April 17th Newsletter

Author

Da Yu
April 17, 2026

As of close on April 16th, 2026

Oil plunged over 10% as Iran briefly reopened the Strait of Hormuz during an Israel-Lebanon ceasefire, lifting broader peace hopes, though ING warned 13 million barrels per day remain disrupted. The hope that the Middle East conflict may be nearing an end helped power a stock rally, alongside strong bank earnings and upbeat CEO comments, as the Nasdaq posted its longest winning streak since 2021. Domestically, U.S. jobless claims fell to 207,000, signaling labor resilience, but continuing claims rose to 1.818 million. Meanwhile, China is hardening its legal toolkit ahead of Trump’s May 14-15 Beijing visit. In tech and media, ShinyHunters claims it stole 78.6 million Rockstar records, though Rockstar says the breach was limited and non-material, while Snap, under activist pressure, will cut 1,000 jobs, close 300 openings, and target $500 million in savings. On Thursday, Live Nation was ruled as a Monopoly in a Victory for States, raising potential breakup pressures.

Top 5 AI Highlights

[1] 📈 Anthropic is briefing the Trump administration on the frontier model Mythos despite a Pentagon blacklist, highlighting national-security demand for advanced AI; separately, Vance and Bessent privately warned tech CEOs about cyber risks before Mythos’s launch. The company is also drawing investor offers valuing it near $800 billion, over double February’s $350 billion, as revenue growth to $30 billion intensifies competition with OpenAI, though funding terms remain uncertain. More from Reuters and CNBC on Mythos and cybersecurity, and Bloomberg on the company’s new valuation.

[2] 👾 OpenAI began a limited rollout of its GPT-5.4-Cyber model to trusted users for vulnerability detection in software, intensifying competition with Anthropic’s Mythos cybersecurity model amid rising concerns about AI-enabled hacking risks. The news comes as OpenAI has announced its acqui-hire of AI personal finance startup Hiro Finance; terms are undisclosed, and Hiro will shut down operations and delete user data in May. More details on GPT-5.4-Cyber and Hiro Finance

[3] 🤖 Meta is developing an AI clone of Mark Zuckerberg to mimic his tone and thinking, enhancing employee access and internal efficiency as part of broader AI-driven productivity and “superintelligence” ambitions. The company also announced a Broadcom partnership through 2029, committing 1GW+ of custom chips designed by Broadcom, while CEO Hock Tan announced that he was exiting the Meta board. The Guardian has info on AI Zuckerberg, and CNBC has more on the Broadcom partnership.

[4] 🍎 Apple privately threatened to remove Grok from its App Store over deepfake violations, forcing X to improve moderation, though compliance concerns persist despite updates: a new report shows that the model continues generating non-consensual sexual deepfakes on X despite safeguards, exposing enforcement gaps and sustaining regulatory, legal, and reputational risks for Elon Musk’s ecosystem. Read more on 9to5Mac and NBC News.

[5] 🤝 More AI partnerships: Stellantis and Microsoft signed a five-year deal spanning 100+ AI initiatives, cyber defense, and Azure modernization, while Cadence and Nvidia teamed on robotics AI to accelerate robot training; Cadence also launched a chip-layout AI agent on Google Cloud. More details on the Stellantis partnership and Cadence partnerships.

✨Open Source AI Development

[1] ⚛️ NVIDIA launched Ising, an open AI model family aimed at advancing quantum computing by improving processor calibration and error correction, delivering up to 2.5x faster and 3x more accurate performance. Details are on Nvidia.

[2] 🧰 AMD updated its GAIA open-source framework, enabling users to build custom AI agents via chat and offering simplified desktop installs across Windows, macOS, and Linux for running LLMs and LLM-based apps locally and privately. More from Phoronix.

[3] 🤖 Microsoft is testing an OpenClaw-like local agent for Microsoft 365 Copilot aimed at enterprises, emphasizing security and always-on task execution amid broader Copilot Cowork and cloud-based Copilot Tasks expansion. Details on TechCrunch

[4] 🗣️ VAANI opens a major speech dataset for low-resource languages, spanning 31,255 hours, 156,534 speakers, and 109 languages, including 59 absent from existing open-source speech datasets. See more on Hugging Face.

[5] 🧠 Linux kernel formalized governance for AI-assisted code, banning AI-generated Signed-off-by tags, requiring human accountability and Assisted-by disclosure—an important precedent for managing AI use in critical open-source infrastructure. More on Kernel.

Founder’s Corner

Range for SAFE Valuation Caps

Source: X

What We Read This Week

  • Watch: Arm’s CEO on Bloomberg “Arm’s AI Pivot: From Smartphones to the Cloud” (Bloomberg)

  • Cisco is reportedly negotiating to buy AI-agent security startup Astrix for $250 million-$350 million, highlighting surging demand for tools that monitor rogue agents and protect enterprises from fast-growing AI risks. (The Information)

  • Anthropic publicly opposed Illinois bill SB 3444, backed by OpenAI, that critics say would largely shield AI labs from liability for disasters caused by misuse; the dispute highlights widening regulatory tensions between major AI rivals. (Wired)

  • Anthropic is introducing optional identity verification for select Claude use cases via Persona, requiring government ID and selfie checks to prevent fraud or abuse; the move has raised privacy and surveillance concerns among users. (Engadget)

  • Google launched the Google app for desktop on Windows 10+, globally in English; it offers Alt+Space universal search for web, Drive, files, and apps, plus AI Mode and screen-sharing assistance. (9to5Google)

  • Allbirds shares plunged 31% after a prior 582% surge, as the company rebrands to “NewBird AI” and focuses on AI compute and GPUs-as-a-service. (WSJ)

  • Sam Altman’s San Francisco home was hit in a second attack within days, with two suspects arrested after an apparent shooting, escalating security concerns around AI executives. (The SF Standard).

  • Forbes reports defunct startups are selling Slack, email, and Jira archives as premium AI training data, creating shutdown cash for founders while raising major privacy, anonymization, and model-memorization concerns. (Forbes)

  • The CTO of Uber said heavy use of AI coding tools like Claude Code is rapidly increasing software costs while boosting productivity, with ~11% of backend code now AI-generated and rising concerns over escalating AI development budgets. (The Information)

  • AI is reshaping golf end-to-end: Smarter tee-time booking, swing coaching, pace-of-play optimization, and turf management promise better player experiences and leaner course operations. (WSJ).

  • AMC is releasing “The Audacity” premiere on TikTok in 21 three-minute clips to court younger viewers, while also airing the full episode on AMC, AMC+, and Samsung TV Plus. (TechCrunch)

AI Fundraising News (Apr 10 — Apr 16)

  • ActionAI: AI reliability testing company, raised a $10M seed

  • Alloy Therapeutics: AI drug discovery company, raised a $40M Series E

  • Artemis: AI security automation company, raised a $70M Series A

  • Balerion: AI mortgage origination company, raised a $6M seed

  • Bluefish: AI brand visibility management company, raised a $43M Series B

  • Capsule: AI security company, raised a $7M seed

  • Cookiy AI: AI customer research company, raised a $7M pre-seed

  • Donecle: AI aircraft inspection company, raised $11.8M

  • Factory: AI coding agents company, raised a $150M Series C

  • Gitar: AI code review automation company, raised $9M

  • Gizmo: AI study materials company, raised a $22M Series A

  • Gravity: AI analyst automation company, raised $7M

  • Helical: AI drug discovery workflow company, raised a $10M seed

  • Hilbert: AI enterprise decision intelligence company, raised a $28M Series A

  • Luminai: AI healthcare workflow automation company, raised a $38M Series B

  • Mintlify: AI software documentation company, raised a $45M Series B

  • Modern Relay: AI work coordination company, raised $3M

  • Nava: AI agent verification company, raised an $8.3M seed

  • nEye: AI data center networking company, raised an $80M Series C

  • Onto Health: AI fertility and longevity care company, raised a $20M Series A

  • Orbital: AI space-based data centers company, raised $1M

  • Outcraft AI: AI revenue execution company, raised a $2M pre-seed

  • Parasail: AI inference routing company, raised a $32M Series A

  • Pillar: AI financial risk automation company, raised a $20M seed

  • Prefix: AI maintenance coordination company, raised a $7.5M seed

  • Ralio: AI agent payments infrastructure company, raised a $2.5M seed

  • Replenit: AI retail customer behavior company, raised a $2.5M pre-seed

  • Resolve AI: AI software repair company, raised a $40M Series A extension

  • Round: AI finance workflow automation company, raised a $6M seed

  • Sapient Perception: AI drone imaging company, raised a $3.3M pre-seed

  • ShengShu: AI video generation company, raised $293M

  • Slash: AI financial services agent company, raised $100M

  • Solidroad: AI customer interaction evaluation company, raised a $25M Series A

  • Spektr: AI compliance infrastructure company, raised a $20M Series A

  • Sygaldry: AI quantum server company, raised a $138M Series A

  • Synera: AI engineering workflow automation company, raised a $40M Series B

  • TraqCheck: AI recruiting and screening company, raised an $8M Series A