Because I use Claude more than the other AIs, this announcement was exciting for me (Limbikani) but I'm unhappy about the whole limited preview thing for Claude Code! Anyway... *First, Sonnet 3.7* * Can switch between quick responses and extended thinking mode (visible step-by-step reasoning thing that most AI interfaces are copying from DeepSeek) * API users can control how long the model thinks (up to 128K tokens) * Same pricing as previous models ($3/million input tokens, $15/million output tokens) * Available on all Claude plans (Free, Pro, Team, Enterprise) - but I'm assuming free has the same annoying limitations ekunzi, mirira kusvika mangwana to continue 😔 * Extended thinking mode not available on free * Some improvements in coding capabilities - they are claiming it outperforms competitors but that jury is obviously still out *🧑‍💻 Claude Code* As indicated above, this is in limited research preview so you will have to ask to be let in (here https://docs.anthropic.com/en/docs/agents-and-tools/claude-code/overview) There's a waitlist and they are saying it's at capacity ☹️. It's a new command-line tool for agentic coding letting devs delegate engineering tasks directly from their terminal It san search/read code, edit files, write/run tests, commit/push to GitHub Anthoropic says Claude Code is already indispensable for its own developers and it has reduced development time for complex tasks there. Full announcement here: https://www.anthropic.com/news/claude-3-7-sonnet *Other Updates* * GitHub integration now available on all Claude plans * Claude 3.7 Sonnet reduces unnecessary refusals by 45% compared to predecessor * Detailed system card available explaining safety evaluations