Signal SentrySearchUpdated Nov 19, 17:30 UTCPM Drop
Play the News · Drop #096 · Evening

AI Connections

How it works: 16 terms from this Drop's news make 4 groups of 4. You can make 4 mistakes. About 3 minutes. Your streak lives only on this device.

Show the groups
  1. Names in AI proof work

    Theorem · Isabelle · I/Q · I/R

    Anthropic worked with Theorem on its proofs; Amazon Science says I/Q and I/R connect an agent to the Isabelle proof assistant.

    From the Drop: Anthropic: Claude proved fuzz targets in 123 open-source libraries never crash · AWS releases tools that let AI agents write and check Isabelle proofs over MCP

  2. Levers a gateway gives admins

    Deny rules · Approvals · Model aliases · Cost centers

    LangSmith Tool Gateway applies deny and approval policies; AUMOVIO says its gateway uses model aliases and charges spend to cost centers.

    From the Drop: LangSmith Tool Gateway puts agent tool credentials, deny rules and approvals in one layer · AUMOVIO routes all model access through one Databricks gateway with Entra ID sign-in

  3. Numbers from coding-agent preprints

    33% · 117 · 36 · 38–82%

    Up to 33% fewer GPU-hours from 36 models tested; 117 of 200 replays repeated an action; 38–82% of successful runs were right on the first edit.

    From the Drop: Preprint: trying a small open model first cut coding-agent GPU-hours up to 33% · Preprint: crashed agent tool calls repeated in 117 of 200 replays without a full runtime · Preprint: coding agents kept working after a correct patch, and 1–8% then undid it

  4. Anthropic IPO figures

    $83 · $220 billion · $17 billion · $900 billion

    TechCrunch says Anthropic priced at $83 a share, expects about $220 billion in cash, had $17 billion in Q3 revenue and $900 billion in commitments.

    From the Drop: Anthropic's IPO leaves it expected to hold about $220 billion in cash, TechCrunch says

Games are written by AI from this Drop's sourced items. Scenarios are illustrative.