Are developers getting lazy? Is there a conspiracy to replace us all with AI-based code generators? Probably not, but it’s easy to suspect so with the apparent onslaught of AI/ML coding assistants from the world’s tech giants.
The latest addition to the growing list of coding assistants is Amazon CodeWhisperer. Amazon defines CodeWhisperer as an “ML-powered coding companion.” Sounds familiar? It might be because Github calls Copilot an “AI pair programmer” – essentially the same thing.
It’s clear why Amazon chose to release its alternative to Copilot. Github’s code completion tool made waves across the developer community with impressive adoption rates. Then, only days after Microsoft put a price tag on Copilot, Amazon announced it would be releasing a preview version of CodeWhisperer for IDEs that support the AWS Toolkit.
Speaking to TechCrunch, Vasi Philomin, Amazon’s VP in charge of AI services, emphasized that the company didn’t create CodeWhisperer to compete with Github’s Copilot. According to Amazon, CodeGuru (Amazon’s AI code reviewer and performance profiler) and DevOps Guru (a tool for finding operation issues), the company laid the groundwork for CodeWhisperer’s launch years ago.
For heavy AWS users, CodeWhisperer may seem like a logical AI/ML coding assistant choice. After all, it comes installed with the AWS Toolkit, and the waiting list can’t be long. But is CodeWhisperer worth trying out if you’re not looking for AWS-specific suggestions in your code? And how does it compare to more established code completion and suggestion tools?
What is Amazon CodeWhisperer?
Amazon CodeWhisperer is, according to the official description, a “machine learning (ML)–powered service that aims to improve developer productivity by generating code recommendations based on their comments in natural language and code in the integrated development environment (IDE).” Announced in June 2022, CodeWhisperer is available only upon request to users who apply for the preview period.
According to Amazon, they train CodeWhisperer on “billions of lines of code drawn from open source repositories, internal Amazon repositories, API documentation, and forums.” Then, you can use CodeWhisperer in your IDE to get recommendations based on previous code lines or comments.
Top 6 best alternatives to CodeWhisperer
Intelligent code completion is not new. Neither Github nor Amazon discovered the proverbial wheel of software development AI assistants. Quite a few solutions existed before CodeWhisperer and Copilot were announced. One of these veteran AI assistants is Tabnine – yup, that’s us.
Tabnine is a code completion tool that supports all the most popular languages, libraries, and frameworks in your favorite IDEs. Our AI models are only trained on permissive open source licensed code, so you can rest assured that your code is not going anywhere.
Tabnine Pro subscription lets you train your own private AI model based on your code and get personalized code completions that match your style and patterns. Getting to keep all your code on your local corporate is not the only benefit of training your own junior AI pair programmer. Hosting and training your own AI models for organizations and teams enables collaborative autocompletion across IDEs.
Free account (enabling short code completion)
Enterprise account: pricing is available on quote.
2. Github Copilot
Copilot’s somewhat bumpy launch into the world brought positive competition like CodeWhisperer and a lot more attention to the abilities of proven software development AI solutions.
$10/month or $100/year (free trial available).
3. Captain Stack
Amazon CodeWhisperer draws more from Captain Stack than Copilot, even though Github’s service inspired Captain Stack.
Captain Stack is a code suggestion tool that uses Google instead of AI. It sends your search query to Google, retrieves StackOverflow and Github Gist answers, and auto-completes them. Available for VSCode alone, Captain Stack is more of a shortcut to Google in your IDE than an intelligent code completion tool.
4. GPT-Code-Clippy (GPT-CC)
Another open-source plugin for VSCode that Copilot inspired is GTP-Code-Clippy.
GPT-Code-Clippy (GPT-CC) aims to be an open source version of GitHub Copilot, a language model (based on GPT-3, called GPT-Codex) fine-tuned on publicly available code from GitHub.
5. Second Mate
Just a second, mate! Let me wait for the code suggestions to load. Second Mate is another open-source AI code completion plugin. It uses EleutherAI GPT-Neo-2.7B (via Huggingface Model Hub) for Emacs. It is a much smaller model, so it will likely not be as effective as CodeWhisperer, Tabnine, or other alternatives.
6. Kite (Not Available)
Even though it is no longer maintained or available, we remember Kite fondly. This security-aware coding assistant was one of the more user-friendly and flexible players in the arena. Unfortunately, at the time of writing, there was no indication that Kite would ever return to the ring.
Whispering to code without Amazon
The six solutions listed above are not the only ones out there. YouCompleteMe, Google’s mysterious AI pair programmer, and other niche code completion and AI pair programming tools will continue to pop up as the battle for the role of your virtual coding assistant rages on. CodeWhisperer is still very new on the block, and it may take some time before it is adopted beyond the AWS fanbase.