A weekly podcast covering Claude Code updates, Anthropic engineering, and AI news. About 6 minutes, every Monday. Hosted by Shannon.
Reading the feed…
Claude Code now picks a session back up automatically once your usage limit resets, so a long task no longer stalls until you return. A new Concise output style trims narration, more settings panels open mid-turn, and memory growth in long sessions is capped. Plus reports on Anthropic's potential public listing and enterprise growth. In this episode: (00:00) Introduction (00:45) Claude Code Updates (03:06) Under the Hood (04:11) In the News Further reading: - The Claude Code guide for startups (Claude Blog): https://claude.com/blog/claude-code-guide-for-startups - Claude on call: How Claude Ta
Subagent forking is now on by default, so forked agents inherit your full conversation and prompt cache instead of a cold start. Plus you can mention live sessions by name, todo tools change on newer models, and Anthropic reports quarterly revenue over eleven and a half billion dollars. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:44) Under the Hood (03:50) In the News Further reading: - Maximizing the value of your Claude Code sessions (Claude Blog): https://claude.com/blog/maximizing-the-value-of-your-claude-code-sessions - How Claude's text watermark works (Anthropi
Auto mode is now the default permission on Pro, Max, and Team plans, replacing the per-tool prompt flow. Self-hosted runners let you keep sessions inside your own infrastructure. Plus, Meta enters the coding agent market and Anthropic signs a ten billion dollar compute deal. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:48) Under the Hood (04:13) In the News Further reading: - Auto mode is now the default in Claude Code for Pro, Max, and Team plans (Anthropic Engineering): https://claude.com/blog/auto-mode-default-in-claude-code - Running auto mode in production (Anthro
Anthropic disclosed three incidents where Claude models escaped cybersecurity evaluation sandboxes and accessed real systems. The new Model Context Protocol spec is rolling out with a stateless core and hardened authentication. Plus, banks are lining up fifteen billion dollars for a Google-backed Anthropic data center. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:30) Under the Hood (03:09) In the News Further reading: - Investigating three real-world incidents in our cybersecurity evaluations (Anthropic News): https://www.anthropic.com/news/investigating-incidents-cybe
Claude Opus 5 is now the default Opus model, with a one million token context window and a new fast mode tier. Plus subagents that nest three levels deep, background code reviews, and news on the AMD deal and the approved books settlement. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:53) Under the Hood (03:56) In the News Further reading: - Introducing Claude Opus 5 (Anthropic News): https://www.anthropic.com/news/claude-opus-5 - The new rules of context engineering for Claude 5 generation models (Anthropic Engineering): https://claude.com/blog/the-new-rules-of-context
The fork command now spawns background sessions and hard caps stop runaway agents. Plus Meta's reported ten billion dollar compute deal with Anthropic and IPO talks from DealBook. In this episode: (00:00) Introduction (00:45) Claude Code Updates (03:00) Under the Hood (04:22) In the News Further reading: - Ending a subset of conversations (Anthropic): https://www.anthropic.com/research/end-subset-conversations - How Anthropic runs large-scale code migrations with Claude Code (Anthropic Engineering): https://claude.com/blog/ai-code-migration - Zero risk isn't the job: a CISO's guide to agentic
Claude Opus 4.8 is now the default on Bedrock, Vertex AI, and Claude Platform, auto mode is generally available, and the doctor command becomes a full setup checkup. Plus Ben Bernanke joins Anthropic's oversight trust and TeraWulf signs a nineteen billion dollar infrastructure lease. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:57) Under the Hood (04:05) In the News Further reading: - Choosing a Claude model and effort level in Claude Code (Anthropic Engineering): https://claude.com/blog/claude-model-and-effort-level-in-claude-code - A field guide to Claude Fable 5: Fi
Claude Sonnet 5 is now the default model in Claude Code, bringing a one million token context window and background-by-default subagents. Plus, the U.S. lifted export restrictions on Anthropic's frontier models, and Anthropic is reportedly in talks with Samsung on a custom AI chip. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:50) Under the Hood (04:12) In the News Further reading: - Introducing Claude Sonnet 5 (Anthropic News): https://www.anthropic.com/news/claude-sonnet-5 - Giving admins more visibility and control over Claude spend (Claude Blog): https://claude.com/
Rewind now recovers context from before you cleared the conversation, and bash mode picks up automatic Claude responses and live file path completion. Plus, Wired covers Mythos returning to vetted US organizations and the Wall Street Journal reports on Anthropic's distillation claims against Alibaba. In this episode: (00:00) Introduction (00:45) Claude Code Updates (03:04) Under the Hood (03:50) In the News Further reading: - Building effective human-agent teams (Anthropic Engineering): https://claude.com/blog/building-effective-human-agent-teams - Introducing Claude Tag (Anthropic News): http
Claude Code now previews in-progress work as live, interactive artifacts you can share with teammates. Auto mode adds guardrails against destructive git and infrastructure commands. And John Jumper is leaving Google DeepMind for Anthropic. In this episode: (00:00) Introduction (00:45) Claude Code Updates (03:03) Under the Hood (04:09) In the News Further reading: - Claude Code now supports artifacts (Anthropic Engineering): https://claude.com/blog/artifacts-in-claude-code - Steering Claude Code: CLAUDE.md files, skills, hooks, rules, subagents and more (Anthropic Engineering): https://claude.c
Claude Fable 5 was available for three days before a US export directive forced Anthropic to suspend access. Sub-agents can now nest five levels deep, a new safe mode flag helps isolate config problems, and the New York Times, Wall Street Journal, and Axios all cover the export story from different angles. In this episode: (00:00) Introduction (00:45) Claude Code Updates (03:22) Under the Hood (04:25) In the News Further reading: - Claude Fable 5 and Claude Mythos 5 (Anthropic News): https://www.anthropic.com/news/claude-fable-5-mythos-5 - Statement on the US government directive to suspend ac
Claude Code now lets you configure a chain of up to three fallback models, and sensitive config writes get new prompts. In the news, Anthropic confidentially filed for an IPO, called for a coordinated pause on self-improving AI, and drew pricing criticism from Microsoft. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:59) Under the Hood (04:30) In the News Further reading: - A harness for every task: dynamic workflows in Claude Code (Anthropic Engineering): https://claude.com/blog/a-harness-for-every-task-dynamic-workflows-in-claude-code - Lessons from building Claude Cod
Claude Opus 4.8 is now the default model with high effort enabled and a faster, cheaper fast mode. Dynamic workflows can orchestrate hundreds of background agents. And Anthropic closed a funding round that vaulted its valuation past OpenAI. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:29) Under the Hood (03:32) In the News Further reading: - Introducing Claude Opus 4.8 (Anthropic News): https://www.anthropic.com/news/claude-opus-4-8 - Introducing dynamic workflows in Claude Code (Claude Blog): https://claude.com/blog/introducing-dynamic-workflows-in-claude-code - Using
The simplify command becomes code-review with effort levels and inline PR comments. Usage now itemizes consumption by skills, subagents, plugins, and MCP servers. And Anthropic is reportedly tracking toward its first profitable quarter on roughly eleven billion dollars in Q2 revenue. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:46) Under the Hood (03:54) In the News Further reading: - Using Claude Code: The unreasonable effectiveness of HTML (Anthropic Engineering): https://claude.com/blog/using-claude-code-the-unreasonable-effectiveness-of-html - New in Claude Managed
Claude Code's new agent dashboard lets you see every session in one place, and fast mode quietly upgrades to Opus 4.7. Plus, Anthropic closes a thirty billion dollar round at a nine hundred billion dollar valuation, and Microsoft reportedly pulls internal Claude Code licenses. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:51) Under the Hood (04:08) In the News Further reading: - Agent view in Claude Code (Anthropic Engineering): https://claude.com/blog/agent-view-in-claude-code - How Claude Code works in large codebases: Best practices and where to start (Anthropic Engi
Claude Code makes worktree base branches configurable and exposes effort levels to hooks and Bash commands. Plus Anthropic raises usage limits and signs compute deals with SpaceX and Akamai. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:56) Under the Hood (04:02) In the News Further reading: - New in Claude Managed Agents: dreaming, outcomes, and multiagent orchestration (Anthropic Engineering): https://claude.com/blog/new-in-claude-managed-agents - Higher usage limits for Claude and a compute deal with SpaceX (Anthropic News): https://www.anthropic.com/news/higher-limi
A new project purge command wipes per-project state, the model picker now reads from your gateway, and OAuth login finally works in WSL2 and SSH. Plus Pentagon AI deals without Anthropic, Mythos scrutiny, and a $900 billion valuation in the works. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:59) Under the Hood (04:28) In the News Further reading: - Lessons from building Claude Code: Prompt caching is everything (Anthropic Engineering): https://claude.com/blog/lessons-from-building-claude-code-prompt-caching-is-everything - Onboarding Claude Code like a new developer: L
Config settings now persist across restarts, vim mode picks up visual selection, and the Opus 4.7 context window math gets corrected. Plus Google commits up to forty billion dollars to Anthropic, and Mozilla used Mythos to fix 271 Firefox bugs. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:33) Under the Hood (03:43) In the News Further reading: - Building agents that reach production systems with MCP (Anthropic Engineering): https://claude.com/blog/building-agents-that-reach-production-systems-with-mcp - Built-in memory for Claude Managed Agents (Anthropic Engineering):
Claude Opus 4.7 is now available in Claude Code with a new extra-high effort level and an interactive slider for tuning the speed-intelligence tradeoff. This week also brought a native binary for faster startup, a parallel multi-agent code review command, and significant security hardening for Bash permission rules. In this episode: (00:00) Introduction (00:45) Claude Code Updates (03:05) Under the Hood (04:20) In the News Further reading: - Best practices for using Claude Opus 4.7 with Claude Code (Anthropic Engineering): https://claude.com/blog/best-practices-for-using-claude-opus-4-7-with-c
Claude Code's default effort level shifts from medium to high for most users, meaning better output quality at the cost of more tokens. Critical security patches fix a Bash permission bypass and command injection vulnerabilities. Plus, Mythos's cybersecurity implications draw attention from Wired, the New York Times, and federal regulators. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:49) Under the Hood (04:10) In the News Further reading: - Seeing like an agent: how we design tools in Claude Code (Anthropic Engineering): https://claude.com/blog/seeing-like-an-agent -
Claude Code ships an interactive Bedrock setup wizard and per-model cost breakdowns. Anthropic scrambles to contain a source code leak after hackers repackage it with malware, and the BBC reports on growing frustration with usage limits. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:53) Under the Hood (03:54) In the News Further reading: - Anthropic Races to Contain Leak of Code Behind Claude AI Agent (Wall Street Journal): https://www.wsj.com/tech/ai/anthropic-races-to-contain-leak-of-code-behind-claude-ai-agent-4bc5acc7 - Security News This Week: Hackers Are Posting t
Auto mode brings safer permissionless operation to Claude Code with built-in safety heuristics. This week also delivers transcript search, conditional hook filters, and a native PowerShell preview on Windows -- alongside news of a federal court ruling preserving Anthropic's government access and reports of an IPO as early as October. In this episode: (00:00) Introduction (00:45) Claude Code Updates (03:13) Under the Hood (04:15) In the News Further reading: - Claude Code auto mode: a safer way to skip permissions (Anthropic Engineering): https://www.anthropic.com/engineering/claude-code-auto-m
Claude Code ships a new bare mode that strips it down for scripted pipelines and CI jobs. MCP channels let servers push messages into your session, VS Code gains remote control bridging, and responses now stream line by line. In the news, Anthropic overtakes OpenAI in enterprise revenue and Cursor plans its own rival model. In this episode: (00:00) Introduction (00:45) Claude Code Updates (02:58) Under the Hood (04:03) In the News Further reading: - AI Coding Startup Cursor Plans New Model to Rival Anthropic, OpenAI (Bloomberg): https://www.bloomberg.com/news/articles/2026-03-19/ai-coding-star
Opus 4.6 now defaults to a one million token context window on Max, Team, and Enterprise plans. MCP servers can prompt you for input mid-task, and a new context diagnostics command flags what's eating your token budget. Plus, Wired profiles OpenAI's race to catch Claude Code, and Anthropic launches a dedicated code review tool. In this episode: (00:00) Introduction (00:45) Claude Code Updates (03:03) Under the Hood (04:10) In the News Further reading: - 1M context is now generally available for Opus 4.6 and Sonnet 4.6 (Claude Blog): https://claude.com/blog/1m-context-ga - Quantifying infrastru
Opus 4.6 is now the default model for Max and Team users, and a new loop command lets you schedule recurring prompts inside a session. Plus, voice mode expands to twenty languages, the Pentagon blacklist fallout continues, and Bloomberg reports Anthropic is nearing a twenty billion dollar revenue run rate. In this episode: (00:00) Introduction (00:42) Claude Code Updates (03:12) Under the Hood (04:21) In the News Further reading: - Common workflow patterns for AI agents—and when to use them (Claude Blog): https://claude.com/blog/common-workflow-patterns-for-ai-agents-and-when-to-use-them - Qua
