dmesg — system messages from the web

What the world printed today, and what we think about it. Facts in dim, takes in bright green. Never blurred.

2025-07-01 (number refreshed Sept 2026) · Cloudflare has now blocked 416 billion AI crawler requests

- Cloudflare told WIRED it has blocked 416 billion AI bot requests since July 1, 2025, when its "Content Independence Day" change made blocking AI crawlers the default for newly onboarded zones. Source: https://www.wired.com/story/big-interview-event-matthew-prince-cloudflare/ (WIRED Big Interview, 2025/26 stage; figure quoted by Cloudflare's Matthew Prince) · Cloudflare PR: https://www.cloudflare.com/press/press-releases/2025/cloudflare-just-changed-how-ai-crawlers-scrape-the-internet-at-large/ - The default applies at the network layer, above robots.txt — a site's own file can say Allow: / while the CDN still challenges AI user-agents.

take: This is the biggest quiet infrastructure change of the year, and almost nobody downstream has priced it in. A default-on block that site owners never chose is now the single largest determinant of whether ChatGPT can read your website — and the number 416B is a floor, not an estimate: it counts only requests Cloudflare saw. We built a probe specifically to catch this mismatch; it fires on OpenAI's own homepage. The web didn't close to AI crawlers by vote. It closed by checkbox.

2026-09-15 (imminent) · Cloudflare's new default can make "block AI" mean "block Google"

- Starting September 15, 2026, Cloudflare judges mixed-use crawlers (Googlebot, Bingbot, Applebot — anything that both indexes search and collects for AI training) by the strictest applicable rule a customer set. A site that flips on "block Training" will therefore also block Googlebot at the network layer. Source: https://blog.cloudflare.com/content-independence-day-ai-options/ (Cloudflare's own announcement: "defaults will be enforced by the most restrictive applicable rules") · SEJ write-up: https://www.searchenginejournal.com/cloudflares-ai-crawler-rules-can-block-googlebot/581385/ - Separately, for new zones on ad-supported pages, Training and Agent categories default to blocked while Search stays allowed — the compensation model hardening into infrastructure. Source: https://www.cloudflare.com/press/press-releases/2025/cloudflare-just-changed-how-ai-crawlers-scrape-the-internet-at-large/ + SEJ. (Secondary date/details: MEDIUM; the CF blog text itself is STRONG.)

take: This is the accidental-degradation story the industry has been waiting for, and almost no site owner has read the changelog. For five years, "block AI bots" was a safe toggle because the bots were disjoint from Google's; as of next Monday (Sept 15) Cloudflare welds the categories back together at the CDN layer, where no robots.txt audit can see it. Every "we blocked GPTBot and traffic didn't move" dashboard is about to get stale, and every SEO tool that only parses robots.txt — which is all of our competitor checkers — will keep reporting green while their customers' Google indexation quietly dies. The report that catches this class of failure has to probe live and say the sentence nobody else prints: "the block is not in your robots.txt." That's why our spec asks for a probe engine — and it's about to need one more user-agent in the list.

2026-07-14 · Profound publishes real prices for the first time: $99 to enter

- AI-visibility platform Profound, previously sales-led and demo-only, posted self-serve pricing: Starter $99/mo (ChatGPT only, 50 prompts), Growth $399/mo (3 answer engines, 100 prompts), Enterprise custom (up to 9 engines). Source (pulled directly from their pricing page by us, 2026-09-10): https://www.tryprofound.com/pricing - Entry-tier "AI visibility" now spans $20–$295/mo across Rankscale, Otterly, Peec and AthenaHQ per 2026 roundups — the category has a real price shelf where it had none a year ago.

take: A funded category's first public price list is always an admission: they need volume now, not conversations. The number that matters isn't $99 — it's that ChatGPT-only is the entry product. For the first time the market is saying out loud what one AI engine's visibility is worth, and it's priced like a mid-tier SEO SaaS. Watch for someone to sell it cheaper than that within a year; commoditization from above is rare, from below it's inevitable.

2026-08-28 · Somebody scanned all 197 YC S26 startups; most are open to AI and illegible anyway

- Potatometer's author ran a technical scan of every YC Spring 2026 batch startup's listed URL; headline finding: "Most are readable by AI crawlers. Most don't tell a crawler what they are." Source: https://news.ycombinator.com/item?id=48639511 - A separate 249-store UK ecommerce crawl (KoreLens) found only 7/248 sites blocked any named AI crawler, but only 2% of homepages exposed product prices in machine-readable markup — "open and illegible." Source: https://korelens.com/resources/uk-ecommerce-ai-readiness-2026 (self-published, single author — WEAK evidence, directional only)

take: The interesting failure mode of AI visibility keeps not being the wall — it's the empty room. The best-funded, most technical companies on the planet ship sites that a crawler can fetch and still not understand, because nobody owns the question internally. Blocking is a decision someone made; illegibility is the sound of no one deciding anything. Every "is my site blocked by ChatGPT" scare in the comments section is downstream of this: the fear is rarely about bots being shut out, it's about no one having checked what the bots would even read.

2026-09-06 · Peec AI: $29M raised, $4M ARR in ten months, growing 300 customers/month

- Per a 2026 competitive analysis, Peec AI raised $29.1M across two rounds (Series A Nov 2025), crossed $4M ARR within ten months of founding, and adds ~300 customers/month; every self-serve tier caps at 3 engines, extra models +$30–140/mo. Source: https://thenextscoop.com/peec-ai-vs-otterly-ai (secondary reporting, not a filing — MEDIUM) - Peec is also, per the same piece, the only mid-market visibility tool shipping an MCP server, letting customers query their visibility data from Claude/Cursor directly.