Anthropic expanded Sonnet's capabilities by introducing a 1M token context window option in beta testing. This represents a significant jump in context handling for the mid-tier model, enabling processing of much longer documents and codebases without truncation, which should benefit users working with large-scale projects.
Anthropic published a postmortem addressing quality issues in Claude Code and related products, acknowledging that recent default reasoning effort changes and session-idle thinking bugs had degraded performance. The team reverted problematic updates and fixed underlying issues, signaling a course correction after community complaints on Hacker News.
Anthropic expanded Claude's managed agents infrastructure with new capabilities including dreaming (internal reasoning), multiagent orchestration, outcomes tracking, and webhook support. These tools enable developers to build more sophisticated autonomous agent systems with better visibility and coordination between multiple agents.
Claude Code's SDK received tooling improvements including `claude agents --json` for programmatic session listing and `/resume` support for background sessions. The update also introduced stricter auto-mode rules via `settings.autoMode.hard_deny` and memory efficiency improvements, reflecting refinements to developer ergonomics.
Discussion on tech forums including Hacker News revealed community concerns and questions around Anthropic's approach to persistent, always-on agents through its Dispatch and Cowork features. The debate touches on use cases, competitive positioning, and whether Anthropic's strategy differs meaningfully from always-on agent offerings from competitors.
Anthropic has acquired Stainless, a company specializing in API development tools and infrastructure. The move strengthens Anthropic's ability to manage and scale its API services for Claude integrations. This acquisition signals investment in developer tooling and platform reliability as Claude adoption grows.