• Hack AI
  • Posts
  • HackAI Newsletter — June 9, 2026 Draft

HackAI Newsletter — June 9, 2026 Draft

Edition #8 | Your 2-minute pulse on AI + Texas tech

Author

Bryce Miles
June 09, 2026

Missed our last update? Read the May 2026 newsletter →

🗓️ Upcoming Events

  • 🤖 HackAI Meetup — Monday, June 15th at Capital Factory — Our next community meetup is right around the corner. Come build, demo, network, and hang with the Austin AI community. Pizza and good vibes as always. RSVP here →

  • 🎓 Gauntlet AI — Cohort 6 Applications Open (Starts July 6 in Austin) — Our friends at Gauntlet AI are accepting applications for their 10-week AI engineering fellowship — 3 weeks remote, 7 weeks onsite in Austin building production-grade AI systems every week: RAG pipelines, multi-agent architectures, fine-tuning, and RL environments. No tuition. Travel, housing, food, and compute are fully covered (Austin locals get a paid downtown studio). Companies like Zapier, Carvana, and GoFundMe hire directly from Gauntlet based on what you build, not your résumé — roles at $200k+ across multiple cities and remote. Cohort 5 generated 246 interviews in a single week. If you're an experienced software engineer ready to go AI-first, this is the move. Apply here →

🌐 Big AI Headlines

💰 New: Anthropic Raises $65B at $965B — Files for IPO Days Later

Anthropic just had the most consequential two weeks in AI startup history. On May 28, the Claude maker closed a $65 billion Series H led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia — pushing its valuation to $965 billion and leapfrogging OpenAI ($852B) as the most valuable AI startup on Earth. Run-rate revenue crossed $47 billion earlier in May. Then on June 1, Anthropic filed a confidential S-1 with the SEC for an IPO. OpenAI is reportedly targeting September at ~$1 trillion. The biggest AI IPO race in history is officially on.

🧠 New: Claude Opus 4.8 Ships — Dynamic Workflows + Mythos "Coming in Weeks"

Anthropic dropped Claude Opus 4.8 on May 28 alongside the mega-raise — their strongest model yet for agentic coding, with ~4x fewer undetected code flaws than Opus 4.7. The big new feature: Dynamic Workflows in Claude Code, which lets Opus spin up hundreds of parallel sub-agents to tackle codebase-scale migrations. Fast mode is now 3x cheaper. And the buried lede: Anthropic said it expects to bring Mythos-class models to all customers "in the coming weeks" — the tightest public timeline yet for the most capable (and controversial) AI model ever built.

⚡ New: Google Drops Gemini 3.5 Flash at I/O — Plus Personal AI Agent "Spark"

Google I/O 2026 (May 19) was the most AI-dense keynote in the company's history. Gemini 3.5 Flash launched same-day — outperforming their own Gemini 3.1 Pro on coding and agentic benchmarks while running 4x faster than comparable frontier models. Pricing: $1.50/$9.00 per million tokens — significantly cheaper than GPT-5.5 or Claude Opus. The Gemini app now has 900 million+ monthly users. The bigger play: Gemini Spark, a 24/7 personal AI agent that runs on your behalf, rolling out to AI Ultra subscribers ($100/mo). Google is betting big that the future is agents, not chatbots.

⚔️ Update: Pentagon vs. Anthropic Gets Even Wilder — NSA Uses Mythos While DOD Bans Claude

Last time: The Pentagon signed AI deals with OpenAI, Google, Microsoft, and five others — pointedly excluding Anthropic. But Trump called Anthropic "high-IQ" and said a deal was "possible."

Now: This saga just reached peak absurdity. The Financial Times reported on June 5 that the NSA has ~6 Anthropic engineers embedded inside the agency, customizing Claude Mythos for offensive cyber operations — potentially targeting networks in China and Iran. This is happening while the same Department of Defense still officially bans Anthropic as a "supply chain risk." On June 5, Hegseth formally rejected Anthropic's plea to reconsider the designation. The DC Circuit appeals court heard oral arguments in late May, with judges appearing divided. One Pentagon, two realities.

⚖️ Update: Musk Loses OpenAI Trial — Jury Says He Waited Too Long

Last time: The Musk v. Altman trial kicked off April 28 in Oakland, with bombshell revelations including Musk admitting xAI uses OpenAI's models for training and Brockman's journal entries about being "warm to steal the nonprofit."

Verdict: On May 18, after less than two hours of deliberation, the jury ruled unanimously that Musk waited too long to sue — his claims fell outside a three-year statute of limitations. The court never reached the merits of whether Altman and Brockman actually breached their duties to OpenAI. Musk called it a "calendar technicality" and vowed to appeal to the 9th Circuit. The verdict clears a major overhang for OpenAI's planned IPO. Both companies are now racing toward the public markets — OpenAI targeting ~$1 trillion, Anthropic at $965 billion. The AI IPO wars are on.

🏛️ New: Congress Drops the "Great American AI Act" — 269 Pages of Federal AI Regulation

On June 4, Reps. Jay Obernolte (R-CA) and Lori Trahan (D-MA) released a discussion draft of the Great American Artificial Intelligence Act — the most comprehensive federal AI framework ever proposed. The 269-page bill targets frontier AI models and would freeze state AI development laws for three years, create mandatory safety evaluations, require independent auditing, and establish a new federal AI coordination office. The bill has bipartisan co-sponsors but is already drawing fire: unions called it "a giveaway to the AI industry," while Public Citizen says it "strips states' authority to protect consumers." We’ll continue to monitor the discussions and see what the actual bill looks like in the coming months.

Meanwhile, Colorado's AI Act takes effect June 30 — the first major state-level AI law. Federal vs. state: the clock is ticking.

🤠 Texas AI Spotlight

🏭 Update: Terafab Production Fab Lands in Grimes County — $119B Tax Deal Approved

Last time: Tesla confirmed Terafab would use Intel's 14A process, with SpaceX handling high-volume manufacturing. A public notice put SpaceX's initial investment at $55 billion and total potential cost at $119 billion.

Big update: On June 3, the Grimes County Commissioners Court approved a full property tax abatement for the Terafab production facility at the Gibbons Creek Reservoir — the former site of a coal power plant, roughly 90 miles northeast of Austin and 70 miles northwest of Houston. The site brings existing power infrastructure, water access, and transmission capacity. Musk defended the deal on X, arguing that employee and contractor taxes will "far exceed all revenue that Grimes County currently earns." The prototype R&D fab remains at GigaTexas in Travis County.

⚡ Update: ERCOT Approves "Batch Zero" — 450 GW in Queue, Record Summer Demand Forecast

Last time: Gas generation overtook wind in ERCOT's queue for the first time since 2016, with 360,000 MW of data center demand and total incoming demand at 410,000 MW.

Now: The numbers keep climbing. Data centers and large loads are now requesting 438–450 gigawatts of power in Texas — more than 5x the entire grid's all-time peak. On June 3, ERCOT's board approved "Batch Zero" — an entirely new process for evaluating which data center projects actually get grid access, prioritizing based on regional transmission capacity. The PUCT (Public Utility Commission of Texas) will vote on final approval June 18. New rules also require data centers to "ride through" brief power disruptions instead of tripping offline and cascading grid emergencies. And ERCOT is forecasting a record 92 GW of summer demand — smashing the previous 85.5 GW record from 2023. The AI boom is stress-testing the Texas grid in real time.

🚗 Update: Austin Robotaxi Wars Intensify — Tesla Goes Metro-Wide, Waymo Recalls 3,800

Last time: We covered Zoox coming to Austin and the growing robotaxi competition.

Big update: The Austin robotaxi scoreboard just got a major shake-up. On June 3, Tesla expanded its unsupervised Robotaxi service to the entire Austin metro — but the fleet is still tiny: roughly 20 active vehicles vs. Waymo's 300+ in Austin alone. New Texas DMV filings (from a new state oversight law) reveal Tesla has just 42 registered AVs statewide vs. Waymo's 577 and Avride's 317. Wait times can exceed 30 minutes. Meanwhile, Waymo recalled 3,791 robotaxis nationwide after one drove into San Antonio floodwaters on April 20 and was swept into Salado Creek (no injuries). The fix is an OTA software update. And Zoox has begun testing its purpose-built robotaxis in Austin. Four companies, one city, very different approaches.

🛡️ Update: Glasswing's First Report — 10,000+ Critical Vulns, Now 150+ Orgs Worldwide

Last time: Anthropic launched Project Glasswing with 12 partners and committed $100M+ in credits to secure the world's critical software using Claude Mythos Preview.

Big update: The May 22 progress report is sobering. In its first month, Mythos Preview and ~50 partners found more than 10,000 high- or critical-severity vulnerabilities in systemically important software. Cloudflare alone found 2,000 bugs (400 high/critical). Mozilla fixed 271 Firefox vulnerabilities — a 10x increase over previous Claude models. A critical wolfSSL flaw (CVE-2026-5194, CVSS 9.1) would have let attackers forge certificates and impersonate banks. Over 90% of flagged issues are confirmed true positives. On June 2, Anthropic expanded Glasswing to ~150 organizations in 15+ countries, including NATO, ENISA, and Australia's Signals Directorate. The bottleneck is now patching speed — only 75 of the first 530 disclosed high/critical bugs have been fixed.

🎓 Update: UT Austin's Horizon Supercomputer Approaches Launch — 300 Petaflops, 4,000 GPUs

Last time: The Horizon supercomputer (4,000+ NVIDIA Blackwell GPUs) was being installed at TACC and expected online this year, enabling UT to "train close to frontier-size models."

Now: The nation's largest academic supercomputer started production in Spring 2026 at Sabey Data Centers' facility in Round Rock — the first time TACC has operated off the UT campus. Horizon packs 300 petaflops of simulation power (10x Frontera) and a 100x AI improvement over the previous system, with 1 million CPU cores and 4,000 NVIDIA Grace Blackwell GPUs. Early projects will focus on Texas-specific challenges including extreme weather forecasting and medical breakthroughs. For context, this puts UT Austin's computing power on par with many commercial AI labs. The Austin-Round Rock corridor is becoming a supercomputing powerhouse.

🗣️ This Month's AI Debate

Should the federal government preempt state AI laws — or do states need to lead?

The Great American AI Act proposes freezing state AI development laws for three years while Congress builds a federal framework. Colorado's AI Act takes effect June 30 regardless. And Texas's TRAIGA has been in effect since January. The question is no longer hypothetical.

Last month we debated whether AI models should be vetted before release. This month, the question shifts to who gets to make the rules.

🏛️ Pro Federal Framework: A patchwork of 50 different state laws creates compliance nightmares for AI companies building one model that ships everywhere. The Great American AI Act's co-sponsors argue that "risks created by AI don't stop at state lines" and protections shouldn't "depend on your zip code." A unified national standard — with independent auditing and frontier model safety evaluations — gives companies certainty and gives Americans consistent protection. Three years is enough time to get it right.

🤠 Pro State Authority: When Congress takes 10+ years to pass tech legislation, states fill the vacuum — and that's working. Colorado, Texas, and the EU all moved faster than Congress on AI guardrails. Public Citizen calls the federal bill "a giveaway to the AI industry" that strips states' ability to protect consumers, workers, and children. If the federal framework is weak, a three-year preemption means three years of zero accountability. States are the laboratories of democracy — let them experiment.

💬 What do you think? Bring your hot takes to the next HackAI Meetup!

💬 Community CTA

Got an AI project, job posting, or demo to share? Reply to this email or message us on Luma — we'd love to feature you!

Stay curious, Austin 🤖💡

— The HackAI Team, Reid & Bryce