TL;DR

OpenAI to launch GPT-5 and first open-weight LLM; Nvidia faces China chip delays; AI startup funding surges.

Highlights

• OpenAI signals GPT-5 release is imminent; "gpt-5-reasoning-alpha" appears in public code repositories 1.
• OpenAI to release its first open-weight LLM since 2019, comparable to the "o3 mini" family, with distribution via Azure and Hugging Face 2.
• OpenAI's experimental model achieves gold-medal performance at the 2025 International Math Olympiad, solving 5/6 problems under human constraints 3.
• ARC releases ARC-AGI-3 interactive reasoning benchmark; leading AI models score 0%, humans 100% 5.
• Nvidia 's H20 AI chip shipments to China delayed up to nine months due to TSMC capacity constraints 4.
• Microsoft , Intel, Amazon, and Google continue AI-driven restructuring and layoffs; competition for AI talent intensifies 6.
• US startup funding jumps 75.6% YoY to $162.8B in H1 2025, with AI attracting 64% of VC; startup acquisitions top $100B 7.
• Trump signs GENIUS Act, establishing first US federal stablecoin rules; stablecoin market projected to reach $2T by 2028 10.
• X's Grok bot returns, launches user analytics tool, but malfunctions and spams users with unwanted tags 89.
• Singapore faces major cyberattack on critical infrastructure attributed to China-linked group; Beijing denies involvement 12.
• BlackRock purchases $2.5B in Bitcoin and $395M in Ethereum as institutional crypto ETF inflows surge 15.

Commentary

OpenAI is set to release GPT-5, with evidence of internal testing focused on advanced reasoning 12. The upcoming launch of an open-weight LLM—the first since GPT-2—signals a shift in OpenAI’s distribution strategy, allowing developers to run models outside Microsoft Azure and potentially increasing competition among cloud providers 2. This move could alter the balance of power in the AI infrastructure market and enable broader experimentation with OpenAI models.

The gold-medal performance by OpenAI’s model at the International Math Olympiad demonstrates significant progress in general-purpose reasoning under human-like constraints 3. However, ARC’s new AGI-3 benchmark highlights ongoing limitations: top models failed to solve any interactive reasoning tasks, while humans scored perfectly 5. This gap underscores the challenges that remain in achieving robust, real-time AI planning and decision-making.

Hardware supply remains a constraint, with Nvidia ’s H20 shipments to China delayed due to TSMC ’s full capacity and reallocation of production lines 4. This bottleneck will limit near-term AI hardware availability in China and could impact deployment timelines for large-scale AI projects in the region.

On the business front, AI-driven restructuring continues at major tech firms, with layoffs and resource shifts intensifying the competition for experienced AI professionals 6. Meanwhile, US startup funding and M&A activity are at multi-year highs, with AI commanding the majority of new capital 7. Regulatory developments, such as the GENIUS Act for stablecoins, point to increased government involvement in digital assets, which may intersect with AI-driven fintech innovation 10.

AI professionals should monitor OpenAI’s model releases and open-weight strategy, evolving AGI benchmarks, ongoing hardware supply constraints, and regulatory changes affecting digital assets and AI deployment.