<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>The USASI weekly</title>
    <link>https://unitedstatesofamericasuperintelligence.com/newsletter/</link>
    <atom:link href="https://unitedstatesofamericasuperintelligence.com/newsletter/feed.xml" rel="self" type="application/rss+xml"/>
    <description>Each issue of the USASI weekly email, archived on the site as it was sent. Independent project. Not a United States government website.</description>
    <language>en-us</language>
    <lastBuildDate>Sun, 11 Oct 2026 12:00:00 GMT</lastBuildDate>
    <item>
      <title>Issue 1: open decision models, EmbeddingGemma 2, and a learning center</title>
      <link>https://unitedstatesofamericasuperintelligence.com/newsletter/issue-1/</link>
      <guid isPermaLink="true">https://unitedstatesofamericasuperintelligence.com/newsletter/issue-1/</guid>
      <pubDate>Sun, 11 Oct 2026 12:00:00 GMT</pubDate>
      <description>What changed in the USASI catalog from October 4 to 11, 2026. Every item links to its sources. Open models: Decision models. Three items this week cover models that return an answer or a score instead of generating text: AWS's Strands Decider 2B under Apache 2.0, Liquid AI's open-weight d1 models for text, images, and speech, and support for decision models in the llama.cpp server. EmbeddingGemma 2. Google released a 740-million-parameter embedding model for text, code, images, video, and audio under Apache 2.0. It now has its own record, and Transformers 5.19.0 added support for it. AstaBrief 8B. Ai2 open-sourced a model that turns a research question and retrieved literature excerpts into a cited report, along with its training data. The model card lists Apache 2.0. Tool releases: vLLM 0.31.0: a preload command that keeps weights in GPU memory across engine restarts. Diffusers 0.41.0: a Qwen-Image 2.1 pipeline; ONNX support is deprecated in favor of Optimum. MLflow 3.17.0: fine-grained permissions for shared resources. SQL-backed servers need a coordinated database upgrade. TRL 1.15.0: a fused LM head, on by default in six trainers. ONNX Runtime 1.31.0: stable model-package and EPContext data APIs. Supported native builds now default to telemetry, which can be turned off. Codex CLI 0.161.0 makes GPT-6.1 Sol the default model, and 0.162.0 adds managed Git worktree tools. Hosted models and policy: Claude Haiku 5.5: Anthropic announced a hosted model for high-volume tasks such as summaries and classification. The announcement does not describe a release of model weights. Anthropic's Usage Policy update takes effect November 12, 2026. It changes the company's own rules for use of its models; it is not a government rule. Anthropic's Genesis Mission commitments: Claude, Claude Code, and API credits over three years for research projects in the federal Genesis Mission, as described in Anthropic's own announcement. xAI's developer documentation now uses the name SpaceXSI,</description>
    </item>
  </channel>
</rss>
