Claude Code

88 stories

Claude Code is Anthropic's command-line AI coding tool, which reads and edits code and runs commands in the terminal. Recent coverage has focused on security and staffing: its "Don't Ask Again" option was reported to have exposed secrets in 1,122 public GitHub files, and a GitHub issue claims it sends user emails in the User-Agent header. Founding product lead Catherine Wu left for OpenAI, the third departure from the team in two weeks.

Related topics

Study of 207,000 Claude sessions shows how scientists use AI

Researchers analyzed 207,000 Claude and Claude Code research sessions and found heaviest use in formal sciences like computer science and math, less in health and lab-based physical sciences. Data analysis accounts for about a third of Claude sessions and nearly half of Claude Code sessions, while idea generation and data gathering—rated most important by scientists—are used less. Tasks AI performs well get two to four times the usage predicted by researcher time spent.

Hacker News ·

Stroq locally audits what coding agents do

Open-source tool Stroq reads Claude Code and Codex session logs to show how many commands originated from text an agent read, and which key names left the machine (never values). It runs via npx with only a local cache and no telemetry, and can guard agents or replay known attacks.

Hacker News ·

Gauge open-sources CLI to iteratively improve docs for coding agents

Gauge has open-sourced a CLI that improves documentation for coding agents, including a local backend that needs no account. It runs tasks that depend on the docs, analyzes session transcripts for friction, generates doc variants and reruns the tasks with those variants injected, iterating until issues are resolved; the loop also works for Skills and CLI surfaces, while a network gateway swaps docs transparently and blanks out credentials.

Hacker News ·

Kimi K3.1 Pro Open Weights Released

Moonshot released weights for Kimi K3.1 Pro, described as its most capable 3T-class open-weight, natively multimodal agentic model. Post-trained from Kimi K3 with large-scale joint reinforcement learning, it offers a 1-million-token context, swaps full attention for Gated DSA, and improves coding, 3D creation and financial research.

Hacker News ·

Anthropic launches tools to show AI spending value

Anthropic released reporting tools that let enterprise customers view, by team, what AI work produced, what it cost and where workflows broke down, plus improvement suggestions. It said the data is not used for model training by default and added pooled team budgets. The move aims to prove ROI ahead of a planned IPO; Anthropic told investors it posted second-quarter operating profit, with Claude Code driving 900% net revenue retention.

TechCrunch ·

Anthropic files for IPO expected in November

Reuters reports Anthropic has filed a preliminary prospectus with the SEC for an IPO expected in mid-November. The filing describes AI evolving from assistive tools into autonomous systems that run projects for days or weeks, spawning 'virtual colleagues' with access to corporate files, software and databases. Claude Code writes, updates, tests and debugs code automatically, Claude Cowork extends this to administrative work, and the company envisions running millions of agents in parallel.

صوت الإمارات ·
Load more