Product Hunt

producthunt.com — Marketing-heavy, but the daily top-5 often includes a Mac app or utility worth a closer look. Feed validated 2026-07-03 (producthunt.com/feed). 2026-08-07: 50 entries, zero passing the bar — the feed is now almost entirely B2B AI tooling (agent SDKs, spend consoles, LLM-markdown converters). Recording the pattern rather than acting on it: the source is healthy, and the occasional consumer app is still worth the scan, but it has not produced a find in several runs. 2026-08-09: feed healthy (50 entries) but again ~95% B2B AI/SaaS self-promo; nothing captured. Also worth recording as a fetch fact: individual producthunt.com/products/<slug> pages now serve a Cloudflare 'Just a moment...' interstitial to curl and WebFetch, so item-level detail has to come from the feed body or a web search — the feed itself is unaffected. 2026-08-22: feed healthy (50 entries, 29 in-window after a three-day gap), zero captured for the sixth recorded run in a row — the in-window set was agent/observability SaaS, an IDE extension pack, an OCR wrapper and a Calendly relaunch. The 08-07 diagnosis holds unchanged; kept daily because the scan is one cheap feed fetch. Dated line updated in place rather than appending a sixth paragraph. **2026-08-25: two corrections and the first productive run on record.** (a) The feed is ATOM, not RSS — entries are <entry> with <link href=...>, so a parser looking for <item>/<link>text finds nothing and the source silently reads as empty; this is almost certainly part of why it has scanned as zero-yield for seven runs. (b) The 08-09 note said producthunt.com/products/<slug> pages are blocked to curl AND WebFetch; only the first half still holds — curl gets a ~5.7KB Cloudflare challenge, but WebFetch resolves those pages fine and returns tagline, maker site, platform and price. Item-level detail is therefore available again. Two finds captured off it this run (openlogi, wavepocket), both consumer software rather than B2B AI, which is the first break in that pattern since 08-07. 2026-08-27: product pages return HTTP 403 to curl but WebFetch succeeds. Feed gives only a one-line tagline, so WebFetch the product page for anything worth writing, then link the maker's own site. 2026-08-29: feed healthy but only 7 in-window entries, all agent/AI tooling (a notch utility, a multi-model prompt runner, a video model, two Slack agents, a 3D platform). Zero captured. The 08-07 diagnosis holds; kept daily because the scan is one cheap fetch. 2026-08-31: feed healthy, 3 in-window entries (oMLX, Superagent, Prequel) — a Mac LLM server, an agent host and a screen recorder. **Seventh consecutive zero-capture run.** The pattern recorded on 08-07 and 08-09 is now the steady state, not a blip; the source is live and the scan is cheap, but it has not produced a find in over a week. 2026-09-01: **second productive run, and it names the extraction rule.** Two finds from a 50-entry feed (Teenage Engineering's EP–2350 FX–MIC; Tether, a Mac desk toy) after seven consecutive zeros. Both were reached the same way and neither was legible from Product Hunt's own copy: open the /products/<slug> page and follow the outbound maker link in it (playtether.app, teenage.engineering), then verify on the maker's own store. The 08-09 Cloudflare note is superseded — product pages fetched fine with a Safari UA and --compressed this run; og:description on the PH page carries the full pitch. The 08-07 diagnosis of the feed's composition is otherwise unchanged: the other 48 entries were B2B AI tooling.