Tech • IA • Crypto
Thinking Machines’ First AI Model, California Loses $3.2B to Texas, TSMC Adds $100B | Diet TBPN
7/10TL;DR
A new open-weights AI model from Thinking Machines Lab intensifies competition in the global AI race while highlighting tensions over openness, distillation, and geopolitics.
KEY POINTS
Thinking Machines launches “Inkling”
Thinking Machines Lab, led by former OpenAI CTO Mira Murati, released its first model, Inkling, as an open-weights system that allows users to modify and fine-tune it. The model contains 975 billion parameters, with roughly 41 billion active at a time, positioning it among the larger open models. It is designed as a flexible, general-purpose foundation model rather than claiming top performance in any single benchmark.
Strategy built around customization
Inkling is closely tied to the company’s Tinker API, which focuses on fine-tuning and enterprise integration. The business model mirrors the “Red Hat” approach, offering openness while monetizing services and support. This allows clients to retain control over model weights while relying on Thinking Machines for optimization and deployment.
Performance and positioning
Early benchmarks place Inkling between models such as Kimmy K2.5 and K2.6, outperforming alternatives like Neotron 3 Ultra. While not considered the strongest overall model, it is viewed as a competitive open-weight option outside China, particularly for coding and customization tasks.
Debate over distillation
The release has reignited debate around AI distillation, the practice of training models on outputs from other systems. While some claimed Inkling avoided distillation entirely, disclosures indicate limited use of synthetic data from open models during fine-tuning. This places it in a gray area, reflecting broader industry ambiguity over what qualifies as distillation.
Global competition and security concerns
The launch comes amid rising geopolitical tension in AI. U.S. firms and policymakers remain cautious about Chinese open-source models, citing security and alignment concerns. At the same time, Chinese labs have been accused of aggressively distilling Western models, with companies reportedly shutting down millions of suspicious accounts per week linked to such activity.
Shifting access to AI models
Reports that Beijing may restrict overseas access to its most advanced models add urgency to Western open-weight efforts. Inkling’s timing suggests a strategic push to ensure alternatives remain available globally, particularly for enterprises seeking non-Chinese solutions.
Creative limitations and design patterns
Separate analysis of newer AI systems highlights how models often develop recognizable stylistic patterns, especially in web design outputs. Efforts to improve quality have focused on removing disliked patterns—such as “bento box” layouts or oversized typography—rather than fundamentally increasing creativity, raising questions about long-term differentiation.
Industrial policy and competition in the U.S.
Beyond AI, competition for major technology projects remains intense. A $3.2 billion automated shipyard project from defense startup Seronic chose Texas over California, citing faster approvals and $211 million in tax incentives. The decision underscores how regulatory speed and incentives are shaping industrial investment.
Semiconductor investment surge
TSMC reported strong earnings and expanded capital expenditure plans, including an additional $100 billion investment in U.S. facilities. Despite this, markets reacted cautiously, with the NASDAQ dipping 1%, reflecting investor concern over potential overspending even amid strong demand for AI infrastructure.
Media and creator economy shifts
In digital media, brand partnerships are evolving. A new deal between Lexus and creators Colin and Samir involves multi-episode sponsorships rather than traditional ads, signaling deeper integration between brands and content creators.
Reassessing corporate value
Analysis of Disney’s $129 billion acquisition spree—including Marvel, Pixar, and Fox—shows the company’s current valuation near $169 billion, prompting debate over whether those deals created value. However, tens of billions returned to shareholders through dividends and buybacks complicate that assessment.
CONCLUSION
The release of Inkling highlights a pivotal shift toward open, customizable AI systems amid intensifying global competition, regulatory divergence, and unresolved questions about how models are built and controlled.