AI-ingest is my personal filter for the AI firehose.
Instead of opening X, blogs, newsletters, and random links every morning, the idea is simple: collect the useful AI updates, score them against what I care about, and turn them into one short brief I can read quickly.
Small README#
AI-ingest collects AI-related sources, ranks them against my interests, and generates a short morning brief.
The flow is:
- Collect updates from sources.
- Score them against my personal context.
- Summarize what matters.
- Serve a small morning page.
The repo has scraper and brief-generation scripts, a GitHub Actions daily workflow, stored brief data, and a small Next.js web view.
How To Use#
The public version is simple: open the live page and read the morning brief.
For the repo workflow:
npm run scrape
npm run briefThe daily workflow keeps generating new brief files, so the tool becomes more useful as a repeatable routine instead of a one-time page.
What broke after launch#
The useful engineering work started when the routine became boring enough to depend on. A source could fail silently, an LLM could return malformed output, or a date boundary could make the pipeline read the wrong file. A broken scrape could even look like a genuinely quiet news day.
I hardened the pipeline so low-volume or heavily failed scrapes stop loudly instead of overwriting a good brief. Model output is shape-checked, titles and links come from the scraped source data rather than being retyped by the model, and raw scrape files remain available for debugging. The scheduled workflow was still producing brief commits in September 2026.
What I Learned#
The main lesson was that information overload is not solved by collecting more information.
The useful part is ranking.
AI-ingest taught me that a good agent workflow needs personal context, a narrow goal, and a final artifact someone can actually use. In this case, the artifact is not a giant dashboard. It is a short morning brief.
Why It Matters#
This project is useful because it turns research into a habit.
I do not want to chase every AI update. I want to notice the few updates that connect to what I am building, learning, or writing about.
This is still a personal filter, not a guarantee that every selected story or generated explanation is correct. The sources remain the place to verify anything important.