●RELEASE — Claude Code v2.1.243 shipped on August 25 with broad improvements across usage reporting, model selection, sign-in, and reliability●USAGE — /usage now breaks results down per loop, showing run count, total tokens, tokens per run, and last run, which makes a chatty /loop task easy to spot●SETTINGS — modelPicker lets you curate the /model list with your own order and labels, while promptCacheTtl and subagentPromptCacheTtl let the main conversation and subagents keep different cache lifetimes●LOGIN — /login now offers keyless sign-in with an Anthropic Console account, so organizations that do not permit API keys can still get in●PERFORMANCE — The native binary is now zstd-compressed, dropping from roughly 340MB to 75MB on Linux x64, and each session uses about 40 to 60MB less memory●FIX — v2.1.245 resolves a startup crash on distributions shipping glibc 2.44, including Arch Linux, CachyOS, and Fedora Rawhide●RELEASE — Claude Code v2.1.243 shipped on August 25 with broad improvements across usage reporting, model selection, sign-in, and reliability●USAGE — /usage now breaks results down per loop, showing run count, total tokens, tokens per run, and last run, which makes a chatty /loop task easy to spot●SETTINGS — modelPicker lets you curate the /model list with your own order and labels, while promptCacheTtl and subagentPromptCacheTtl let the main conversation and subagents keep different cache lifetimes●LOGIN — /login now offers keyless sign-in with an Anthropic Console account, so organizations that do not permit API keys can still get in●PERFORMANCE — The native binary is now zstd-compressed, dropping from roughly 340MB to 75MB on Linux x64, and each session uses about 40 to 60MB less memory●FIX — v2.1.245 resolves a startup crash on distributions shipping glibc 2.44, including Arch Linux, CachyOS, and Fedora Rawhide