These Last Two Weeks in AI (6/30 - 7/11)

wanglersteven
Jul 12
4 min read

Updated: Jul 20

Editor’s note: Sorry for missing last week—I was busy enjoying hot dogs, deep dish pizza, lakes, and NASCAR for the Fourth of July!

Steven posing with Dale Jr. — How cool is that?

TL;DR

OpenAI delays its open-weight model amid safety and reliability concerns (especially after xAI’s issues). GPT-5 rumors persist, but expect incremental rather than breakthrough improvements.
Meta and OpenAI are battling for AI talent, with record pay packages, rising culture concerns, and a likely shift in how tech leaders retain teams.
Google acquires Windsurf’s AI coding tech for $2.4B, after OpenAI’s failed bid. The dev tooling arms race is heating up.
AI coding assistants may slow expert devs, per new METR data, but most see value with the right workflow. Results are still up for debate.
Intel’s RealSense division spins out with $50M for robotics and edge AI ambitions.
OpenAI alumni launch Thinking Machines Lab with $2B+ in funding—a new heavyweight contender enters the research ring.
Big developer platform moves: Google Gemini’s Batch Mode (50% cheaper tokens), Amazon Bedrock’s long-term API keys, AWS’s 600Gbps C8gn instances, LangGraph v0.3’s orchestration upgrade.
xAI seeks up to $200B in new funding—an ambitious ask as it trails market leaders.
EU GPAI Code drops (voluntary—for now), but AI Act deadlines (August 2, 2025) are locked in. Compliance automation is essential.

Model Landscape Shake-Ups

OpenAI Hits the Brakes—Again

OpenAI has postponed its first open-weight model release indefinitely for more safety testing, following recent public issues (notably with xAI). This highlights a broader industry shift toward reliability and responsible deployment. (Business Insider, TechCrunch)

GPT-5: Incremental or Transformative?

Industry insiders expect a July launch window for GPT-5, with larger context windows and reduced hallucinations. Many expect more evolution than revolution; the real advances may come from tools and integrations, not core models. (Medium)

AI Talent Wars: Meta vs. OpenAI

Meta’s Superintelligence Lab Blitz: Meta is (still) offering unprecedented compensation—some packages range from $100M to $300M over four years—to poach top AI talent from Apple, OpenAI, Anthropic, and more. Notably, Ruoming Pang reportedly received a $200M+ deal. (SFGATE, Reuters)
OpenAI Responds: Sam Altman has called these huge offers "mercenary" moves, claiming “missionaries will beat mercenaries.” OpenAI is also recruiting heavily from rivals like Tesla, xAI, and Meta.
Culture & Retention Risks: The pay arms race is creating internal friction. Michael Dell has warned of resentment; Reid Hoffman says big offers are rational but retention depends on mission and culture. (MarketWatch)

This level of competition signals a critical inflection point—either a breakthrough era or a coming reckoning.

New & Noteworthy AI News

Google’s $2.4B Windsurf Deal

Google struck a $2.4B non-exclusive license deal for Windsurf’s AI coding tech, absorbing Varun Mohan’s team into DeepMind. OpenAI’s $3B bid reportedly fell apart after Microsoft flagged product maturity concerns. Why Google took the risk: a sign of the escalating race for AI dev tools, talent, and IP. (TechCrunch, WSJ)

AI Coding Tools: Help or Hindrance?

A METR study found Copilot and Cursor lengthened task times by 19% for experienced devs, mainly due to prompt lag and confusion in large codebases. Not everyone buys these results—if AI tools slow you down, you’re either an ultra-elite dev (unlikely) or you’re not leveraging them effectively. For most, the boost is real if workflows are set up right. Either way: product polish and integration still trump pure model hype. (TechCrunch)

RealSense Spins Out from Intel

Intel spun off RealSense (AI vision and robotics) with a $50M Series A led by Intel Capital and MediaTek. With 60% of the global robotics depth-camera market, RealSense aims to drive edge AI further. (Tom’s Hardware)

OpenAI Alumni Launch “Thinking Machines Lab”

OpenAI CTO Mira Murati and cofounder John Schulman launched Thinking Machines Lab, already backed by $2B+ from a16z, drawing talent from OpenAI, Anthropic, and Meta. (Wikipedia)

Why It Matters

Trend	Why You Should Care
AI dev tools arms race	IP and talent deals are accelerating—watch for more M&A and licensing moves.
Product/UX over raw performance	AI coding tools are not a cure-all—benchmark and tailor for your own dev teams.
Edge AI expansion	Real-time perception is fueling smarter, more autonomous production systems.
Next-gen research labs	Veteran-led new labs could disrupt the landscape with faster model and tool innovation.

Funding & Hardware Hype

xAI seeks up to $200B in new funding—even as it lags front-runners in capability and adoption. Elon Musk claims the company isn’t short on cash, but this ambitious ask reflects a flood of capital chasing AI hardware. With xAI still in catch-up mode, investors are watching closely. (Reuters)

Bottom line: Expect GPU/ASIC supply to stay tight and on-prem inference costs to stay volatile.

Governance & Compliance

The European Commission published its General-Purpose AI Code of Practice on July 10, a voluntary blueprint on transparency, copyright, and systemic risk. (Reuters, EU Digital Strategy)
Despite lobbying, Brussels confirmed AI Act deadlines—GPAI obligations kick in August 2, 2025. (Reuters)

Final Thoughts

AI is obviously evolving rapidly, but the real advances are now in practical tools, infrastructure, and compliance—not just model size. Stay adaptive, keep your standards high, and watch the details.

✌️Steven