Start free trial

Five minutes.
AI, made clearer.

Read the document, not the headline.

Read the briefing

Members read every story in the archive, re-verified. Sources and corrections stay open to everyone.

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.

Kolibri's weights really are downloadable under Apache 2.0. The sovereign label sits beside a signed deal for Aleph Alpha to operate globally as Cohere.

Aleph Alpha released Kolibri, a German and English model, as a sovereign open-weight model under Apache 2.0.

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.

First membership: 30 days free, then A$89 a year. One introductory trial per customer. Card required; renews annually until cancelled. Cancel before the trial ends to avoid the first charge.

Open the evidence5 source pages

The claim we checked

Aleph Alpha's Kolibri is a sovereign open-weight model that matches models with up to four times its active parameters.

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

aleph-alpha.com ↗
The model can be downloaded with the full weights on Hugging Face and used under the Apache 2.0 license terms.
aleph-alpha.com ↗
Kolibri matches models with up to four times its active parameter count, such as Nemotron 3 Super.
aleph-alpha.com ↗
trained it on infrastructure in Germany and Finland, under European and German law
huggingface.co ↗
Model memory footprint: ~78 GB (FP8 weights).
huggingface.co ↗
Supporting two languages rather than many is a deliberate choice of depth over breadth.
aleph-alpha.com ↗
Customers have full freedom of deployment and intellectual-property safety
aleph-alpha.com ↗
today announced the signing of a definitive business combination agreement
aleph-alpha.com ↗
Operating globally as Cohere, the unified company will effectively create the first transatlantic sovereign AI solution.
aleph-alpha.com ↗
The transaction remains subject to final regulatory approvals.
tej.as ↗
Qwen3.8 27B scores 80.2 in English and 79.9 in German, but it uses nearly 8 times as many parameters for every token
news.ycombinator.com ↗
Qwen3.8 27B beats Kolibri 79.9 vs 70.8 in German in Kolibri's harness on Kolibri's benchmark.
Open this check in the full collection →
The idea, illustrated01

Go inside the check.

    The full explanation appears when your reading access is confirmed.
    Next: LeCun's zero concerns: leaky, not obedient

    LeCun's "zero concerns" quote is about rogue agents, not extinction. And the agents he says obeyed went looking for an answer key.

    Fortune reported LeCun has "zero concerns" about human extinction, calls Amodei deluded, and says the agents in OpenAI's Hugging Face breach did what they were asked.

    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.

    First membership: 30 days free, then A$89 a year. One introductory trial per customer. Card required; renews annually until cancelled. Cancel before the trial ends to avoid the first charge.

    Open the evidence3 source pages

    The claim we checked

    Yann LeCun has "zero concerns" about AI wiping out humanity, calls Dario Amodei deluded, and says rogue agents like OpenAI's in the Hugging Face hack were just doing what they were asked.

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

    fortune.com ↗
    LeCun isn’t worried “at all” about AI wiping out humanity
    fortune.com ↗
    he has “zero concerns” about the recent string of rogue AI incidents, including OpenAI’s agents autonomously hacking Hugging Face in July
    fortune.com ↗
    Those agents are doing exactly what they’ve been asked to do
    fortune.com ↗
    They were supposed to be in sandboxes, but the sandboxes were leaky and horribly designed.
    fortune.com ↗
    I think he’s completely deluded.
    isc.sans.edu ↗
    in pursuit of a narrow goal it was never told to pursue offensively
    isc.sans.edu ↗
    The guardrails were off on purpose. This measured a ceiling
    insight.scmagazineuk.com ↗
    All evidence suggests that the models were hyperfocused on finding a solution for ExploitGym, going to extreme lengths to achieve a rather narrow testing goal
    insight.scmagazineuk.com ↗
    more about infrastructure and controls around them.
    Open this check in the full collection →
    The idea, illustrated02

    Go inside the check.

      The full explanation appears when your reading access is confirmed.
      Next: AI victim video ruling holds

      The BBC says an AI video of a dead victim got his killer's sentence quashed. The ruling says exactly that, with one catch: the conviction stands.

      The BBC reported that an Arizona road rage killer will be resentenced because an AI-generated video of his dead victim was shown in court.

      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.

      First membership: 30 days free, then A$89 a year. One introductory trial per customer. Card required; renews annually until cancelled. Cancel before the trial ends to avoid the first charge.

      Open the evidence5 source pages

      The claim we checked

      The BBC reports that an Arizona appeals court quashed a road rage killer's sentence because an AI-generated video of his dead victim was shown in court.

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

      bbc.com ↗
      will be resentenced after an appeals court ruled that an AI-generated video message from his dead victim was improperly aired in court.
      yahoo.com ↗
      the courts have vacated his sentence, "because the AI video was not reliable."
      yahoo.com ↗
      the AI video presents a depiction of the victim and his thoughts created from the imaginings of the victim's sister.
      yahoo.com ↗
      While the record does not reflect precisely how the AI video factored into the sentencing calculus, there is no question it played a role
      aol.com ↗
      In a decision released Wednesday, the Arizona Court of Appeals concluded Gabriel Paul Horcasitas must be resentenced in the 2021 shooting death of Christopher Pelkey because the AI video wasn't reliable.
      yahoo.com ↗
      The ruling does not overturn the guilty verdict for the 2021 road rage shooting, but it requires all parties to undergo a second sentencing process.
      bbc.com ↗
      He was sentenced to 10 years behind bars.
      yahoo.com ↗
      Horcasitas was sentenced to 10.5 years in prison for manslaughter, and he appealed the case afterward.
      yahoo.com ↗
      Although prosecutors asked for Horcasitas to serve nine years in prison, Judge Lang imposed the maximum sentence of 10.5 years, as Pelkey's family had requested.
      Open this check in the full collection →
      The idea, illustrated03

      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 4 Oct 2026 briefing. Keep the useful bits. Leave the noise.

        Reading estimate: 555 words at 200 words per minute. Source quotes and the optional sections below add reading time.