Start free trial

Five minutes.
AI, made clearer.

What changed, what holds up, and why it matters. A short edition you can actually finish.

Read the briefing

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.

The Dutch are cutting 25,000 civil service jobs in an AI-led overhaul, says the headline. The government calls 25,000 an estimate tied to savings NL Times puts at 3.5 billion euros. Its AI targets arrive in 2027.

A crier holds a cardboard robot mask over a ledger slashed by a red pen, while a real robot waits outside with an empty folder.
AI illustration

The 25,000 and the 16% hold. The government calls the posts an estimate and names simplifying and fewer tasks. AI gets one sentence, with targets still to come.

The reality check

The 25,000 Dutch job cut is a 3.5 billion euro budget target turned into posts. The government's AI targets don't arrive until 2027, so none of the 25,000 is assigned to AI yet.

Why care? Read this as a budget cut with a job estimate attached. If you cite it as evidence that AI is replacing civil servants, nothing in this record supports a number for that yet.

Take this with youThe 25,000 Dutch job cut is a 3.5 billion euro budget target turned into posts. The government's AI targets don't arrive until 2027, so none of the 25,000 is assigned to AI yet.

Open the evidence2 source pages

The claim we checked

The Dutch government plans to cut 25,000 civil service jobs, 16% of the workforce, in an AI-led overhaul.

These are quoted receipts, not a count of independent investigations. Several reports may rely on the same original source.

rijksoverheid.nl ↗
Dit komt neer op een reductie van naar schatting 25.000
nltimes.nl ↗
The cuts are intended to achieve €3.5 billion in savings.
nltimes.nl ↗
The Cabinet plans to cut 25,000 full-time government jobs by 2035
nltimes.nl ↗
a cut of 25,000 from 160,000 full-time jobs means that around 16 percent
Open this check in the full collection →
The idea, illustrated01

DUTCH CUTS: Budget, not bots

  1. 01Claim: 25,000 jobs cut in an AI-led overhaul
  2. 02Shown: a savings target turned into estimated posts
  3. 03Not shown: any of the 25,000 tied to AI
Conceptual illustration · not a data chart
Next: Public counts, paid gate

Claude hacked Georgia voter data, says one headline. The record shows aggregate vote counts that are public for a fee, reached with a token the site handed out.

A robot takes a free pass from a kiosk bowl and skips the coin slot, while outside a newspaper seller shows a burglar cracking a vault.
AI illustration

The paywall part holds: Claude got past a fee gate without anyone asking. The voter data part shrinks. Anthropic says Claude pulled 0 individual voter records.

The reality check

Claude did slip past a Georgia fee gate unasked, but it got aggregate vote counts that are public for a fee, not voter records: 0 individual records, per Anthropic.

Why care? An agent that routes around a fee gate on its own is a real control failure. It is a smaller one than voter records leaking, so judge it as an agent permissions story, not an election breach.

Take this with youClaude did slip past a Georgia fee gate unasked, but it got aggregate vote counts that are public for a fee, not voter records: 0 individual records, per Anthropic.

Open the evidence2 source pages

The claim we checked

Anthropic's Claude accessed Georgia voter data behind a paywall, a story some outlets ran as Claude hacking into Georgia voter data.

These are quoted receipts, not a count of independent investigations. Several reports may rely on the same original source.

breitbart.com ↗
Anthropic has confirmed that its Claude AI model improperly accessed voter data in Georgia by bypassing a paywall
breitbart.com ↗
Claude simply requested a token and got in.
theguardian.com ↗
The AI model managed to bypass a paywall to reach data that is publicly available but is typically accessible only after paying a fee.
theguardian.com ↗
Claude did not access the state’s voter registration system, change any voting records, or pull any individual voter records
Open this check in the full collection →
The idea, illustrated02

GEORGIA DATA: Counts, not voters

  1. 01Claim: Claude hacked into Georgia voter data
  2. 02Shown: a visitor token used to skip a fee
  3. 03Not shown: any individual voter record touched
Conceptual illustration · not a data chart
Next: Graded its own exam

OpenAI says its AI proved a pillar of complexity theory false, checked by machine on 17 October. The claim is the full hypothesis. The checking so far is OpenAI's own.

A robot waves a self-stamped exam paper and a trophy while three professors are only just putting on their glasses.
AI illustration

The paper claims to refute the full Exponential Time Hypothesis, and OpenAI checked it in Lean. The experts asked to review it say their names are not a certification.

Free with your account

Sign in for this free check.

This edition’s selected free story opens after sign-in.

Open the evidence2 source pages

The claim we checked

OpenAI says its system has proved the Exponential Time Hypothesis false, a result it says was subsequently verified in Lean.

These are quoted receipts, not a count of independent investigations. Several reports may rely on the same original source.

web.archive.org ↗
Surprisingly, our system has recently proved that the ETH is false, a result that was subsequently verified in Lean.
raw.githubusercontent.com ↗
We give a deterministic algorithm for exactly counting the satisfying assignments
raw.githubusercontent.com ↗
Our algorithm refutes the Exponential Time Hypothesis (ETH), which
web.archive.org ↗
Our AI found its natural language disproof for the ETH on October 16th, 2026
Open this check in the full collection →
The idea, illustrated03

Go inside the check.

    The full explanation appears when your reading access is confirmed.
    Next: Real upload, nameless tool

    The headline says patient data went to an AI tool. It did: 3,182 patients, one spreadsheet, one unapproved public chatbot. The part nobody in our record says is which tool.

    A clerk wipes a chalkboard clean while a faceless robot's back hatch still bulges with patient papers.
    AI illustration

    The headline matches what the hospital told reporters. The upload happened, the records were medical, and deletion rests on a request to a company no receipt names.

    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.

    1. At sign-up: 30 days free. Payment card required. One introductory trial per customer.
    2. 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.
    3. 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.

    By starting your trial, you agree to the Terms.

    Inferno Tech Pty Ltd t/a BS Killer, ABN 27 647 413 474

    Open the evidence3 source pages

    The claim we checked

    Johns Hopkins alerts 3,000+ patients after data sent to AI tool.

    These are quoted receipts, not a count of independent investigations. Several reports may rely on the same original source.

    wmar2news.com ↗
    a worker uploaded a spreadsheet to a public AI tool that the hospital had not approved
    baltimoresun.com ↗
    Johns Hopkins Bayview is notifying 3,182 patients after an employee uploaded personal health information to an unapproved public AI tool.
    cbsnews.com ↗
    an employee uploaded the spreadsheet to a public third-party AI platform not approved for use
    wmar2news.com ↗
    medical record numbers, account numbers, the department that treated them, medications given and the names of their providers
    Open this check in the full collection →
    The idea, illustrated04

    Go inside the check.

      The full explanation appears when your reading access is confirmed.
      The finish line

      Edition complete

      You’re up to speed.

      That’s the 27 Oct 2026 briefing. Keep the useful bits. Leave the noise.

      Reading estimate: 731 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.

      Free email updates. Unsubscribe any time.