reduce discovery overhead, preserve human judgement for verification and context, and publish transparently. It’s a practical step toward making the relentless flow of WordPress news manageable for freelancers, agency owners, and developers who need to stay current without spending hours each day reading every feed.
If you want to try it yourself, the repository, the published reports, and setup notes are available from Damon’s public pages. The system is flexible: you can run it locally, tune prompts and sources to your interests, and keep the human-in-the-loop step that makes the output trustworthy and memorable.