Your signal radar
You built a working radar in the room. This is everything you need to run it again on Monday, on your own.
The idea in one line: a feed is a stream of things happening in your market. Your brain (ICP, offer, triggers) decides which of them are signals worth acting on. You run it with two prompts in Claude Code. No terminal.
What you need
- 01
Claude Code open (the tool from Wednesday). You just chat with it.
- 02
Your brain: your ICP, your offer, and the 2–3 trigger types you care about.
- 03
A feed: one RSS URL tied to your market (starter list at the bottom).
Collect
Paste this into Claude Code, with your feed URL dropped in:
Collect the stream
Fetch this RSS feed and save the recent items so I can review them.
Feed: [YOUR RSS URL]
Save all items from the last 7 days into one file: signals/<today's date>/raw.md
(leave older dated folders untouched). List one item per entry, each with:
title, date, link, and the full summary text.
Then tell me how many you saved.Claude Code writes a single raw.md. Open it — that’s your raw stream, before any judgement.
Why a dated folder? Each run lands in signals/<today’s date>/, so today never overwrites last week. Your raw items and scored lists pile up into a history you can look back on.
Score
Fill in your brain, then paste this:
Score against your brain
Read signals/<today's date>/raw.md and score each item against who we are.
OUR BRAIN:
- ICP: [who we sell to]
- Offer: [what we sell]
- Triggers we care about: [e.g. new VP Sales, hiring surge, funding, new market]
Write the results to signals/<today's date>/scored.md as a markdown table, ranked
by priority (highest first), with EXACTLY these columns, in this order:
priority High / Medium / Low
company the organisation
trigger which trigger type fired
what_happened one factual line
why_now the timing
why_them the ICP fit
why_you how our offer connects (the angle)
confidence High / Med / Low
evidence the exact sentence from the source
link the source URL
Anything that is not a genuine buying signal: list it below the table with a
one-line reason. Do not invent signals.What good output looks like:
| priority | company | trigger | what_happened | why_now | why_them | why_you | confidence | evidence | link |
|---|---|---|---|---|---|---|---|---|---|
| High | Northwind | new VP Sales + hiring | Named a VP Sales, posted 12 SDR roles | first 90 days = motion rebuild | mid-market SaaS, our ICP | we sell outbound tooling | High | "…appointed Jane Doe as VP Sales and is hiring 12 SDRs…" | techcrunch.com/… |
Skipped:
- "AI funding roundup" — a list, no single company we serve.Now read it and judge for yourself: real signal, or a plausible miss? Tune the brain or the trigger list and rerun.
Why two prompts? One prompt that fetches and scores drifts and cuts corners. Collecting first, scoring second, keeps each job honest — and you can inspect the raw files in between.
A mention is not a signal.
A real signal answers all three:
- Why now
- — the timing. Why this week, not last month.
- Why them
- — the fit. Why this account matches your ICP.
- Why you
- — the angle. How your offer connects.
If you can’t fill all three, it’s noise.
Your Signal Map
One row per trigger that matters to you. This is the map you point the radar at.
| ICP segment | Trigger type | Feed / source | Why-now logic | Your angle | Priority |
|---|---|---|---|---|---|
| Mid-market SaaS | New VP Sales + hiring | company news, LinkedIn | first 90 days = motion rebuild | "saw you're rebuilding outbound…" | High |
| | | | | | |
| | | | | | |Your source
Start with one RSS feed tied to your market. Most news sites and blogs publish one — look for an RSS link, or try adding /feed to the URL. For example:
https://techcrunch.com/category/artificial-intelligence/feed/No feed on a page you care about? Hand Claude Code the page through a reader instead:
https://r.jina.ai/https://the-page-urlThat’s the whole toolkit: one feed, plus a reader for the odd page that doesn’t have one.
Two checklists worth keeping
A good source is
- about companies you care about
- updates often
- readable without a login
- predictable in format
Before you trust a signal
- is it real or invented?
- would the why-now survive a skeptical "so what?"
- is the who specific?
- would you get the same one again tomorrow (or is it a repeat)?
One caution
You’re pasting into an LLM. Public sources and your own ICP only — nothing confidential.