Daily auto-collect · weekly ranked report (opt-in) · runs unattended
AI Scout Pipeline — Auto-Track the Latest AI Papers & Tools
The AI Scout Pipeline runs on your own machine and does it for you. Every day it stages the newest papers and fastest-rising GitHub repo candidates — spending zero AI tokens. If you enable the weekly run, it runs unattended: your agent ranks the pile into one report and writes a 'review is ready' line to the log (or emails you if you set up SMTP) — INTEGRATE / INVESTIGATE / WILDCARD / SKIP, each with a one-line reason. You just open it and decide.
Instant download · one-command installer · no extra API keys · no monthly fees
The problem
A plain keyword search misses candidates
Manually scanning feeds is slow, and a plain keyword search alone can miss the repo that's quietly gaining stars fastest this week.
▌ Doing it by hand
- You check HuggingFace and GitHub manually — or you don't, and fall behind
- A keyword search misses fast-rising repos that don't match your terms
- No filter — you read everything or you read nothing
- Burns your time daily, or burns AI tokens summarizing noise
▌ AI Scout Pipeline
- A daily collector stages paper + repo candidates automatically, spending $0
- Two feeds: GitHub keyword search AND an OSS Insight trending (fastest-gaining-stars) feed
- Enable the weekly run once and it ranks the pile on its own schedule
- You get one ranked report with a one-line reason on every item
What it does
Two jobs: collect daily for free, rank weekly if you enable it
The daily collection spends nothing. If you enable the weekly run, the pipeline runs unattended, ranks the pile with your agent, and drops one report where you'll find it.
Auto-collect for $0
Every day it pulls the newest papers from HuggingFace and the newest + fastest-rising repo candidates from GitHub, and stages them as clean, readable files. No tokens, no extra keys, no fees.
Ranks the pile, then logs it
Once you enable it, it runs unattended on a schedule: your agent ranks the week's pile into one report and tags each item INTEGRATE / INVESTIGATE / WILDCARD / SKIP with a one-line reason. Then it writes a 'review is ready' line to the log — or emails you if you've set up SMTP in scout.conf.
Catch more candidates
It pulls two feeds — a GitHub keyword search for new-with-traction repos AND an OSS Insight trending feed for repos gaining stars fastest this week — so it catches more candidates than a keyword search alone.
Surface the surprises
Anything the ranking is unsure about lands in a Wildcards list, so a candidate you didn't have a keyword for still gets surfaced for you to scan.
What's in the box
Everything to get it running
- ✓One-command installer — paste one line, no coding
- ✓The daily collector + the weekly auto-run + the walkthrough skill for Claude Code
- ✓An illustrated setup guide and a full explainer of how it works
- ✓Fully editable — change what it hunts for in one list
- ✓Two crons scheduled on install — daily collection + the weekly ranking (off until you flip the switch)
- ✓The /scout-weekly walkthrough — walk the report with your agent and decide what to build
⚠ Requirements — please read before buying
- Claude Code (drives the weekly ranking — it spends Claude Code tokens)
- macOS or Linux — Windows works via WSL
- Python 3 (already installed on Mac/Linux)
- No extra API keys, no paid services beyond Claude Code
Stop falling behind
Let your machine watch the feeds for you
Auto-collect daily for $0. Enable the weekly run and it ranks the pile on its own, writing one report to review. Unzip, run the installer, flip the switch when you're ready.
Instant download · 30-day Gumroad refund policy applies
Setting up Claude Code from scratch? Start with the Claude Code Setup Playbook — $29.
See all my Claude Code tools → expressive446.gumroad.com