A weekly AI intelligence briefing that builds itself. Every Monday morning it scans what AI builders are actually shipping: new GitHub projects, trending tools, builder forums. Your personalized agent then sends you one decision-ready email: the 3–5 'hot AI developments' worth knowing, one thing to try this week, and an honest "hype check" on what looked big but wasn't.
Built for recruiters, TA and People-Ops professionals — or anyone who wants to stay genuinely current on AI without living on Hacker News. No coding required for the core setup (email digest + searchable archive) — the guide assumes you've never used n8n.
- An email digest — the movements, why each matters to you, one concrete "try this" action, with links to the evidence.
- A Google Sheet archive — every scored item, searchable, week over week. (Prefer Slack delivery, or no Google at all? Both are two-minute swaps — see CALIBRATE.md.)
And if you want to go further: the one-page visual card above. That one is an
advanced, optional extra — it needs a few manual steps beyond the core setup
(see card/). You lose nothing essential by skipping it.
This isn't "AI summarizes the news." Three guardrails are built in:
- Evidence or it doesn't ship. A trend needs at least two independent sources — or one genuinely exceptional release, clearly flagged as single-source.
- A harsh filter. Most items score 2–5 out of 10 and never reach you. Marketing posts, hype threads, and starter templates get cut.
- Honest quiet weeks. If nothing clears the bar, it says so — it never pads.
- AI usage: typically under $2/month at weekly runs. The workflow is set up for Anthropic (Claude) out of the box, but any provider key works — OpenAI, Google Gemini, and others are a model-node swap away (see CALIBRATE.md). Funding an API key requires a payment card and ~$5 of starting credit.
- n8n: the cloud free trial is time-limited; keeping scheduled workflows running afterwards needs a paid plan (Starter is ~$24/month), or self-hosting n8n for free if someone technical can set that up for you.
Follow SETUP.md — click-by-click, written for non-engineers. You'll need: an n8n account, an Anthropic API key (or an OpenAI / Gemini key — see CALIBRATE.md for the swap), and a Gmail account.
This is what you're importing — the whole pipeline on the n8n canvas:
Left to right: ten sources are fetched and parsed, a fast model (Haiku) scores every item, a smarter model (Sonnet) groups the survivors into movements, and the result fans out to the email digest, the archive, and the visual card. The dashed connections are the AI model and format-repair nodes.
- Change what it watches: open the
Emit Feed Listnode — the sources are a short list you can edit in plain text. - Follow specific people: point it at anyone on LinkedIn — free and scrape-less, using LinkedIn's own post-alert emails (see the last section of SETUP.md).
- Change who it's for: the "builder profile" inside the
Scorernode describes the reader. Rewrite one paragraph and the whole radar re-aims.
| File | What it is |
|---|---|
radar-workflow.json |
The automation — import this into n8n |
SETUP.md |
Click-by-click setup for non-engineers |
CALIBRATE.md |
Every tuning knob, in plain English |
sample-card.png |
Example weekly card (illustrative sample) |
workflow-canvas.png |
The workflow's node graph, as it appears in n8n |
card/ |
Advanced, optional: the visual-card template + fill kit |
Stuck or have a question? Open an Issue — no question is too basic; this repo exists for non-engineers.
MIT licensed © 2026 onepromptman (Bryan Hong). Use it, adapt it, share it.