Claude code insights, Spotify’s AI music in the wild, plus Anthropic’s $10B raise.
NVIDIA announced the Rubin platform yesterday, bringing six new chips for the next generation of AI compute. Builders using LLMs get a look at Claude's code generation insights. And in user-facing AI, Spotify is deploying AI-generated music, while nearly 40% of consumers now start searches with AI directly.
Anthropic is reportedly in talks to raise $10 billion, which would push its pre-investment valuation to $350 billion, nearly doubling it in four months. GIC and Coatue Management are expected to lead the round, adding to previous investments and a reported $15 billion commitment from Nvidia and Microsoft. The company, known for its Claude chatbot, reportedly expects to break even by 2028.
Read full story→OpenAI launched ChatGPT Health, a new service designed to link with user medical records and wellness apps. It operates in a dedicated space with a separate context boundary, using a connector layer via partners like b.well. The company clarifies it's not for diagnosis or treatment, setting clear policy boundaries for its use and explicitly stating non-training boundaries.
A new study finds that in a survey of 500 consumers who actively use AI tools, 37% now start searches directly with AI tools, bypassing traditional engines. They cite clearer, faster answers (60%) over ad clutter, though 85% still verify information.
Adam Wathan, CEO of Tailwind Labs, stated that AI caused "75% of the people on our engineering team lost their jobs," with "Traffic to our docs ... down about 40% vs early 2023" and "our revenue is down close to 80%." This happened despite Tailwind CSS's growing popularity, highlighting AI's disruptive effect on developer tools.
NVIDIA announced its Rubin AI platform, featuring six new chips. This hardware supports long-context and agentic workloads by treating the rack as the unit of compute, addressing power and reliability constraints. The linked deep dive is dated Jan 05, 2026.
Anthropic's Claude Code (Opus 4.5) autonomously planned tasks, wrote code, and deployed a working site, using task decomposition plus reusable 'skills' and sub-agent delegation. It achieved this by 'compacting' context windows and coordinating 'subagents' for specialized work.
Spotify's Smart Shuffle recommended an AI-generated song, prompting questions about platform transparency. User studies from Deezer/Ipsos show 80% want AI music labeled and 73% want to know if platforms recommend it.
AI's impact on software will vary significantly across eight distinct markets, not uniformly. The article predicts an explosion of new software in areas like internal tools (e.g., custom data dashboards) and personal software (e.g., AI companions), but less in enterprise SaaS (e.g., ERP systems) or safety-critical systems (e.g., medical devices), based on market constraints and demand elasticity. If this is true, expect shifts in pricing or distribution over 12 months in these areas.
Mercor founder Brendan Foody reveals their human-in-the-loop strategy for training frontier AI models: hiring experts like poets at $150/hour to build rubrics and evaluate outputs. This moves beyond raw text feedback towards structured evaluation using techniques like pairwise ranking, goldens, and calibration rounds, where expert evaluators create the rubric.