Claude
The AI for problem solvers. Built by Anthropic to be safe, accurate, and secure. Talk to Claude on claude.ai or download the app ...
The thinking lever
The talk introduces test time compute for Claude, demonstrates its impact with a traffic simulation example using Opus 4.7, and explains token types, constraints, and the evolution to adaptive thinking and budgets to balance time, cost, and intelligence.
The expanding toolkit
The talk reframes the model as an expanding toolkit where tool use, context management, code execution, and computer use are built into the model, greatly simplifying development and boosting reliability. It uses demos and tips to show practical ways to leverage these capabilities through Claude and Claude Code, with a look ahead to broader agent level front doors for software.
Refining a PowerPoint with Claude
The video demonstrates how Claude can iteratively build and refine a deck by adding slides, making targeted edits, and converting bullets into visuals and native charts, all while preserving styling and context. It emphasizes a workflow where Claude asks clarifying questions and operates precisely on selected content.
Why do AI models hallucinate?
The video explains what AI hallucinations are, why they happen, and how Claude is trained and evaluated to reduce them, plus practical tips for spotting and mitigating hallucinations when using AI.
Get daily AI recaps from
Claude in your inbox
Get AI-powered summaries delivered to your inbox. Save hours every week while staying fully informed.