Start with two full samples. Sign in for this edition’s free story. Sources and corrections stay open.
What the pen marks mean
rememberThe takeaway: what to remember
figureThe number that matters
evidenceThe evidence line
rulingThe ruling
claimThe claim that fails
Each mark is drawn once, as you reach it. Numbers are only circled when they come from the story’s own figures.
01 / True, butFree sample
An army of AI agents did plant shells on real websites. But the companies were named by Reuters, the hits were websites, and the $5,000 is a guess.
Slashdot, quoting The Verge, 19 October 2026, with the $5,000 figure from coverage of Reuters' report ·
AI illustration
NorForge watched exposed Cairn panels plant uploader shells on shop sites. Deere, SharkNinja and Cummins describe limited website incidents, and the cost figure is a researcher's estimate.
The reality check
NorForge found exposed Cairn panels planting uploader shells on websites. Reuters named the firms; Deere saw no sign of sensitive data accessed and SharkNinja says a standalone support site was hit. The $5,000 is an estimate; the report says it has no token numbers. THE VERDICT: Needs context.
Why care? Automated website intrusion is getting cheap, which matters to anyone running a web shop. But a shell on a support site is not a breach of Deere, and an estimate is not a receipt.
Take this with youAI agents did plant web shells on more than 100 websites. The companies were named by Reuters, the hits were websites, and the $5,000 is a rough guess.
Open the evidence4 source pages +
The claim we checked
A Hacker Broke Into John Deere, SharkNinja With an Army of AI Agents, breaching more than 100 company websites for about $5,000.
These are quoted receipts, not a count of independent investigations. Several reports may rely on the same original source.
Four AI agents did kill the Ender Dragon on their own. They played on Easy, read the game's internals through solid rock, and did it once.
Google DeepMind (@GoogleDeepMind post on X, as carried by Digg), October 2026, describing its arXiv paper ·
AI illustration
The arXiv paper backs the numbers and the no-human-help claim. Its own limitations section says the agents saw hidden blocks, played on Easy and ran a single attempt.
The reality check
The paper confirms 2,783 programs, a dragon kill in 27 hours 14 minutes and no human input. It also says the agents read hidden blocks through code, played on Easy, and this was a single run. THE VERDICT: Needs context. Real result, narrower game.
Why care? This is a strong long-horizon coding result, not proof that AI plays Minecraft the way a person does. Scope any 'first' to the authors' own conditions.
Take this with youFour AI agents did kill the Ender Dragon with no human input. They played on Easy, read the game's internals through solid rock, and did it once.
Open the evidence5 source pages +
The claim we checked
Four Gemini 4 Argon agents beat Minecraft from scratch in 27 hours.
These are quoted receipts, not a count of independent investigations. Several reports may rely on the same original source.
Two community projects have recently reported end-to-end dragon kills: a single Claude Opus 5 agent calling 37 high-level Minecraft tools on Normal difficulty
An Anthropic co-founder told the IMF the singularity has begun. His evidence is test scores and anecdotes, and his own speech calls them first signs.
Jack Clark, Anthropic co-founder, in his Per Jacobsson lecture at the IMF annual meetings, republished in Import AI 477; headline via The Rundown AI ·
AI illustration
Jack Clark's lecture in Bangkok rests on benchmark jumps, AI-assisted maths and physics results and a tenfold rise in code at Anthropic. None of that is yet a measured speed-up in science or the economy.
Members · 30 days free
See what the evidence actually shows.
Members read the full check on every story: what the evidence shows, why it matters to you and the one line to take with you. Every past edition, re-verified, and the Receipts Pack come with it. A$89 a year, about A$0.24 a day.
At sign-up: 30 days free. Payment card required. One introductory trial per customer.
Your trial ends 30 days after you start. Unless you cancel before then in Account → Manage subscription, we charge A$89 for the first year. It renews automatically at A$89/yr until cancelled.
You can ask for a full refund within 14 days after any annual payment, renewals included, with no reason needed, by emailing hello@bskiller.com. This voluntary refund does not limit your rights under the Australian Consumer Law.
Reading estimate: 762 words at 200 words per minute. Source quotes and the optional sections below add reading time.
Get tomorrow's check in your inbox.
One AI claim a night, checked against independent sources, receipts attached. Free.
At sign-up: 30 days free. Payment card required. One introductory trial per customer.
Your trial ends 30 days after you start. Unless you cancel before then in Account → Manage subscription, we charge A$89 for the first year. It renews automatically at A$89/yr until cancelled.
You can ask for a full refund within 14 days after any annual payment, renewals included, with no reason needed, by emailing hello@bskiller.com. This voluntary refund does not limit your rights under the Australian Consumer Law.
