A nightly scan that turns layoffs, shutdowns, and funding distress into a short, ranked list of people worth reaching out to - delivered as a one-page card.
Every night it reads the day's talent-market signals, weighs each one against the role you're hiring for, and emails you a single card: who just became reachable, why now, the line to open with, and the concrete edge your company can offer.
Built for recruiters, TA, and People-Ops. No coding required.
The image above is a real run on fictional companies. Open
card/sample-card.htmlin any browser to see the live version.
- One card, ranked. The few strongest opportunities, tiered S / A / B, each with a why now, an outreach hook, and the specific edge your company offers (equity, relocation, comp, ownership).
- A fit fingerprint. A six-axis hex per target - openness, role, product, region, size, stage - so you read fit at a glance.
- A running memory. Every signal it has already seen is remembered, so you're never pinged about the same event twice.
It arrives as a clean HTML email. Nothing is ever sent to a candidate - the card is for you.
This isn't one model reading headlines and guessing. Every candidate is argued by two independent AI judges - an Advocate makes the case, a Skeptic pushes back - and the two are reconciled before anything reaches your card. One model never grades its own work. Freshness is weighed in, so an hours-old layoff outranks a stale one. You always get the final call.
Left to right: a nightly trigger pulls two signal feeds (WARN layoff notices + funding-distress news), a fast model shortlists the candidates, an Advocate and a Skeptic argue each one, the two are reconciled into a ranked set (freshness weighed in), and the survivors are rendered into the card and emailed. Every step has a sticky note on the canvas.
Full click-by-click is in docs/setup.md. The short version:
- Import
n8n/talent-alpha.workflow.jsoninto n8n. - Add keys: Anthropic (required), a free WARN Firehose key, Perplexity (optional), and connect Gmail.
- Create one Data Table named
r04_events_seenwith a text columnevent_id(the dedupe memory). - Open the
Confignode and describe your company + the role you're hiring for. - Hit Execute to test, then set the schedule when you like what you see.
Just want to watch it run first? Import n8n/talent-alpha-demo.workflow.json instead - it ships with sample signals pinned in, so it runs end to end with only an Anthropic key. Every step also has a matching sticky note on the canvas.
| Item | Cost |
|---|---|
| n8n (cloud Starter, to keep it scheduled) | ~$24/mo - or free self-hosted |
| Anthropic (the scoring + judging) | a few cents per run; ~$5 starting credit is plenty |
| WARN Firehose (layoff feed) | free (25 calls/day) |
| Perplexity (funding-distress feed) | optional; a few cents per run |
| Gmail delivery | free |
- Who you're hiring - the Config node. Be specific about your leverage (equity band, comp stance, relocation, benefits, ownership); the judges cite it in each card's edge line.
- How strict it is - the Reconcile Judges node. It shows between 4 and 10 targets and blends the two judges; loosen or tighten to taste.
- What it scans - this ships with WARN layoffs + funding distress (category 1 of many). Duplicate the fetch → normalize pair to add M&A, RTO mandates, exec departures, and more; they all merge into the same judge panel.
- The prompts - the scanner and both judges live in the workflow; reference copies are in
prompts/.
| Path | What it is |
|---|---|
n8n/talent-alpha.workflow.json |
The workflow - import this |
n8n/talent-alpha-demo.workflow.json |
Same, with sample data pinned in - runs with only an Anthropic key |
card/sample-card.html |
The card, as a real HTML page |
docs/setup.md |
Click-by-click setup for non-engineers |
prompts/ |
The scanner + Advocate + Skeptic prompts, in plain text |
sample-card.png |
The card image shown above |
Related: a recruiting-tuned cousin of build-radar - the same scan → judge → card idea pointed at a different signal.
Questions? Open an Issue. Nothing is too basic.
MIT licensed.