tyler cosgrove

Chinese labs releasing competitive open-weight models (DeepSeek, Kimmy K3) are accused of distilling US frontier models; Kimmy K3's license demands profit-sharing from cloud providers, creating a geoeconomic flashpoint where US policy may ban Chinese open-source AI despite industry opposition.

Nasdaq's new fast-entry rules will force billions in passive index funds to buy SpaceX, OpenAI, and Anthropic shares immediately upon IPO, driving up prices and forcing sales of rival stocks, creating a structural tailwind for these listings.

Post-Durbin Amendment, big banks (JPM, BofA, Wells, PNC) explore buying Fiserve's Star/NYCE networks to replicate Capital One/Discover model — owning issuer, network, and volume. This pressures Visa/Mastercard and merchants, while creating tailwinds for abstraction layers (Ramp) that sit above network fragmentation.

Tesla's vision-only FSD and vertical integration create a structural lead that legacy OEMs cannot close despite data being theoretically scrapeable; regulatory capture (NHTSA) and cultural inertia prevent adoption of superior alternatives like comma.ai.

Tech companies are growing total opex despite layoffs because token bills ($5B) exceed labor savings ($2B); CEOs are now scrutinizing whether inference spend drives conversion, retention, and revenue or just 'token maxing' vanity metrics.

By releasing a 975B-parameter open-weights model (41B active) purpose-built for fine-tuning, Thinking Machines aligns product and business model — customers get weight ownership and portability, while Tinker API monetizes the integration and customization layer, a potential blueprint for sustainable open-source AI economics.

The shift to AI agents requiring longer context windows drives unprecedented demand for high-bandwidth memory, creating a supply shortage that benefits Micron as the leading US DRAM manufacturer with its new 1-alpha production.

Chinese authorities meeting with Alibaba, ByteDance, Z.ai, DeepSeek to discuss export controls on advanced models — replicating US restrictions on GPT-5/6, Fable, Mythos. Open-weight models already released (Hugging Face) are hard to claw back, but future frontier models could face pre-release security review or outright bans, raising costs for US AI users who rely on cheap Chinese open-source supply.