Skip to content

Weekly digest

2 – 8 November 2026

8 posts

Claude Max, Ultra and Team subscribers now get a monthly API credit for building with the Claude API and Agent SDK, the Claude Code team said in its October recap. MCP servers that support it can now ask a follow-up question mid-task and run long jobs in the background.

Each 30-second status update from a background subagent in Claude Code costs about 99% less, because Haiku 5.5 writes it from short notes instead of re-reading the whole conversation. Anthropic says sessions with background subagents used about 5% less of the limit.

Claude Max, Ultra and Team plans add a monthly API credit

Subscribers on Max, Ultra and Team now get a monthly API credit for building with the Claude API and Agent SDK, the Claude Code team said in its October recap. If you’re on one of those plans, the credit comes every month.

Cloud sessions also start setting up while you’re still typing the prompt, and any line in a cloud session’s diff can take a comment that goes to Claude with your next message. MCP servers that support it can now ask a follow-up question mid-task and run long jobs in the background.

ClaudeDevs TwitterDiscuss

Claude Code cuts subagent update cost 99%, session use 5%

Each 30-second status update from a background subagent in Claude Code costs about 99% less, because Haiku 5.5 writes it from short notes instead of re-reading the whole conversation. The figure that reaches your limit is smaller: Anthropic says sessions with background subagents used about 5% less of it.

That works out to about 6% more prompts on the same limit. The subagent doing the work keeps its own model, and only the status line goes to Haiku 5.5.

It’s live for everyone on Team and Enterprise and for most Pro, Max and Ultra users on v2.1.316 or later.

ClaudeDevs TwitterDiscuss

Claude desktop gets /storage to show and clear used space

According to Crypto Briefing, a new /storage command in the Claude desktop app shows how much local storage Claude takes up. The same command clears inactive tasks and unused worktrees, the extra Git working copies a repository can have alongside the main one.

Testing Catalog ThreadsDiscuss

Investors doubt even the lowered $1.5T Anthropic valuation

Major investors are skeptical that Anthropic’s IPO can reach a $1.5 trillion valuation after a deep market selloff, The Wall Street Journal reports. That figure was already the fallback: three Reuters sources earlier named around $1.5 trillion as the lower number investors were weighing, against roughly $2 trillion expected from a November listing.

Earlier, Reuters reported that 45% quarterly revenue growth disappointed some investors.

Hacker NewsDiscuss

FT: Anthropic strikes $3.4B gas power deal for data centers

The Financial Times reports that Anthropic has struck a $3.4 billion deal to generate gas-fired power for its data centers. Earlier, The Wall Street Journal put Anthropic’s seven-year cloud deal with Akamai at no less than $11.6 billion.

Hacker NewsDiscuss

FT: OpenAI hires a former US Army secretary

The Financial Times reports that OpenAI has brought on a former secretary of the US Army, as the military looks to expand its use of AI.

Earlier, The Intercept reported, citing documents released through a FOIA lawsuit, that the Defense Department asked OpenAI for custom models for national security use with “minimal refusal rates”.

Hacker News CodexDiscuss

OpenAI’s Brockman calls first half “pretty rough”

Greg Brockman, OpenAI’s president, said Wednesday at the WSJ Tech conference that the first half was “pretty rough,” with Anthropic pulling ahead with business customers. That’s his account from a conference stage. The latest outside figure, per the Financial Times, put OpenAI’s run rate at roughly $50 billion in September against Anthropic’s $65 billion.

Brockman said OpenAI’s agentic coding tools only caught on in the third quarter, when enterprise overtook consumer as the company’s biggest revenue driver, and he cited a “revenue inflection” in Q3. Before that, The Wall Street Journal reported, second-quarter revenue grew 18% to $6.7 billion.

Hacker News CodexDiscuss

OpenAI’s dots can now read email codes to pass 2FA

A Threads post reports that OpenAI’s dots can now take sign-up codes from an email to complete a 2FA check on your behalf when prompted, and Crypto Briefing and Leodex carry the same claim. If a dot can read your inbox, it can get past a service’s second login step without you, though reports differ on whether it enters the code itself or hesitates before doing so.

In OpenAI’s introduction of dots, each one gets its own identity, credentials and system access. Dots launched in ChatGPT for Pro and Business Premium users in eligible markets.

Testing Catalog ThreadsDiscuss