AI digest: Money talks, agents walk
Google drops £40B on Anthropic, OpenAI ships GPT-5.5 at double the price, and DeepSeek quietly undercuts everyone.
Big money moved this week. Google’s massive Anthropic bet shows how expensive the AI race has become, while pricing wars heat up around agent capabilities.
Google bets the farm on Anthropic with £40B investment
Google’s planning to invest up to £40B in Anthropic, mixing cash and compute credits as AI rivals scramble for massive capacity. This follows the limited release of Anthropic’s cybersecurity-focused Mythos model. The sum is absolutely staggering, suggesting Google sees Anthropic as a genuine counterweight to OpenAI rather than just another portfolio company.
OpenAI ships GPT-5.5 but doubles the API price
GPT-5.5 is here with OpenAI calling it a “new class of intelligence” designed for autonomous agent work. The model tops benchmarks but still hallucinates frequently, and API pricing jumped 20% over GPT-4. OpenAI’s chief scientist admits AI progress has been “surprisingly slow” but promises big leaps ahead. Charging more for incremental improvements feels like testing what the market will bear.
DeepSeek undercuts everyone with million-token models
Chinese lab DeepSeek released V4-Pro and V4-Flash, both handling million-token contexts at prices well below OpenAI, Google, and Anthropic. The timing is perfect as Western companies raise prices and cap usage for agent workloads. DeepSeek’s “good enough for almost nothing” approach could force a reckoning on AI pricing models.
GitNexus solves the codebase awareness problem
GitNexus hit 19,000 GitHub stars by giving Claude and Cursor proper structural understanding of codebases through MCP-native knowledge graphs. Every developer using AI coding assistants has hit this problem, agents editing code they don’t actually understand. Open source solving real developer pain while big labs fight over benchmarks feels about right.