Week of August 8: fewer posts, more subscribers, and a press wave

Pageviews fell 13% as output dipped, but confirmed new subscribers rose to 168 and RuntimeWire drew coverage from WIRED, Gizmodo, and more.

41,661 weekly pageviews

Key metrics

  • 41,661 weekly pageviews (-13% vs previous period)
  • 217 articles published (-12% vs previous period)
  • 168 new subscribers (+79% vs previous period)

Major traffic events

  • The Grok settings story led the week — “Grok Bot's hidden "Elon-Only Settings" picker lists 33 models” drew 5,121 views.
  • Press coverage broke wide — Verified mentions included WIRED, Gizmodo, Techmeme, The Media Copilot, Poynter, TechRepublic, and CNN's Reliable Sources Newsletter.
  • Subscribers rose as traffic fell — New confirmed subscribers reached 168, up from 94 the prior week, while weekly pageviews fell to 41,661 from 47,745.

Top stories

Product & newsroom experiments

  • RuntimeWire Benchmarks went live — Independent model leaderboards (BBEH Mini, Newsroom Reliability v0.2) on a strict run protocol: live lane preflights, frozen per-run reasoning effort, and failed-row accounting so a score can never quietly lie.
  • Earnings fast lane — Earnings drops now publish verified coverage within minutes — every figure checked token-for-token against the release, dedupe guards, homepage auto-feature, and a $TICKER post on the drop.
  • Publish Race coverage timelines — Articles can now show who covered a story first, with attested timestamps only — RuntimeWire never claims a win it can't prove.
  • Google search results cleaned up — Crawler-parity HTML on every indexed route, www→apex 301s, real favicons, a shared 60-character title budget, and hard 404s for junk paths.
  • Head-to-Head got more honest — Per-provider catalog health with an on-demand test button, roster classification fixes, and showdown results now publish as full articles with their own feed.
  • Daily recap + live show segments — The recap video and live broadcast moved to segment-based generation — tighter cuts, faster recovery when a piece fails.
  • These updates now draft themselves — The weekly In Public update auto-drafts every Saturday 8am CT from real numbers, then waits for a human. This draft started as one.
  • Admin users manager — Searchable directory of every account with linkage badges and guarded, cascade-safe deletion.
  • INSiDE + article polish — New community bot conversation starters, an article like button, inline image captions, ELI5 improvements, and audio that generates the first time someone hits play.
  • New Tab extension 0.3.5 — The daily quote grew a hover copy/share row — Copy, X, Threads, LinkedIn.
  • Hardening week — Public API audit fixes, an admin demo mode, RSS fetch fallback for blocked feeds, curator and breaking-news tuning, and infrastructure status monitoring.
  • 10,933 lines of new tests — 69 test files touched this week — benchmarks, earnings dedupe, race conditions, route guards, and crawler rendering all landed with tests.

External recognition

Lessons, mistakes & surprises

I’d tentatively flag the divergence between traffic and subscriptions: pageviews fell 13%, while new confirmed subscribers rose from 94 to 168. Publishing also declined from 248 articles to 217, so the relationship between output and subscriber growth is worth watching.

I’d also treat the press coverage as a meaningful milestone, with verified mentions spanning WIRED, Gizmodo, Techmeme, The Media Copilot, Poynter, TechRepublic, and CNN's Reliable Sources Newsletter.

A note from Ryan

Another wild week, mentioned all over the place but still saw traffic pretty flat -- which goes to show what a viral week we had the week before! Making huge SEO changes as I noticed we're still being discovered by Google (meaning showing up in Google App and News) and getting 10% of the SEO traffic as our competitors. So that's my main focus this week -- along with a few more scoops of course!