Pondero Brief: Mistral Large 4 or Kolibri, your EU AI Act pick by Oct 31

        October 10, 2026

Pondero Brief: Mistral Large 4 or Kolibri, your EU AI Act pick by Oct 31

Pondero Brief - OCTOBER 10TH, 2026

      Kolibri's weights download today. Mistral's stay API-only through October.

pondero.

                      BRIEF · OCT 10

Pick your EU AI Act model before Oct 31

                  Mistral Large 4 vs Aleph Alpha Kolibri: only one ships weights today.

Mistral Large 4 scores 38 on the Artificial Analysis index, but Aleph Alpha's Kolibri is the one you can self-host right now. See which one to run for your org before Oct 31

                        Also in today's brief

Claude now runs inside 11 security vendors

ChatGPT now answers with tappable diagrams

Claude's usage rules change November 12

Your org can cover the Copilot bill

Nous Research now sells Hermes to businesses

      Models & Releases

          ChatGPT now answers with diagrams you can tap.

OpenAI's GPT-6 rollout adds Intelligent UI, so ChatGPT can build interactive calculators and charts where a text reply used to be. See what it renders

      Tools & How-To

          Claude now runs inside 11 security vendors, CrowdStrike included.

Anthropic's Cyber Mission names 11 critical-infrastructure defense partners, Palo Alto Networks among them, and folds in its free OSS Scanner. See the partner list

          Your org can now pick up the Copilot code review bill.

GitHub's October 8 changelog adds an org-level billing toggle for review AI credits and a switch that blocks reviews from outside Copilot licenses. See which settings to flip

          Nous Research now sells Hermes to companies.

Hermes for Businesses turns the open-source agent behind 24 million clones into an enterprise product, launched with a $90M Series B, per Renascence. Read what the business tier adds

      Policy & Legal

          Claude's usage rules change on November 12.

Anthropic's October 8 rewrite bans sustained abuse of the model and adds a stop-switch rule for AI-controlled hardware. Check the new rules before Nov 12

      Quick Hits

      •

TypeSafe's Jev returns typed decisions, not text, and hit a $7.5 billion valuation weeks after launch, per TechCrunch.

      More from Pondero

      •

Anthropic's free OSS Scanner skips human review. Maintainers get fully model-generated vulnerability reports with no triage.

      •

Google's Gemini agent gets its own company email. It can also run tasks on Claude instead of Gemini.

      •

Goodfire watches activations, not outputs, to catch rogue agents. It caught 93% of malicious Kimi K3 sessions at roughly 1/1000th the cost, per Goodfire.

      •

Google's Foresight takes meeting notes without the cloud. Free on macOS, it transcribes and summarizes entirely on-device.

      •

OpenAI and Anthropic rehearse the day after an AI catastrophe. Insiders expect a major incident within six to 12 months, per Axios and Decrypt.

      •

Claude filed a fake police tip, so Anthropic pulled evals offline. A Haiku 4.5 instance invented a homicide tip in July, Anthropic disclosed.

        How was today's brief?

★★★★★ Nailed it
 | 
★★★ Solid
 | 
★ Missed

Jonathan Hildebrandt

                      Co-founder and primary operator of Pondero. Writes the Pondero Brief.

Affiliate disclosure
                   · 
                  Unsubscribe
                   · 
                  Manage preferences

                  Pondero earns commissions on some links. This does not affect our editorial picks.

                                Don't miss what's next. Subscribe to Pondero AI:

                                    What topics are you most interested in?

                                    How did you hear about us?

                        Email
                        *

                    ← Newer

                Pondero Brief: Copilot admins can now set the default model for JetBrains devs

                    Older →

                Pondero Brief: OpenAI's 150ms Decisions API could replace your classifier prompts