Issue #8 | DailyPrompt

01 product

Anthropic Tests Removing Claude Code from Pro Plan, Triggering Developer Backlash

Anthropic quietly removed Claude Code from its $20/month Pro subscription for around two per cent of new users, updating the pricing page to show the feature excluded — while other documentation still listed it as included. The company's head of growth acknowledged poor communication and framed it as a limited test, but developers reacted sharply, with many cancelling subscriptions and exploring alternatives like local models and competing services. The move is widely read as a cost-management step ahead of a potential IPO, as subscription pricing has become economically unsustainable against actual token consumption.

The Register →

product

Anthropic Launches Prompt Cache Dashboard So Developers Can See Their Cost Savings

Anthropic has released a dedicated prompt caching dashboard in the Claude console, giving API customers real-time visibility into their cache hit rates and the cost savings being generated. Cached input tokens cost just 10% of standard pricing, but until now developers had no way to verify whether caching was working or quantify the benefit. The dashboard addresses a persistent gap for teams running high-volume production workloads.

ClaudeDevs

@ClaudeDevs

Caching is critical for customers to lower both costs and TTFT. We're launching a new dashboard in the Claude console to give you visibility into your prompt cache hit rate and cost savings.

Apr 22 View on X →

product

$ cat story_03.md

Claude Code Terminal Now Shows Session Recaps When You Return From Other Work

A new Claude Code update adds automatic recaps in the terminal: when a developer switches focus away from an active session and the task completes in the background, Claude now displays a summary of what it did when the developer returns. The feature tackles a common frustration for those running long background tasks who had to scroll through extensive output to understand what Claude had actioned.

$ open claudedevs →

research No. 04

Claude Mythos Helped Mozilla Find and Patch 271 Firefox Security Bugs

Mozilla confirmed that Anthropic's restricted Claude Mythos model identified 271 vulnerabilities in Firefox during internal testing, all of which have since been patched. The organisation noted that even a single one of these bugs would have been a red-alert security incident in 2025, and that while an elite human researcher could theoretically find each flaw, AI's ability to operate at speed and scale is fundamentally shifting the economics of defensive security.

Source: Yahoo Tech →

ClaudeDevs

@ClaudeDevs

Claude Code in the terminal will now show recaps when you switch focus away from the session and the session completes in the background.

Apr 22 View on X →

community

Anthropic Argues Training Claude on Song Lyrics Constitutes Fair Use

Anthropic has filed its formal fair use defence in the copyright lawsuit brought by Universal Music Group and other music publishers, arguing that ingesting lyrics alongside trillions of other words to train a general-purpose language model is transformative and creates no market harm to original works. The publishers counter that Claude 2 can reproduce songs nearly identically and that the use creates clear market dilution. The case is one of the most closely watched AI copyright disputes working its way through the US courts.

Billboard →

community

Claude Desktop Accused of Silently Installing Browser Extension Files Across Seven Browsers

Hacker News users have reported that the Claude Desktop app installed native messaging configuration files for a browser extension across Brave, Vivaldi, Arc, Edge, Opera, Chromium, and Chrome without displaying any permission prompt or settings disclosure. Anthropic has not commented publicly. Defenders argue this is standard native messaging setup used by apps like 1Password, but critics say the practice should require explicit user consent, and the thread has reopened broader debate about what desktop AI applications should be permitted to do silently.

Read on Hacker News →

Breaking

Anthropic Reverses OpenClaw Ban But Leaves Policy Confusion in Its Wake

After weeks of contradictory signals, Anthropic has confirmed through OpenClaw's creator that CLI-style usage of Claude — the pattern OpenClaw relies on — is permitted, following a public statement from Claude Code's head that such usage is allowed. However, some developers report the system still blocks OpenClaw-related prompts in practice, leaving the tool in what one described as 'a weird limbo where it should work in theory but doesn't.' The episode has intensified calls for Anthropic to publish clear, documented usage policies rather than communicating policy through ad-hoc social media posts.

Hacker News