Getting This Site Online

I’ve been working for the past three weeks to make this site public. Although it’s still under construction, I’m happy to announce that it is now online. There was a lot to do to get it working—researching hosting options, web frameworks, and more. Throughout this process, I’ve relied on several AI assistants to help me get the site up and running. I primarily use macOS as my development platform with VS Code, and up until about a week ago, I was using GitHub Copilot along with GPT-4o. However, I recently discovered Codeium and have decided to switch over to it as my in-IDE coding assistant.

What sets Codeium apart from other AI coding assistants is its ability to use context that you specify in the form of custom instructions, much like ChatGPT, but it also uses local context in the form of Retrieval-Augmented Generation (RAG) from directories and repositories on your local machine. This is great because it has access to your codebase while you’re working inside the IDE.

Upgrading to the Pro/Teams version gives you access to their Llama 3.1 405B model, which has been fine-tuned on many languages and used for various projects. Its autocomplete is very good and quite fast. It integrates seamlessly with VS Code, but they also offer a command line version called “Termium.” Inside VS Code, you access it much the same way you access Copilot—either from the edit session or the chat sidebar. Codeium also provides usage statistics for how much you have been using Codeium and details on how you have been using it. The Pro/Teams version includes limited access to GPT-4o as well along with team statistics. This is my personal public profile with my current statistics: (https://codeium.com/profile/unixwzrd)

One of the standout features of Codeium is its ability to produce a set of references in response to a prompt, similar to what Perplexity offers. I still use ChatGPT and the GPT-4o model for brainstorming, coding solutions, and double-checking Codeium’s suggestions. Since Codeium doesn’t have Internet retrieval yet, ChatGPT fills that gap. Overall, I’m impressed with Codeium. It even costs less than Copilot—$10 a month with an annual subscription or $12 a month if paid monthly. I think I’ll be sticking with Codeium going forward and canceling my Copilot subscription, as Codeium appears to do everything Copilot can, but faster and better.

You can check out Codeium on their website. They have a free version available, so give it a try—they are constantly updating and enhancing their capabilities. You might make the switch too.

If you want to give Codeium a try head over to Codeium and check it out.

Add your thoughts or questions about the topic 2024-08-02 AI Coding Assistants