Claude Code Update is INSANE!
Anthropic has added push notifications to Claude Code, allowing users to assign long AI tasks and walk away from their computers. When Claude finishes or gets stuck, it sends a phone notification for user input. The presenter argues this fundamentally shifts AI from a 'babysitting' model to an asynchronous workflow model.
Summary
The video covers a new feature released by Anthropic for Claude Code: push notifications. The presenter explains that users can now assign Claude Code a long-running task — such as writing code, running tests, or fixing bugs — and physically leave their computer. When Claude either completes the task or encounters a decision point requiring human input, it sends a push notification directly to the user's phone. The user can then approve, deny, or give new instructions right from the notification itself, without sitting back down at their computer.
The presenter argues this update is more significant than it appears because, until now, every AI tool required the user to remain present and watch the AI work. Even advanced AI agents demanded constant supervision. Claude Code's push notification system breaks this pattern by enabling genuinely asynchronous AI workflows — where the AI reaches out to the human rather than the other way around. The technical underpinning involves a hooks and event system within Claude Code's workflow, triggering notifications on events like task completion, required input, or errors.
The presenter also notes this is not limited to developers. Anthropic's separate 'computer use' capability (referred to as 'co-work') allows Claude to control an entire computer — browsing the web, clicking buttons, and filling out forms. Combined with push notifications, the presenter envisions a near-future scenario where Claude can handle complex, multi-step business tasks like lead research and email drafting entirely in the background, pinging the user only when a decision is needed.
A time-savings argument is made: if a user runs three 15-minute Claude tasks per day, push notifications reclaim 45 minutes of previously wasted 'watching' time daily — over 5 hours per week. For larger tasks, the savings compound further. The presenter frames this as a shift from trading time one-for-one with AI to running multiple workstreams in parallel.
The video closes with a step-by-step setup guide and substantial promotion of the presenter's paid community, the AI Profit Boardroom, and a free community called the AI Success Lab.
Key Insights
- The presenter argues that before this update, every AI tool on the planet — including the most advanced agents — still required users to remain glued to their screen, and Claude Code's push notifications are the first instance of an AI doing complex multi-step work and then proactively reaching out to the user on their phone.
- The presenter explains that Claude Code's notification system is built on a hooks and event system, where specific workflow events — task complete, input needed, or error encountered — trigger notifications with built-in approve or deny buttons, so users never even need to open an app.
- The presenter claims that Anthropic's 'co-work' feature, which lets Claude control an entire computer and browse the web, combined with push notifications, points toward a near future where Claude can handle any computer-based business task in the background — such as researching leads, building spreadsheets, and drafting personalized emails.
- The presenter argues the bottleneck in AI productivity was never AI speed but the requirement for users to watch and supervise, and that push notifications remove this bottleneck by enabling parallel workstreams — running one Claude instance on one task while the user handles another or runs a second Claude instance simultaneously.
- The presenter contends that the core problem Anthropic solved — how does an AI communicate with you when it needs you — is simple and obvious in hindsight, yet no AI company had implemented it until this update, making it a pivotal shift from 'babysitting AI' to 'managing AI.'
Topics
Full transcript available for MurmurCast members
Sign Up to Access