Windsurf Editor Reviews

I've been a "Teams" user since the day 1 I met Codeium. Started my adventure with raw GPT-4, then looked into something more reasonably priced. Copilot seemed to be "that thing", but after checking it for a couple of hours, I realized this is much inferior due to its low context understanding. Raw GPT-4 chat window was eating 8-32K context like pancakes, while Copilot was limited to 4K (basically ex Bing AI, which was free). Then I looked into Codium and Codeium (yep, those are different) and the latter appeared to be just that thing. It much better understood the context of a project and didn't give me those annoying "wait for 30 minutes before typing again" type of messages. They've done it real smart - allow the user to go beyond that limit, and just get this capacity back while they sleep or it's the weekend. This is the part I hate about the raw open AI experience, that you pay for every single day you ain't using the service. Pay for API you say? That is even less reasonably priced. But my adventure wasn't all that fabulous. The assistant got substantially worse with time and later they revealed that it was no longer GPT-4 but GPT-4o and it suddenly made all the sense and confirmed my speculations. 4o is a very, very bad model for coding and complex thinking. I felt cheated as I could have the same thing with Bing yet for free. That day I strongly planned to unsubscribe. But soon after they added Sonnet and I can tell for sure, I'm here to stay, and they definitely too. Thank you for doing the right thing.

- Smart business model. Compared to the raw ChatGPT model you don't feel that message limit, and even when you do, you have plenty of other competitive models to use while you're getting back your capacity.
- Auto-completion is useful, but it's pretty much for any assistant
- They are listening. Compared to the whales in this niche they do hear us and quickly react, reflect, and implement leading ideas.

- The "Ctrl + I" code generator I found to be exceptionally useless. It breaks or deteriorates your code.
- GPT-4o is overrated.

Good experience, would definitely be my AI-powered editor of choice over Cursor. Composer mode would definitely be a useful feature.

Much better than Autocomplete and works across every editor platform. Does recognize context across files very well.

Takes a little bit of time to switch from muscle typing to reading the autocomplete and evaluating if it's what you want, if you know what you want

It is a great service, and it has helped me succeed. I have been able to code my dreams, and I plan to keep using this tool forever!

Fast AI Github Code Assistant, I love that I am able to develop new code with an AI chat, and I do not need an instructor.

Difficulty for noncoders, but with time, you can learn this quickly!

Codeium is a great tool to use if you're wanting to try the AI coding assistant. Codeium as of this review is free but does offer paid plans for teams with more customization.

Gives you options to choose which LLM models to choose.
Can customize your context by adding directories or repositories.
Over 70+ languages supported and provides better user UI interaction.
Oh and it's FREE.

Doesn't have as have as much context awareness of your VSCode environments compared to Copilot.

I evaluated most of the available AI coding assistants out there to help me be more efficient day-to-day and finally settled on Codeium. The base Llama 3.1 70b model is good, but you also get access to a beta Llama 3.1 405B model, which are great.

Even better, you get access to Claude 3.5 Sonnet and GPT-4o with the paid plan, which ends up only being a couple dollars more per month than the other major alternative which only has GPT-4. Easy decision.

Good base models with Llama 3.1 70b and 405b.
Access to Claude 3.5 Sonnet and GPT-4o models with paid plan.
Good performance.

Seemingly small context window, so really only good for small simple codebases.
No support for custom models (OpenAI, Ollama, Anthropic).
No support for "code" formatting in prompts for better readability.

Very great. Easy to install & setup! The responses are great and definitely do make me a better & more efficient programmer.

The model suggestions are much better than Copilot. It can also suggest modifications in the middle of a line, pushing back the rest of the line to the right, which is very useful.

The VSCode extension can be a little buggy sometimes, requiring me to restart my VSCode window. Happens maybe once a day.

Good alternative and useful sometimes, I like the select and explain functionality that they give us.

Good alternative to GitHub Copilot, VScode interface was very nice! The select and explain feature is very useful!

Nothing too far. Only silly issue is that I think it uses a port i use in my own environment.

Excellent product. It's well worth downloading the Windsurf editor to get the full experience. Recommending to friends.

* The context awareness is the best I've seen (knows my whole repo).
* The quality of the autocomplete/supercomplete is excellent.
* New features all the time and they're actually good.

* The documentation is lacking.
* They have the option to use their own models or OpenAI or Anthropic but there's no quantitative comparison of their quality.

Amazing experience, a must-have tool for any professional, specially those that are using multiple programming languages

This VSCode extension is not only much faster than Copilot, it's an order of magnitude more useful!!

Remembering shortcuts can be hard, it would be great to have a quick shortcut to display the ones used the most.

Helping a lot, but fails to stay focused on open folder, topic or chat, its helpful for any project with little bits of help it gets you really far, gets you to the finish line.

It's helpful for any project with little bits of help it gets you really far, gets you to the finish line.

Content awareness, chat focus, working folder awareness.

Incredible as a free product (best available imo,) even better with the $10 monthly price for a better experience.

From personal use, believe the autocomplete suggestions are the best available, somehow better than GH copilot. Also the refactor option and sidebar editor are super useful day to day.

Same as all other AIs, loses context if too long and lacking latest tech documentation.

When I first started using the extension, it was definitely different than traditional coding. Now, I find myself using the extension all the time and it truly does make coding more efficient and easier to keep organized. It makes me much better and providing quality comments as well.

The autocomplete is easily my favorite feature and the one I use the most. It gets better as you develop more and it really does a great job of understanding your coding style and adapting to it. The conversation chat is also really helpful when I have general app architecture or syntax related questions. I also find myself using the inline commenting feature quite frequently and I love not having to type out all the comments myself.

Although I use the inline action buttons a decent amount, sometimes they clutter the code as I am trying to read through the file. I am not sure if there is an easy way to add a keyboard shortcut that allows for toggling these on/off, but if there is, that would be a nice feature to add.

Great, I enjoy coding again now I have an assistant with me at all times. I make less mistakes and find security easier now.

Autocomplete makes my coding faster, the assistant helps me find bugs and silly issues fast. i can refactor and describe code in seconds that used to take hours.

Sometimes defaults to old AI model instead of premium model

Very good, it has saved a significant amount of time, added value in terms of code and served as shortcut for documentation.

It gives me access to a number of different models, provides some local context to the model from your code. It helps solve coding problems, will often suggest fitting solutions. I like the chat functionality and the fact that it will answer generic questions outside of the tech domain.

It is the early days of consumer AI and the models are not as smart as they one day will be. It is possible to end up in a ignorance loop and waste some time, I wish it just admitted when it does not know the right answer. No chat support for Emacs. Would like to see that side get some attention.

I tried all the ai code completion plugins for PyCharm including copilot and Codeium won. I'm a fan. I pay for the premium sub, and tbh i don't use what it comes with but I like supporting them.

Code completion is fast and accurate. It uses context hints from surround code, you clipboard, and other sources. Easy to setup.

It regularly causes some errors in PyCharm. The little toast popups in the bottom right that say 'error'. I've also had a few other minor oddities.

Overall I recommend using Codeium with your IDE! It has multiple features that aid when writing code whether it be improvement or learning!

Using Codeium has made refactor suggestions easy by having the ability to highlight specific sections of the code within an IDE! I specifically use VS Code and it has been working pretty smoothly. I also like having the option of having a chat box so that I can discuss other topics.

Codeium sometimes struggles to connect to my IDE and I am unable to use. I haven't been able to figure out the cause of this but sometimes I am unable to use Coddium for a couple days and then all of the sudden it is back.

Codeium has the best overall copilot experience on the market. The fact that the free tier already provides a product that is better than its competitors is really telling. The pricing for the paid plans is fair and offer features worth the price

On top of a great copilot, you have a community in which the developers actively participate in. Why use anything else?

- Saves A TON of time writing repetitive code. Helps debugging, coming up with solutions, explaining code, and everything else (and more) that you can expect from a copilot.

- Allows switching between additional models such as GPT and Claude

- Tends to have features before its competitors do

- Developers actively engage with its userbase and are always open to feedback

I don't have any immediate complaints with my current use cases.

I do wish however that they added support for more libraries in their live chat feature. While it still works fine for the libraries I use, and they already support at least 12 different major libraries, there are still some libraries I wish they supported such as Flutter, Vue, and .NET which would improve the live chat experience even more.

This is an affordable alternative to GitHub Copilot that is often more intuitive and seamless to use. I love the chat interface and how it allows you to select different foundation models to interact with as some may be better at writing code while others at writing documentation.

- Chat interface to explain, analyze, and review code
- Made good suggestions for improving style
- Fast, context-aware code auto-completion
- Quick suggestions for writing unit-test, documentation, and refactoring your code

- Writing unit test is hit or sometimes miss and can require multiple generations
- For large codebase, you may have to specify the features you want in terms of bite-sized functions

Codeium is one of the best AI coding assistant tools, and its price is very reasonable. I will keep using it.

Support multiple IDEs and Editors, such as VS Code, JetBrains IDEs, Sublime Text, Vim. Codeium is affordable.

Few AI models, only models of OpenAI; it would be better to add some other models.

As someone who has been writing software for over a decade now, I have to say that Codeium is truly a tectonic shift in the way I do my job. I'm faster, more efficient, and more accurate in what I do, and I can't help but wonder if I can go back to not using it anymore!

- Very useful code completion and snippets
- There are many times when it read my mind and created a nice and accurate block of code where I was about to start typing something
- I've had a lot of "a-ha!" moments when it would show me a new way of writing some function that I would not have thought of
- Being able to bring up a chatbot inside VSCode to help me build a new function is a godsend
- Being able to have the Codeium assistant explain a function in plain English is super helpful as well

- Sometimes, it blocks the regular function/variable autocomplete in VS code (you can get around it by hitting backspace, escape, and then the next letter)
- There are times when the AI has hallucinated some code that isn't super relevant, but that's mostly when I'm working on some relatively cutting-edge stuff that hasn't been done before.