Skip to main content

The news desk

AI news, checked against the source.

What changed in AI, one story at a time. Each one links to its primary source, and facts stay apart from interpretation.

Last updated ·

Sunday, November 8UTC

  1. New
    OpenAIAgents SDK 0.24.0-alpha.7Developer

    OpenAI’s Agents SDK alpha adds stored sessions and stricter executor defaults

    OpenAI’s Python Agents SDK 0.24.0-alpha.7 prerelease adds application-owned conversation storage, voice and hosted-tool paths for native agents, and a stricter default write policy for new executor environments. It also changes some approval-mode names. These are developer migration considerations in an alpha, not a new ChatGPT feature or a generally available product.

    OpenAI1 source

Free, email only

Unlock the full brief free.

  • Thirty days of stories to browse, not seven
  • On every story: what to check, each source and why it matters
  • The same signup unlocks every free tool

Continuing subscribes the email on that account to Work Worth Doing, my newsletter, the same as the form below, and unlocks every free tool in this browser. We only receive your email address. Unsubscribe in one click.

This is not an account: there is no password, and the unlock is a cookie in this browser. You also join Work Worth Doing, Demetri Panici's newsletter, every week: what I built and what changed in AI. We'll email you a link to confirm, and you can unsubscribe in one click. The same signup unlocks every free tool on the site. How your email is handled.

Already subscribed? Enter the same email to unlock this browser. You won't be signed up twice.

Friday, November 6UTC

  1. New
    OpenAIBioBasicsBenchDeveloper

    OpenAI introduces a benchmark for agents reading raw lab evidence

    OpenAI and Ligo developed BioBasicsBench with 100 routine wet-lab tasks using raw images, instrument outputs and records. In OpenAI’s own evaluation, GPT-6 Astra averaged 66.7% of available credit at xhigh effort, and the article documents several failures. This measures a narrow research setting; it does not establish that agents can run a laboratory reliably.

    OpenAI1 source

Monday, November 2UTC

  1. New
    NVIDIANVIDIA RecoveryDeveloper

    NVIDIA releases an open-source service to recover interrupted AI training jobs

    NVIDIA described NVIDIA Recovery on November 2 as an open-source, workload-level service for GPU clusters. It detects interruptions from logs, metrics and scheduler signals, diagnoses likely causes and can coordinate job restarts from checkpoints. New deployments begin in monitor mode, and automated actions require operator configuration; the article’s operational measurements come from NVIDIA’s own cluster.

    NVIDIA1 source

  2. New
    NVIDIANVIDIA AttributionDeveloper

    NVIDIA open-sources failure diagnosis for distributed AI workloads

    NVIDIA said on November 2 that NVIDIA Attribution is open source. It correlates workload logs, metrics and scheduler state into a structured diagnosis and evidence trail, with a service API and reference blueprint. Its claimed production scale and failure examples are internal NVIDIA experience rather than an independent benchmark.

    NVIDIA1 source

  3. New
    NVIDIAAI Cluster Runtime recipesDeveloper

    NVIDIA publishes early Kubernetes recipes for Vera Rubin clusters

    NVIDIA’s November 2 technical post describes AI Cluster Runtime recipes that generate deployment bundles and validate Kubernetes configurations for upcoming Vera Rubin NVL72 systems. The post cites tests on early hardware; some driver support is targeted for future releases, so the recipe is an early operator preparation aid, not evidence that all Rubin systems are generally available.

    NVIDIA1 source

  4. New
    NVIDIANVIDIA Workload MapDeveloper

    NVIDIA introduces a common workload map for AI Kubernetes frameworks

    NVIDIA described Workload Map on November 2 as a reusable description of distributed Kubernetes workload structure. It ships definitions for several training and inference frameworks so schedulers and management tools can inspect them without separate parsing adapters. This is a developer infrastructure tool for teams operating AI clusters, not a new model or consumer app.

    NVIDIA1 source

Days are UTC. Some sources report an exact time and others only a date; both are placed on their UTC day.