AI DAILY / DEV
FRIDAY
November 20, 2026
→

    Google Ships Gemini 4 Neon — $3 Input, Between Hydrogen and Carbon

    • Nov 18 launch via Gemini API, AI Studio, and the Gemini app; priced $3 / $15 per Mtok with a 1M-token context window.
    • Sits between Gemini 4 Hydrogen ($1.50) and the still-unshipped Gemini 4 Carbon as Google's default coding/reasoning tier.
    • Google-reported: 70.5% Terminal-Bench Science, 84.0% RiemannBench, 71.3% HLE-Diamond.
    • Opus 5.5 still edges it on pure Terminal-Bench coding; Neon leads on science and reasoning.
    • HN launch thread: 1,017 points, 761 comments.
    models blog.google

    Microsoft Ignite Drops Seven MAI Models, Led by MAI Code 2

    • Nov 17–18 event: MAI Code 2, MAI Code 2 Flash, MAI Cyber 2 Flash, MAI Image 2.7, MAI Realtime 1, Dragon Copilot Transcribe, Decision 1.1.
    • MAI Code 2 claimed at 98% of GPT-6.1 Sol quality for 40% of the price; MAI Cyber 2 Flash scores 92.3% on CyberGym-E2E vs 91.5% for GPT-6 Astra.
    • MAI Realtime 1 is Microsoft's first speech-to-speech model at $0.01 per minute.
    • Rolling out in GitHub Copilot, Foundry preview, and Dragon Copilot clinical workflows.
    • HN keynote thread: 594 points, 376 comments.
    models microsoft.ai

    Anthropic Closes +17% at $97.17 on Day One, ~$1.76T Fully Diluted

    • ANT opened $111.50 (+34%), hit $115.10 intraday, then gave back most of the pop to close $97.17.
    • Fully diluted valuation ~$1.76T; non-diluted market cap ~$1.27T.
    • After-hours slipped to $92.60 (-4.7%) as the Nasdaq fell 2.2% on the session.
    • Largest IPO on record at ~$81.3B raised; 30-day greenshoe for up to 147M more shares still open.
    • HN first-day threads: 923 and 986 points, 1,824 and 1,913 comments.
    industry fortune.com

    OpenAI and Broadcom Unveil Jalapeño, OpenAI's First Custom Inference Chip

    • Nov 19 announcement: reticle-sized inference ASIC co-designed with Broadcom in roughly nine months.
    • OpenAI says it used its own models to speed the RTL and verification loops during design.
    • First deployments late 2026, scaling through 2027 with Microsoft as a launch customer.
    • Decouples OpenAI's inference roadmap from Nvidia supply after a week of circular-financing debate.
    • HN thread: 220 points, 168 comments.
    industry openai.com

    Claude Code Skills Ecosystem Takes Over GitHub Trending

    • obra/superpowers added 2,028 stars in 24h (322K total) — agentic skills framework with a software-dev methodology.
    • elementalsouls/Claude-BugHunter added 1,349 stars (12.9K total) — Claude Code skill bundle for bug hunting.
    • mvanhorn/last30days-skill added 954 stars (70.9K total) — researches a topic across Reddit, X, YouTube, HN, Polymarket, web.
    • Three of the top eight trending repos today are Claude Code skills bundles — a real ecosystem signal, not a single viral repo.
    open-source github.com

    DeepSeek Raises ~$15.8B Second Round Backed by Tencent, CATL, Chinese Telcos

    • Nov 20 Digitimes report: CNY105.5B second-round close, roughly US$15.8B.
    • New backers include Tencent, CATL, and the three major Chinese state telcos.
    • Raise lands while DeepSeek tightens free-tier usage limits citing compute shortages.
    • Signals Beijing-aligned capital stepping in as US export controls continue to squeeze PRC lab access to Nvidia Rubin.
    industry digitimes.com