monroewalker

I frequently see high praise for Kagi here, both for the quality of the search results as well as the ostensibly customer-friendly business decisions.

For the people that use their search and trust the company - why? Not to say I don’t trust them or trust them less than any other company claiming to be privacy and security focused, just would like to know how others made their decision.

The obvious strawman argument seems to be that since they charge for search, their business model is based on subscriptions rather than monetizing search data. Nothing would stop a company from charging for search AND selling data

I'm finally feeling that I've gone through the setup process for a new computer enough times that I'm ready to setup a repo to host some install script I can just curl into bash. Some thoughts I have so far:

- goal is to have something like `curl https://somesiteorgithub.com/myscript | bash` which will bring a given computer up to date with whatever the latest changes are that I've made to my setup - script should be re-runnable without causing issues and to install anything new on machine x that I've added on machine y - dotfiles should be updated too - will probably have the script checkout a setup repo I'll use to store everything and then have the script pull if the repo is already there - will symlink dotfiles from the setup repo - sharing shell history in some way might be nice.. haven't looked into how that might be done yet though

For those that have already gone through this process, any recommendations? Any unexpected obstacles to getting something working? Any good publicly shared references for something like this?

edit: Claude provided the following in response to a similar question: ``` Yes! These are often called "dotfiles," "bootstrap scripts," or "system initialization scripts." Several notable developers and organizations maintain public versions. Here are some popular examples:

Popular Individual Dotfiles: Mathias Bynens' dotfiles - One of the most starred dotfiles repos Paul Irish's dotfiles - Frontend-focused setup Thoughtbot's laptop - Professional development company's setup holman/dotfiles - Organized by topic, very well structured Companies' Setup Scripts: Github's codespaces dotfiles Thoughtbot's dotfiles Tools/Frameworks: mac-dev-playbook - Ansible-based setup Homebrew Bundle - Brewfile approach mackup - Keep application settings in sync Dotfiles Managers: dotbot chezmoi yadm Resources to find more:

dotfiles.github.io - GitHub's guide to dotfiles awesome-dotfiles - Curated list of dotfiles resources ```

I'll check out the references later. Curious if anyone has some experience with any of the above that's worth sharing.

Currently using Terminator when on Ubuntu and iTerm2 on my macbook. Always with zsh + oh-my-zsh plus a handful of plugins and aliases.

Curious to know what the real terminal and shell wizards are up to.

Lately shell interactions have felt very slow and archaic without some kind of AI assistance now that I've become so used to having it when coding with Copilot. I've seen and tried some of the available AI shell tools but haven't liked any yet. Just typing command descriptions and waiting for the AI generated command doesn't feel like the right model. Literally just using copilot in a context where I could execute its output would probably be a better experience. If I remember correctly, when playing with Emacs there was one of the terminal options where the output and command line were in a single buffer and hitting enter on the last line would execute the command there. So I'm thinking that if there's a copilot plugin for Emacs and that terminal option, then combining the two might be pretty nice.

Current list for me is:

- NYT Connections https://www.nytimes.com/games/connections

- NYT Strands https://www.nytimes.com/games/strands

- Betweenle https://betweenle.com/

- Symbl https://www.symble.app/

- Mazele https://mazele.io/

- Hidden Mirrors https://www.hidden-mirrors.com/

Is there a good reference for which models should be preferred for which tasks? As I understand it, there's a decent amount of overlap in the capability of the top models right now (Claude 3, Claude 3.5, GPT4o, OpenAI o1, Gemini 1.5 Pro, Llama 3.1, etc.) but each have their own strengths and weaknesses as well as differences in the API pricing, context windows, usage privacy policies, etc. It'd be nice to know how each stack up against each other for specific use cases. Eg. writing -> use Claude 3, coding -> use Claude 3.5, STEM / logic questions -> OpenAI o1, huge context -> Gemini 1.5 Pro, self-hosting / running locally -> Llama 3.1

I do see a lot of articles on Google from people just personally testing and comparing models based on their own criteria. I haven't found anything yet though that seems to be both thorough and well maintained (given how often new models are released or updated).

I've had mine in the default target retirement funds of the provider (Fidelity / Empower / Vanguard) from different employers. After seeing how much of the fund is invested in bonds (10% for Vanguard) despite the long time until retirement, I'm looking at moving everything to VOO / VTSAX / VFIAX to have a more aggressive allocation in 100% equity.

Curious what others here have opted to do.

I've been buying Nuun but am curious to know what products others have settled on, especially after finding out Nestle owns Nuun now. Or if you make your own, how's that been?