DeepSignal

Hey IH 👋

Tomorrow's my first PH launch. Built DeepSignal solo over 3 months.

The pitch: my AI-news reading was a mess — 30 min/day across X, HN,

3 newsletters, and RSS, still missing things. So I built the tool I

wished existed.

What it does:

- Aggregates 27 sources (frontier labs / model vendors / chip OEMs /

arXiv / HN)

- Every article gets an AI summary + 0–100 Signal Score (source 30 +

technical 25 + business 15 + novelty 30, half-life 1.5d). Formula

is open on every card so you can argue with the score.

- Featured rail caps any single source at 2 slots — no source flooding

- Bilingual EN/中文 at ingest (not request-time translation)

- Daily brief to email/Slack/Feishu, your local 8am — picks YOUR

local-yesterday Top-N, not UTC

Stack: Next.js 15 + Supabase + Vercel Pro Cron + OpenAI gpt-4o-mini.

22 idempotent SQL migrations, 7 cron routes. Solo + bootstrapped.

Cost economics:

- ~$46/mo at launch (Vercel + Supabase + OpenAI)

- ~$75/mo projected at 1k subs

- Daily brief computed at ingest, so per-subscriber LLM cost is ~$0

(caps unit economics tightly even at scale)

Pre-launch numbers: 800+ articles indexed. Past 7 days from SEO alone:

142 visitors, 67% bounce. Not huge, but Google indexing kicked in faster

than expected on a 5-day-old domain (sitemap + IndexNow paid off).

Tomorrow 14:00 UTC+8 (00:01 PT), it goes live.

If you read AI news daily, would mean a lot if you tap "Notify me":

https://www.producthunt.com/products/deepsignal?utm_source=ih&utm_medium=launching_post&utm_campaign=launch

What I'm hoping to learn:

- Sources I'm missing (especially APAC + EU labs)

- Whether the Signal Score weighting feels right to other AI builders

- Daily brief format (length, frequency, vertical ordering)

Happy to answer any technical / build / cost questions in the comments.

Thanks IH 🙏