Everyone and their mother are vibe coding their own solutions like this, all the time.

Memory is not just a matter of retrieval, it's also a matter of knowing what to retrieve and when.

Can we even define AI before getting into this debate?

Also the first voice of KARR from Knight Rider.

How about build whatever interface makes sense for the user? Doesn't matter if it's a TUI, GUI, speech interface, or some physically actuated / haptic feedback based system. Do what makes intuitive sense to the user and provides enough expressive power to get their job done without overwhelming them.

An agent will read your comment and take it as a challenge to show something can go wrong. Thanks for prompt injecting, mate. /s

This will have significant consequences for the military.

Autonomous and semi-autonomous systems address that.

In that they'll have to act like a competitive employer?

It will be interesting to see how South Korea holds its borders 5 generations from now. That population will have shrunk substantially. Perhaps a robotic workforce and a drone military will finally deliver luxury capitalism to them. Or they will simply lose their control of their borders.

Countries with larger populations will see these guys commit civilizational suicide with enough time to recover one can hope. We won’t be funding USAID when the majority of Americans are drawing from Medicare though.

This is beautiful, I can't deny that. But the claim that Claude Code is a fixed set of tools is quite wrong. Claude Code regularly generates new code to carry out tasks, and you can choose to promote those tasks to skills, workflows, etc.

Much of the world believed in geocentrism for a very long time, yet we've known better for a long time now. Currently, this "Ptolemaic" idea of human intelligence being the center of everything sounds similar. What if human intelligence is not at the center? What if other forms, such as AI, emerge as the new "center"?

Basically, as long as you are using an Anthropic library or tool, you can use your OAuth credentials. For example, you can use the Claude Agent SDK with your OAuth credentials. This is sweet because I can prototype all sorts of agents with Claude Code embedded inside, at a predictable monthly cost. One nice use case is turning skills into standalone tools or apps.

You can also do convoluted things like run Claude Code within tmux and send input to it and read the output.

MCP Channels are interesting too for bidirectional communication between your app and a running Claude Code instance, with an MCP server sitting in between. It's slow, but allows for some interesting use cases when you want to step out of an existing CLI session to do work that is easier in a graphical interface, have Claude Code respond and do work, then when you're done, go back to the CLI session and continue, never losing context.

The way I read the Anthropic docs, it seems the term plan is to block the usage of OAuth credentials with the "Claude Agent SDK".

This URL: https://code.claude.com/docs/en/agent-sdk/overview

Says this: "Unless previously approved, Anthropic does not allow third party developers to offer claude.ai login or rate limits for their products, including agents built on the Claude Agent SDK. Please use the API key authentication methods described in this document instead."

Again, it seems Anthropic prefers to bill API token rates (long run), not subscriber effective token rates.

You don't really need to tmux at all for Claude Code CLI. Claude Code CLI supports streaming json input, and streaming json output; you can use stdin/out as a pipe to control Claude Code CLI.

I'm doing this today in https://github.com/Cidan/ask -- works great.

This is awesome. Thank you for pointing out streaming JSON input and output, I'll definitely be taking a look at that.