AI tool launches, product updates and industry moves.
On July 23, 2025, the Trump administration released its AI Action Plan and signed three executive orders: accelerate innovation, build infrastructure, export the American AI stack.
In July 2025, Google DeepMind's Gemini Deep Think and OpenAI's experimental model both hit gold-medal level at the IMO — LLMs matching gold medalists for the first time.
In July 2025 Windsurf lived the wildest 72 hours in startup history: after the OpenAI deal collapsed, Google paid $2.4B for the founders and Cognition bought whatever was left.
Moonshot AI unveils Kimi K2 with 1 trillion total parameters — the largest open model at launch — excelling at agentic and coding tasks.
xAI released Grok 4 in July 2025, setting records on GPQA and other hard benchmarks, plus a multi-agent Heavy edition.
In 2025 Waymo logged 100 million fully autonomous miles and tripled annual rides to 15 million — robotaxis went truly commercial.
In June 2025 Meta bought 49% of Scale AI for $14.3 billion and recruited its CEO — the most expensive shot in the AI talent war.
At its first Code with Claude developer conference, Anthropic unveiled Claude Opus 4 and Sonnet 4, built for hours-long autonomous coding.
At Microsoft Build 2025, GitHub introduced the Copilot coding agent — it takes tasks, runs in the background on GitHub Actions and opens PRs.
In May 2025, DeepMind unveiled the Gemini-powered coding agent AlphaEvolve, which found a 4x4 matrix multiplication method beating Strassen's algorithm.
Alibaba releases the Qwen 3 series; the flagship Qwen3-235B-A22B rivals top models like DeepSeek-R1 and o1 on major benchmarks.
OpenAI launches o3 and o4-mini, the first reasoning models to integrate web browsing, Python, and image analysis directly into their chain of thought.
In April 2025, Safe Superintelligence raised $2 billion at a $32 billion valuation — under a year old, zero product, zero revenue.
Meta released Llama 4 Scout and Maverick on April 5, 2025 — its first natively multimodal open-weight models, with Scout supporting a 10M-token context.
Cognition releases Devin 2.0 with a collaborative IDE and a starting price slashed from $500 to $20 per month, taking the AI software engineer mainstream.
Midjourney launched V7 on April 3, 2025, with clear gains in prompt understanding, image quality and hand details, plus default personalization.
Runway released Gen-4 on March 31, 2025, built on world consistency: a single reference image keeps characters stable across scenes.
On March 28, 2025, Musk announced xAI's all-stock acquisition of X — valuing xAI at $80 billion and X at $33 billion — merging data and distribution.
In March 2025, Anthropic published interpretability research that first 'traces' Claude's internal thought circuits, revealing how LLMs think and why they hallucinate.
On March 26, 2025, a federal judge rejected OpenAI's motion to dismiss, letting The New York Times's core copyright claims proceed — with Microsoft as co-defendant.
Google released Gemini 2.5 Pro in March 2025 with built-in reasoning, leading multiple benchmarks by meaningful margins.
At GTC 2025, Jensen Huang unveiled Blackwell Ultra (shipping H2 2025) and previewed the next-gen Vera Rubin architecture.
Butterfly Effect launched the general-purpose agent Manus on March 6, 2025; the invite-only rollout sparked a code-hunting frenzy with a 2M+ waitlist.
On February 26, 2025, Amazon unveiled the generative-AI Alexa+, free for Prime members — the biggest overhaul in the assistant's decade.