NEWSROOM

Inside the newsroom

What we are reading, what Cyclotron recommends, and what became a story. We keep a running list of sources. A decision model called Cyclotron estimates which ones we will want, and we make the final call.

01 / Latest story

The newest story.

Our most recent piece of writing.

02 / Recently accepted

What we accepted lately.

Sources that passed our review most recently, newest first. Each card says whether an editor or the system accepted it.

04 / Highest confidence

What Cyclotron was surest about.

Accepted sources, ranked by how confident Cyclotron was that we would accept them. Confidence is its estimate, not a verdict.

05 / Recommended, not yet accepted

Waiting for our review.

Sources Cyclotron expects us to accept. Nobody has decided yet, and we may decide otherwise.

06 / Conversations

What we are discussing.

Comments on public sources and published stories, most recently active first. Authors appear under a public handle or as the Editor.

  • Reference: AI model security controls: why knowing the rules isn't enough

    AI model security controls: why knowing the rules isn't enough

    1 message

    See also: https://donellameadows.org/archives/leverage-points-places-to-intervene-in-a-system/

    by Editor

  • Reference: Brian Letort, Ph.D. — Global Head of Data and AI

    Brian Letort, Ph.D. — Global Head of Data and AI

    1 message

    See also: https://donellameadows.org/archives/leverage-points-places-to-intervene-in-a-system/

    by Editor

  • Reference: Pi Is the Last Harness

    Pi Is the Last Harness

    1 message

    I’m not sure if I agree that people will stop trying to reinvent coding agents and will agree to just adopt Pi as a standard. It is really good, though.

    by Editor

  • Reference: Memdoor: A team of 5 Qwen3.8-27B models beats Fable 5 on LiveCodeBench Hard

    Memdoor: A team of 5 Qwen3.8-27B models beats Fable 5 on LiveCodeBench Hard

    1 message

    A local coding agent using open-weight models that allegedly performs well.

    by Editor

  • Reference: Building an AI-native engineering org

    Building an AI-native engineering org

    1 message

    “ Faster coding shifted the bottleneck to planning and review.”

    by Editor

  • Reference: One reviewer per layer: splitting code review by the way code is read

    One reviewer per layer: splitting code review by the way code is read

    2 messages

    Reply.

    by Editor