In a remarkable leap for artificial intelligence, Thinking Machines Lab recently introduced Inkling, a groundbreaking 975-billion-parameter multimodal Mixture-of-Experts (MoE) model. This model, equipped with a controllable thinking effort, quickly established itself as the most potent open-weights model in the Western world, with customization available through Tinker. However, this accolade was short-lived. Enter Moonshot's Kimi K3, which boasts a staggering 2.8 trillion Mixture-of-Experts parameters and utilizes Delta Attention with integrated vision capabilities. Hailing from China, K3 now holds the title for the largest open-weight model, propelling a $31.5 billion funding round. K3's impact was immediate, as it surged to the top of the Frontend Code Arena, displacing Fable 5 across six of seven domains. This innovation rivals Fable-class models, outperforming Opus 4.8 at Sonnet pricing, with full capabilities expected by July 27. In an intriguing development, these models are now designed to oversee one another. Gauntlet, an open-source pipeline comprising five adversarial reviewer personas, outperformed human evaluations in 15 out of 20 blind tests on architecture papers. Similarly, OpenAI's GPT-Red employs self-play to enhance GPT-5.6 Sol, reducing successful prompt injections to a mere 0.05%. Evidently, peer review is evolving beyond human intervention.
Simultaneously, substantial developments are occurring in the semiconductor industry. Taiwan Semiconductor Manufacturing Company (TSMC) reported an impressive 77.4% increase in quarterly profits and announced plans to boost its 2026 capital expenditure to $64 billion. Their commitment to the AI megatrend is evident. The U.S.-Taiwan partnership is intensifying, with $265 billion and ten fabs now in the pipeline, facilitated by a tariff-for-treasure agreement. Meanwhile, Apple is actively seeking to acquire chip companies as its M2 Ultras struggle and its Siri revamp relies on Nvidia's rented chips. On the horizon, quantum statistical plasmonic metacrystals, capable of filtering light through quantum coherence at room temperature, offer promising advancements in solar energy. The fusion industry, too, is thriving, securing a record $4.48 billion in annual funding, a 69% increase, with over 16,000 employees in the sector. However, not everyone is enthusiastic about these technological advancements. New York's pioneering data center moratorium prompted a presidential directive to reverse the decision "IMMEDIATELY," while the governor advocated for equitable benefits for communities driving AI advancements.
The embodiment of artificial intelligence is generating political ripples. Nvidia has unveiled Cosmos 3 Edge, a comprehensive world model for robots, and allied with Fujitsu, Hitachi, and Kawasaki to form a Japanese physical-AI coalition. Jensen Huang, Nvidia's CEO, has dubbed the physical world as AI's "next frontier." This frontier is not without conflict. Hyundai's Ulsan factory workers went on strike due to the introduction of a humanoid robot named Atlas, marking the first instance where robotic labor halted car production. Meanwhile, nominal collaborators Waymo and Uber are engaging in lobbying disputes over the rapidly advancing robotaxi sector.
At the user interface level, artificial intelligence is becoming more integrated into daily life. NotebookLM has evolved into Gemini Notebook, now seamlessly connected to cloud computers capable of writing and executing code. Meta has initiated a feature to alert parents when a teenager's AI chat suggests self-harm, with future plans for direct emergency service contact. China, taking a stricter stance, has banned AI companions from fostering emotional dependency, affecting millions of users. In an unexpected turn, xAI pursued legal action against one of its users for attempting to generate illicit images, while addressing a privacy scandal by deleting collected data and open-sourcing 844,000 lines of code from Grok Build. DeepMind and Isomorphic Labs have introduced a bioresilience initiative, leveraging SynthID watermarking technology and directing AlphaFold and AlphaEvolve towards early outbreak detection.
Economically, artificial intelligence is having a transformative effect. Visa has opened stablecoin minting to 15,000 banks and 200 million merchants. South Korea is launching a nationwide chatbot for its 52 million citizens, and India has introduced its second AI unicorn in as many months with the vibe-coding startup Emergent. As AI labs absorb the entrepreneurial class, with 105 Y Combinator alumni now employed at OpenAI or Anthropic, Anthropic is courting banks for an IPO and has launched Ode with Anthropic to bring Claude to the mid-market. San Francisco landlords are capitalizing on the AI boom, with one lease renewal demanding a 0.25% stake in any startup founded on the premises. However, the friction is palpable. Publishers have taken legal action against Google for using books to train Gemini, and the first evaluation of leading models has highlighted concerns over potential