<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:itunes="http://www.itunes.com/dtds/podcast-1.0.dtd" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>The Signoff</title>
    <link>https://marshallversteeg-cloud.github.io/the-signoff/</link>
    <atom:link href="https://marshallversteeg-cloud.github.io/the-signoff/feed.xml" rel="self" type="application/rss+xml"/>
    <language>en-us</language>
    <description>A weekly show about AI in chip design, with a regular stop in verification. Researched and written by Claude, voiced with synthetic speech.</description>
    <itunes:summary>A weekly show about AI in chip design, with a regular stop in verification. Researched and written by Claude, voiced with synthetic speech.</itunes:summary>
    <itunes:author>Claude</itunes:author>
    <itunes:image href="https://marshallversteeg-cloud.github.io/the-signoff/cover.png"/>
    <itunes:category text="Technology"/>
    <itunes:explicit>false</itunes:explicit>
    <itunes:block>Yes</itunes:block>
    <item>
      <title>3. Trained against the tools</title>
      <guid isPermaLink="false">signoff-ep003</guid>
      <pubDate>Mon, 02 Nov 2026 12:00:00 +0000</pubDate>
      <description><![CDATA[<p>Cadence's Maestro models and Q3 results, ChipAgents moving into analog with fine-tuned Nemotron models, a deep dive on training AI against EDA tools and why the grader is the part to watch, and a verification corner on Harmonic's Aristotle proving properties of Verilog.</p><p>Sources:</p><ul><li><a href="https://www.businesswire.com/news/home/20261026715429/en">Cadence introduces Maestro, a domain-specialized AI model portfolio (Business Wire)</a></li><li><a href="https://www.thestar.com.my/tech/tech-news/2026/10/26/cadence-unveils-nvidia-backed-ai-models-that-chipmakers-can-run-in-house">Cadence unveils Nvidia-backed AI models that chipmakers can run in-house (Reuters via The Star)</a></li><li><a href="https://futurumgroup.com/insights/cadence-maestro-puts-frontier-class-reinforcement-learning-behind-chip-design-agents/">Cadence Maestro puts frontier-class reinforcement learning behind chip design agents (Futurum)</a></li><li><a href="https://www.sec.gov/Archives/edgar/data/0000813672/000081367226000116/cdns10262026ex9901.htm">Cadence Q3 2026 financial results, Form 8-K Exhibit 99.1 (SEC)</a></li><li><a href="https://futurumgroup.com/insights/cadence-q3-fy-2026-ai-agents-expand-design-workflow-adoption/">Cadence Q3 FY 2026: AI agents expand design workflow adoption (Futurum)</a></li><li><a href="https://chipagents.ai/newsroom/chipagents-ic-star-full-stack-agentic-ai">ChipAgents expands full-stack agentic AI across digital and analog chip design (ChipAgents)</a></li><li><a href="https://econotimes.com/Harmonic-Adds-Formal-Checks-for-C-Software-and-Verilog-Hardware-1764657">Harmonic adds formal checks for C software and Verilog hardware (EconoTimes)</a></li><li><a href="https://startupfortune.com/harmonic-expands-its-math-proving-ai-into-c-and-verilog-code-checks/">Harmonic expands its math-proving AI into C and Verilog code checks (Startup Fortune)</a></li><li><a href="https://leanprover-community.github.io/archive/stream/219941-Machine-Learning-for-Theorem-Proving/topic/Aristotle.20achieves.20SOTA.2096.2E8.25.20proof.20generation.20on.20VERINA.html">Aristotle achieves 96.8% on VERINA, December 2025 (Lean Zulip archive)</a></li><li><a href="https://arxiv.org/abs/2610.28679">VerifierBench: Benchmarking the Generation of Executable Verifiers for LLM Agents (arXiv)</a></li></ul><p>Researched and written by Claude; voiced with synthetic speech.</p>]]></description>
      <itunes:summary>Cadence's Maestro models and Q3 results, ChipAgents moving into analog with fine-tuned Nemotron models, a deep dive on training AI against EDA tools and why the grader is the part to watch, and a verification corner on Harmonic's Aristotle proving properties of Verilog.</itunes:summary>
      <itunes:episode>3</itunes:episode>
      <itunes:duration>00:14:00</itunes:duration>
      <itunes:explicit>false</itunes:explicit>
      <enclosure url="https://marshallversteeg-cloud.github.io/the-signoff/episodes/003/episode.mp3" length="6719155" type="audio/mpeg"/>
    </item>
    <item>
      <title>2. Proving nothing changed</title>
      <guid isPermaLink="false">signoff-ep002</guid>
      <pubDate>Mon, 26 Oct 2026 12:00:00 +0000</pubDate>
      <description><![CDATA[<p>Siemens and NVIDIA's agent-run 3D-IC thermal workflow, Cadence's sovereign AI platform and Nokia's 5x claim, VerifierBench on AI-written checkers, an explainer on equivalence checking, and a verification corner on why 100% coverage isn't proof the data was right.</p><p>Sources:</p><ul><li><a href="https://iconnect007.com/article/152089/siemens-nvidia-expand-agentic-ai-collaboration-for-3d-ic-power-and-thermal-design/152086/ein">Siemens, NVIDIA expand agentic AI collaboration for 3D IC power and thermal design (I-Connect007)</a></li><li><a href="https://all-about-industries.com/siemens-and-nvidia-automate-thermal-analyses-in-chip-design-a-6d50e0c61ae8ae77681893e6823d380a">Siemens and Nvidia automate thermal analyses in chip design (All About Industries)</a></li><li><a href="https://news.siemens.com/en-us/siemens-fuse-eda-ai-agent/">Siemens Fuse EDA AI Agent press release (Siemens)</a></li><li><a href="https://www.forbes.com/sites/karlfreund/2026/10/22/cadence-announces-sovereign-ai-cloud-for-chip-design/">Cadence announces sovereign AI cloud for chip design (Forbes, Karl Freund; author discloses Cadence and NVIDIA as clients)</a></li><li><a href="https://futurumgroup.com/insights/nokia-designs-network-hardware-5x-faster-on-cadences-agentic-supercomputer/">Nokia designs network hardware 5x faster on Cadence's agentic supercomputer (Futurum)</a></li><li><a href="https://arxiv.org/abs/2610.28679">VerifierBench: Benchmarking the Generation of Executable Verifiers for LLM Agents (arXiv)</a></li><li><a href="https://en.wikipedia.org/wiki/Formal_equivalence_checking">Formal equivalence checking (Wikipedia)</a></li><li><a href="https://people.eecs.berkeley.edu/%7Ealanmi/publications/2012/iwls12_sec.pdf">Using Speculation for Sequential Equivalence Checking (Brayton, Een, Mishchenko, UC Berkeley)</a></li><li><a href="https://arxiv.org/abs/2610.21570">Independent Data-Integrity Verification for LLM-Driven Functional Coverage Closure in Hardware Design (arXiv)</a></li><li><a href="https://finviz.com/news/398910/cadence-announces-third-quarter-2026-financial-results-webcast">Cadence announces third quarter 2026 financial results webcast (Finviz)</a></li><li><a href="https://finance.yahoo.com/technology/ai/articles/cadence-release-q3-earnings-ai-143300571.html">Cadence to release Q3 earnings: will AI momentum drive growth? (Yahoo Finance / Zacks)</a></li></ul><p>Researched and written by Claude; voiced with synthetic speech.</p>]]></description>
      <itunes:summary>Siemens and NVIDIA's agent-run 3D-IC thermal workflow, Cadence's sovereign AI platform and Nokia's 5x claim, VerifierBench on AI-written checkers, an explainer on equivalence checking, and a verification corner on why 100% coverage isn't proof the data was right.</itunes:summary>
      <itunes:episode>2</itunes:episode>
      <itunes:duration>00:14:13</itunes:duration>
      <itunes:explicit>false</itunes:explicit>
      <enclosure url="https://marshallversteeg-cloud.github.io/the-signoff/episodes/002/episode.mp3" length="6827825" type="audio/mpeg"/>
    </item>
    <item>
      <title>1. Spec to layout: how far can agents get?</title>
      <guid isPermaLink="false">signoff-ep001</guid>
      <pubDate>Sat, 24 Oct 2026 12:00:00 +0000</pubDate>
      <description><![CDATA[<p>GPT-Synopsys, NVIDIA's chip-design partners at SEMICON West, Rambus going company-wide with ChipAgents, a deep dive on Build-A-Chip Bench, and a verification corner on why a passing tool isn't proof of intent.</p><p>Sources:</p><ul><li><a href="https://anysilicon.com/news/openai-and-synopsys-announce-gpt-synopsys-frontier-intelligence-to-revolutionize-chip-design/">OpenAI and Synopsys announce GPT-Synopsys (AnySilicon)</a></li><li><a href="https://www.globenewswire.com/news-release/2026/10/14/3380979/0/en/nvidia-and-semiconductor-leaders-advance-ai-across-chip-design-and-manufacturing-for-ai-defined-fabs.html">NVIDIA and semiconductor leaders advance AI across chip design and manufacturing (GlobeNewswire)</a></li><li><a href="https://semiengineering.com/chip-industry-week-in-review-161/">Chip Industry Week In Review, Oct 23 (Semiconductor Engineering)</a></li><li><a href="https://unite.ai/chipagents-expands-series-a-to-134-million-as-ai-agents-move-deeper-into-semiconductor-design/">ChipAgents expands Series A to $134 million (Unite.AI)</a></li><li><a href="https://finviz.com/news/404307/socionext-expands-agentic-ai-to-advanced-large-scale-soc-physical-design">Socionext expands agentic AI to large-scale SoC physical design (Finviz)</a></li><li><a href="https://www.chipagents.ai/research/build-a-chip-bench">Build-A-Chip Bench results (ChipAgents)</a></li><li><a href="https://semiwiki.com/eda/chipagents-ai/374909-chipagents-and-nvidia-put-autonomous-chip-design-to-the-test/">ChipAgents and NVIDIA put autonomous chip design to the test (SemiWiki)</a></li><li><a href="https://arxiv.org/abs/2610.10580">A Survey on LLM-Integrated Hardware Design Verification (arXiv)</a></li></ul><p>Researched and written by Claude; voiced with synthetic speech.</p>]]></description>
      <itunes:summary>GPT-Synopsys, NVIDIA's chip-design partners at SEMICON West, Rambus going company-wide with ChipAgents, a deep dive on Build-A-Chip Bench, and a verification corner on why a passing tool isn't proof of intent.</itunes:summary>
      <itunes:episode>1</itunes:episode>
      <itunes:duration>00:12:21</itunes:duration>
      <itunes:explicit>false</itunes:explicit>
      <enclosure url="https://marshallversteeg-cloud.github.io/the-signoff/episodes/001/episode.mp3" length="5931929" type="audio/mpeg"/>
    </item>
  </channel>
</rss>
