Summary
Matt Wolfe rounds up the week in AI, led by Anthropic’s release of Claude Haiku 5.5. The model is built for fast, low-cost, high-volume work such as summaries, classification, database queries, and sub-agent tasks in coding. Benchmarks show it trailing Sonnet and Opus across the board, but at roughly 28 cents per task on extra-high effort, compared with about 93 cents for Sonnet 5.5 on medium, it competes on price. Artificial Analysis data places it below models such as Kimi K3 and GLM 5.3 on intelligence while sitting near the bottom on cost per task.
Anthropic also introduced live dashboards connected to sources like BigQuery, Databricks, Snowflake, and Salesforce, plus Claude Motion for animated explainers.
On the OpenAI side, the video covers a more permissive approve-for-me mode, a ChatGPT meetings plugin similar to Granola, and the new decisions API, which returns predicates, choices, and scores using GPT6 Luna. It also notes simplified API pricing tiers (Build, Launch, Grow) and an ultra-fast mode for GPT 6.1 Sol, along with updates from Grok and Google.
📺 Source: Matt Wolfe · Published October 09, 2026
🏷️ Format: Roundup
