🤖 HackAI Newsletter — February 2026

Missed our last updates? Read the Late January newsletter → or the January 6th newsletter →

🗓️ Upcoming Events

• 🚀 AutoHDR Hackathon — This Friday! — Huge opportunity for AI devs! Austin's fastest-growing AI startup, AutoHDR, is hosting a hackathon awarding $8K in cash prizes along with up to three potential $120K–$180K+ job offers to winners. Bring your friends and build together all weekend — ML experience is not required, just be full-stack and an exceptional problem solver. Sign up here →

• 🎤 HackAI x Atlassian Meetup — Wednesday, March 11 — A special collab event! Mark your calendars! Sign up here →

• 🖥️ Texas AI House — Friday, March 13 - A HackAI <> AITX collab event. Join us for a full day summit exploring Texas’s role in the AI revolution. Sign up here →

• 🎶 SXSW 2026 — March 12–18 — AI is the #1 topic this year, with 250+ sessions covering runaway AI risks, human-centric design, and AI ethics. Note: SXSW is using the new pop-up SXSW Village while the Convention Center is under construction.

🌐 Big AI Headlines

🧠 New: The February Model Rush — 10+ Major Releases in 3 Weeks

This may be the most concentrated month of major AI model launches ever. Here are the biggest:

🚀 New: OpenAI Ships GPT-5.3-Codex (Feb 5) OpenAI's most capable agentic coding model yet — combining the Codex and GPT-5 training stacks. ~25% faster than GPT-5.2, new SOTA on SWE-Bench Pro across 4 languages. OpenAI flagged it as the first model that could potentially automate cyberattacks, gating full API access behind a trusted-access program. A lighter variant, Codex-Spark (Feb 12), runs on Cerebras chips (not NVIDIA!) at 1,000+ tokens/sec — the first product of OpenAI's $10B+ Cerebras partnership.

Last time: We covered GPT-5.2-Codex as their most capable coding model. This is a major leap forward.

(Side note from Bryce: GPT-5.3-Codex is the first time I’ve felt like Claude Code has had major competition. I’d definitely recommend giving it a try sometime)

⚡ New: Anthropic Drops Claude Opus 4.6 & Sonnet 4.6 (Feb 5 / Feb 17) Opus 4.6 launched the same day as GPT-5.3-Codex — Anthropic reportedly moved their release up by 15 minutes to beat OpenAI. Built for deep financial research and complex analysis. Sonnet 4.6 followed on Feb 17, approaching Opus-class intelligence at Sonnet pricing. Anthropic also launched Claude Cowork (agentic work mode) and 11 open-source plugins for legal, sales, finance, and more — the legal plugin triggered a market earthquake (more below). Plus: Anthropic donated its Model Context Protocol (MCP) to the Linux Foundation, and OpenAI, Microsoft, and Google have all adopted it as the emerging "USB-C for AI agents" standard.

🔥 New: Google Gemini 3.1 Pro (Feb 19) Launched today and early results are stunning: 77.1% on ARC-AGI-2 (more than double Gemini 3 Pro), 94.3% on GPQA Diamond, and Artificial Analysis ranks it #1 most powerful AI model. Google also shipped Deep Think updates, Computer Use tools, and Veo 3.1 video improvements — all at unchanged pricing.

Last time: We covered Gemini 3 Flash going live at $0.50/million tokens. Google keeps accelerating.

🤖 New: xAI's Grok 4.20 Beta (Feb 17) Not a single model but a multi-agent system — four specialized AI agents (Grok, Harper, Benjamin, Lucas) that think in parallel and debate each other. It's reportedly the only profitable AI model on stock-trading benchmarks (10–34% returns). Available to SuperGrok and X Premium+ subscribers.

🌐 New: Mistral 3 Family (Late Jan / Early Feb) Massive open-source release: Mistral Large 3 (675B params, Apache 2.0), Ministral 3 models, Voxtral Transcribe 2 for on-device speech-to-text in 13 languages, and Mistral OCR 3 for document processing.

🇨🇳 New: China's Spring Festival AI Offensive A wave of Chinese models with narrowing capability gaps. Highlights: Zhipu AI's GLM-5 — a 744B open-source model trained entirely on Huawei chips (zero NVIDIA hardware). Alibaba's Qwen 3.5 with 397B params supporting 201 languages. Moonshot's Kimi K2.5 can coordinate 100 sub-agents at just $0.60/M input tokens. DeepSeek expanded context from 128K to 1M tokens, with V4 expected imminently.

💥 New: The "SaaSpocalypse"

Anthropic's Claude Cowork legal plugin triggered one of the most dramatic market reactions in recent tech history. On February 3 ("Black Tuesday for Software"), the software sector collapsed 13% in a single session — two S&P software ETFs shed $300.6B in combined market value. Thomson Reuters fell ~16%. The fear: foundation model companies packaging domain-specific workflow tools could undercut the entire specialized SaaS ecosystem. Through mid-Feb, Palantir dropped ~22% YTD, Adobe ~27%, Salesforce ~25-30%.

AI-driven layoffs are accelerating: Salesforce cut ~1,000 jobs, Meta Reality Labs cut ~1,500, Baker McKenzie eliminated 600-1,000 support staff, and Autodesk cut ~1,000. In total, 107 tech companies have conducted layoffs in 2026, impacting ~40,000 workers.

💰 Update: Record-Breaking AI Funding

Last time: We covered xAI's $20B raise and Anthropic's $10B term sheet.

The numbers since then are staggering:

• OpenAI is closing a ~$100B+ round at $850B+ valuation — the largest private funding round in history. Amazon (~$50B), SoftBank, NVIDIA, and Microsoft participating. IPO possible Q4 2026.

• Anthropic raised $30B in a Series G (Feb 12) at $380B valuation — the second-largest venture deal ever. Total funding now ~$64B.

• SpaceX–xAI merged (Feb 2) — creating the most valuable private company on Earth at $1.25 trillion. Six of xAI's 12 co-founders have since departed.

• Waymo raised $16B (Feb 2) at $126B valuation

• World Labs raised $1B (Feb 18) for spatial AI from Fei-Fei Li's lab

• Runway raised $315M (Feb 10) at $5.3B for AI video generation

🏛️ New: Pentagon vs. Anthropic Standoff

At a February 18 summit, the Pentagon demanded "all lawful use cases" from its four major AI providers (Anthropic, OpenAI, Google, xAI). Anthropic — the only holdout — wants guarantees its models won't be used for autonomous weapons or mass surveillance of Americans. The Pentagon threatened to label Anthropic a "supply chain risk" — a designation typically reserved for foreign adversaries. One to watch closely.

📜 Update: AI Policy Surge Continues

Last time: We covered Trump's executive order targeting state AI laws and TRAIGA facing federal pushback.

The FY2026 spending package (passed Feb 3) included multiple AI provisions: $40M for the Defense Innovation Unit, FCC briefings on AI robocalls, and OMB guidance for AI-ready datasets. The AI Overwatch Act advanced in Congress, giving 30-day review authority over AI chip export licenses to China. In a first, the Trump administration directly opposed Utah's AI Transparency Act — a Republican-sponsored bill — signaling real friction between federal and state AI regulation.

The EU Commission missed its February 2 deadline for AI Act high-risk system guidelines (now pushed to March/April). The EU also charged Meta with antitrust violations for excluding third-party AI from WhatsApp — potential fines up to 10% of global revenue. A $3 billion copyright lawsuit was filed against Anthropic over 20,000+ copyrighted songs.

🤠 Texas AI Spotlight

🚗 Update: Tesla Robotaxis — Cybercab Production Begins + 14 Crashes Last time: Tesla went fully driverless in Austin with FSD v13 — no human safety monitors. Big updates: On Feb 17, the first Cybercab rolled off the Gigafactory Texas line — a purpose-built vehicle with no pedals or steering wheel, priced at $30,000. Full production starts April 2026. However, Tesla robotaxis have been involved in 14 reported crashes in Austin, drawing regulatory scrutiny.

🚕 Update: Waymo Expands Austin + $16B Raise Last time: We discussed the vision-only vs. LiDAR debate. Waymo expanded its Austin service area to 140 square miles (up from 90), reaching Manchaca, more of east Austin, and north of US-183. They raised a massive $16B round and are deploying 6th-gen Ojai robotaxis. Dallas, Houston, and San Antonio expansions planned for 2026.

⚖️ Update: Texas AI Law + Austin's TRUST Act Last time: TRAIGA took effect Jan 1 and faced federal pushback. Austin City Council unanimously passed the TRUST Act (Feb 5), creating a new surveillance technology framework — after twice postponing votes on AI-powered cameras in parks. Meanwhile, a Texas Tribune investigation (Feb 19) found AI in political ads remains a "free-for-all" after a disclosure bill stalled in the Senate.

🏗️ Update: Texas Data Center Boom Raises Water Alarms Last time: Stargate was building a 1.2 GW facility in Milam County and Texas was projected to become the #1 US data center market. Now: A second Stargate site was announced in Shackelford County (combined ~7 GW planned capacity, $400B+ investment). Apple announced a $500B US investment plan including an AI data center in Houston. But: Texas data centers consumed 25 billion gallons of water in 2025, projected to hit 161 billion by 2030. Hays County officials are pushing back on a proposed data center that could use 70,000 gallons/day.

🎓 Update: UT Austin's Horizon Supercomputer Progress Last time: UT leads the nation in NSF research funding at $176M+. The Horizon supercomputer — the nation's largest academic supercomputer (4,000+ NVIDIA Blackwell GPUs, funded by a $457M NSF award) — is being installed at TACC and expected online in 2026. UT also secured $20M in renewed funding for its NSF AI Institute for Foundations of Machine Learning.

💡 New: Samsung Taylor Fab 90% Complete Samsung Austin Semiconductor celebrated its 30th anniversary, with the $17B+ Taylor fab now 90% done and on track for late 2026 operations producing 2nm chips for AI and automotive. Total planned investment across Taylor and Austin: ~$40 billion.

🚀 New: Austin AI Startup Raises Austin-based Circuit raised $30M (Feb 18) for its AI manufacturing workflow platform, founded by former Silicon Labs CEO Tyson Tuttle. webAI is developing private AI that runs on phones rather than data centers, targeting healthcare and defense.

💬 HackAI Community

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

Stay curious, Austin 🤖💡

— The HackAI Team, Reid & Bryce