<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Latent Minds Institute · papers and notes</title>
    <link>https://latentmindsinstitute.com/</link>
    <atom:link href="https://latentmindsinstitute.com/rss.xml" rel="self" type="application/rss+xml"/>
    <description>Research from an independent frontier AI research lab studying the hidden representations, computations, cognitive structures, and deployment dynamics of advanced AI systems.</description>
    <language>en</language>
    <lastBuildDate>Sat, 11 Jul 2026 08:44:29 GMT</lastBuildDate>
    <item>
      <title>Underwriting the Machine: A Field Guide to GPU Credit Risk</title>
      <link>https://latentmindsinstitute.com/papers/gpu-credit-underwriting/</link>
      <guid isPermaLink="true">https://latentmindsinstitute.com/papers/gpu-credit-underwriting/</guid>
      <pubDate>Sat, 01 Aug 2026 00:00:00 GMT</pubDate>
      <description>Working paper · economics strand, How should lenders price credit risk on GPU-collateralised debt, and what does that market structure imply for compute dependence? Epistemic status: Analytical framework with illustrative models · not investment advice.</description>
    </item>
    <item>
      <title>Latent Observatory: SAE Feature Explorer</title>
      <link>https://latentmindsinstitute.com/instruments/latent-observatory/</link>
      <guid isPermaLink="true">https://latentmindsinstitute.com/instruments/latent-observatory/</guid>
      <pubDate>Sat, 11 Jul 2026 00:00:00 GMT</pubDate>
      <description>Interactive instrument, What has a sparse autoencoder actually learned about a model, and does the auto-interp story survive contact with the activation evidence? Epistemic status: Live model data via the open Neuronpedia API; explanations labelled as hypotheses.</description>
    </item>
    <item>
      <title>Evaluation State in Language Models</title>
      <link>https://latentmindsinstitute.com/papers/evaluation-state/</link>
      <guid isPermaLink="true">https://latentmindsinstitute.com/papers/evaluation-state/</guid>
      <pubDate>Fri, 10 Jul 2026 00:00:00 GMT</pubDate>
      <description>Working paper, What evidence would show that a model represents evaluation as a shared internal state rather than reacting to surface cues? Epistemic status: Research programme · five hypotheses, experiments proposed, none run.</description>
    </item>
    <item>
      <title>Transformer Explainer: GPT-2 Live in the Browser</title>
      <link>https://latentmindsinstitute.com/instruments/transformer-explainer/</link>
      <guid isPermaLink="true">https://latentmindsinstitute.com/instruments/transformer-explainer/</guid>
      <pubDate>Fri, 10 Jul 2026 00:00:00 GMT</pubDate>
      <description>Interactive instrument · adaptation, What does a transformer's forward pass actually compute, step by step, on real input? Epistemic status: Adapted instrument · original by Cho et al., Georgia Tech Polo Club.</description>
    </item>
    <item>
      <title>Model Entrenchment: Why Useful AI Systems Become Difficult to Remove</title>
      <link>https://latentmindsinstitute.com/papers/model-entrenchment/</link>
      <guid isPermaLink="true">https://latentmindsinstitute.com/papers/model-entrenchment/</guid>
      <pubDate>Wed, 01 Jul 2026 00:00:00 GMT</pubDate>
      <description>Working paper, When an AI system resists removal, where does the resistance live: in the model's internal representations and behaviour, or in the web of dependence around it? Epistemic status: Conceptual framework · experiments proposed, none run.</description>
    </item>
    <item>
      <title>The Interpretability Map</title>
      <link>https://latentmindsinstitute.com/instruments/interpretability-map/</link>
      <guid isPermaLink="true">https://latentmindsinstitute.com/instruments/interpretability-map/</guid>
      <pubDate>Wed, 01 Jul 2026 00:00:00 GMT</pubDate>
      <description>Research map · interactive instrument, What has mechanistic interpretability actually established, in what order, with what dependencies, and what should a new researcher do this week? Epistemic status: Literature synthesis · sources cited per node.</description>
    </item>
    <item>
      <title>A Forward Market for GPU Compute</title>
      <link>https://latentmindsinstitute.com/papers/gpu-forward-market/</link>
      <guid isPermaLink="true">https://latentmindsinstitute.com/papers/gpu-forward-market/</guid>
      <pubDate>Wed, 01 Jul 2026 00:00:00 GMT</pubDate>
      <description>Working paper · economics strand, What would a functioning forward market for GPU compute look like, and what would it change about the economics of AI deployment? Epistemic status: Market-design proposal · conceptual.</description>
    </item>
  </channel>
</rss>
