Current model hierarchy for models with multiple names. This is not a comparison of performance between brands.

(Last updated August 3, 2026.)

How to easily setup per directory API keys to, for example, scope different AI subscriptions in different folders. This will use the CLI direnv to accomplish that.

Here is a script using the GitHub gh CLI to iterate over all of your repos and make them more secure by setting the following settings on them:

• pull requests only allowed for collaborators (to prevent cache poisoning attacks)
• immutable releases (because immutable releases are good)

A quick guide on how to track session cost with a Claude hook. This will both keep a running cost counter below your Claude CLI and persist session costs to ~/.claude/cost/.

Add the following to your ~/.bashrc or ~/.zshrc:

alias ghpr='git push && gh pr create --fill'

Now just run ghpr after you've committed, and now you have a PR!

(for security purposes)

• Globally:
  • run npm config set ignore-scripts true
  • this will populate ~/.npmrc
• Per repo:
  • create ./.npmrc
  • add ignore-scripts=true into that file

How to setup a Linux VM for an AI CLI to go crazy in (without risk to your host system).

I thought this would be easy (because it's easy to stream from a backend to a frontend and vice versa) but it was not. Let's look at how to stream a file from an S3 connection on a server to another server in Node.js.

I recently had to figure out client-side mTLS certificates. I've never had to deal with them before, and it was quite annoying, so I documented what I learned.

Npm has recently been making a lot of changes around package publish security because of recent supply-chain attacks targeting Npm.

What does this mean for the lonely dev self-publishing packages from their local CLI?