Voice Mode Launches for Claude Code — Free Transcription, Real-Time
Anthropic is rolling out voice mode for Claude Code, letting developers issue commands entirely by voice. Hold the spacebar to record, release to transcribe — voice input flows directly into the cursor position in real time alongside normal keyboard use. The transition between voice and typing is seamless. Voice transcription tokens are completely free, with no additional billing. Anthropic engineer Thariq Shihipar confirmed the rollout, noting users activate it with the /voice command.
The feature is currently live for approximately 5% of Claude Code users, with a broader rollout planned over the coming weeks. It's a meaningful quality-of-life upgrade for developers who spend long sessions in the terminal — particularly useful for narrating agentic runs, reviewing diffs out loud, or simply keeping hands off the keyboard during intensive coding sprints.
Automatic Prompt Caching Now Available — No Manual Breakpoints Required
Anthropic added a new simplified cache_control field to the Messages API that handles prompt caching automatically. The system identifies the last cacheable block in your request and moves the cache point forward as conversations grow — no manual cache breakpoint management needed. The feature works alongside existing block-level cache controls for developers who want more granular control. Auto-caching is especially valuable for long-running agentic workflows where context windows grow incrementally across many turns.
Claude Code Analytics API: Org-Level Usage Data Now Programmable
Anthropic's new Claude Code Analytics API gives organizations programmatic access to daily aggregated usage metrics for their Claude Code deployments. Each endpoint returns a date-specific snapshot covering productivity metrics, tool usage stats, cost breakdowns, and more — all scoped per organization, per day. For engineering managers and platform teams running large Claude Code rollouts, this closes a major gap: cost attribution, ROI measurement, and capacity planning no longer require manual console exports. The API is live today.
Claude Code Revenue Tops $2.5B Annual Run-Rate — Doubled Since January
Claude Code's annualized revenue surpassed $2.5 billion in February 2026, more than doubling from its run-rate at the start of the year. Weekly active users also doubled over the same period. The numbers make Claude Code Anthropic's fastest-growing product line and a significant force in the AI coding assistant market — putting real pressure on GitHub Copilot, Cursor, and other incumbents. The Pentagon dispute, which drove a wave of new Claude sign-ups broadly, appears to have had a halo effect on the developer product as well.
ChatGPT Uninstalls Jump 295% in the US — Users Flood to Claude
Data from app analytics firms shows US uninstalls of ChatGPT jumped 295% following OpenAI's announcement of a Pentagon partnership — the deal that Anthropic publicly refused. Claude download counts surged in parallel, with the app holding the #1 spot on Apple's free app chart. Social media chatter skewed heavily toward users explicitly citing the Pentagon deal as their reason for switching. The episode is turning into a real-world test of whether AI safety commitments move consumer behavior at scale — and early data suggests they do.
Haiku 3 Retirement: 46 Days Left — Migrate to Haiku 4.5 Now
Claude Haiku 3 (claude-3-haiku-20240307) retires on April 19, 2026. Any application calling that model string will break hard after the deadline — no graceful fallback. The migration path is Haiku 4.5, which delivers better performance at comparable pricing. With 46 days on the clock, teams running Haiku 3 in production should be running parallel tests against Haiku 4.5 this week. Developer forums are starting to surface first-mover migration reports, most of which are positive with minor prompt tuning required.
Anthropic Goes to Court — Pentagon Challenge Moves to Legal Phase
Anthropic confirmed it will file a federal court challenge against the Pentagon's "supply-chain risk" designation, which bars Anthropic products from use across US military and federal agencies. Legal scholars at Lawfare flagged procedural irregularities in how the designation was applied and raised First Amendment concerns around compelled speech in commercial contracts. The general consensus: the designation is on shaky legal ground and unlikely to survive full judicial review.
The legal battle sets up a landmark case for the AI industry — one that could define the limits of government authority to dictate how AI companies set safety guardrails in commercial products. Meanwhile, Anthropic continues to benefit in the court of public opinion, with new user sign-up records every day of the dispute and a 295% spike in ChatGPT defections following OpenAI's decision to fill the void Anthropic vacated.