Codium enables developers to automate the critical process of creating tests. According to the business, $11 million in seed capital was used to create this AI model.
The Cost of Wrong Software
Such a tool has considerable potential. Many businesses are now doubting the caliber of their software after the astonishing $2 trillion cost of software faults in the U.S. alone in 2020. The cost of fixing errors increases as they spread across the software development life cycle. Yet, testing software is a difficult and time-consuming procedure.
Itamar Friedman and Dedy Kredo had direct experience with these difficulties as leaders of product and R&D teams at organizations like Alibaba Cloud. They founded Codium AI in 2022 after being persuaded by their backgrounds in software development, machine learning, and product management that AI large language models (LLMs) could be used to validate software tests.
TestGPT Makes Testing Easier
The co-founder and CEO of Codium, Friedman, emphasized the value of testing code as a developer. Along with finding flaws, it also helps you understand your code and demonstrates that you’re programming for a reason.
He felt it was tiresome and frustrating to write non-trivial test cases. The alternative of allowing a defect to enter production might be disastrous, therefore sometimes developing tests is even painful.
The first Codium tool is an IDE (integrated development environment) extension that enables an iterative process of creating tests and then modifying code based on the results of those tests. Through communication with the developer, the tool is able to better comprehend the code and produce more accurate and insightful tests, all the while assisting the developer in writing better code.
Developers that use Codium AI, according to the company, may anticipate finding errors and gaining insightful information about their code, enhancing the quality and functionality of their product.
Improved Code Integrity for Faster Development
The TestGPT system makes use of generative AI models, just like ChatGPT, Copilot, and other generative development tools. Friedman claims that TestGPT, however, is concentrated on ensuring that the code adheres to the desired specification. High code integrity is intended to be made possible so that developers can work more quickly.
It incorporates testing best practices into its prompting process and performs a number of pre-and post-processing stages to guarantee high-quality results, according to Friedman.
Nowadays, prominent IDEs like VS Code and PyCharm offer Codium as an add-on. There are plans to enable additional features and collaborations, as well as other IDEs and programming languages. Since the closed-alpha release of Codium in January 2023, the business claims that thousands of users have already installed it.
In order to continue to maintain high code integrity, Codium AI wants to grow and integrate into more phases of the software development life cycle. Its growth is anticipated to include management of tests and test data, CI/CD integration, automatic issue fixes, suggestions for code improvements, and support for test-driven development of the next generation.