Highlights:

  • The GPT-4 large language model from OpenAI serves as the foundation for the AI model TestGPT that CodiumAI has developed.
  • TestGPT analyzes the freshly written code, docstring, and comments and then recommends which tests should be run to ensure the functionality and integrity of that code.

After closing a USD 11 million seed financing round, software code quality startup CodiumAI Ltd. is coming out of stealth mode with the beta test launch of its generative AI-powered code integrity tool.

Vine Ventures and TLV Partners co-led the round today, and angel investors including some unnamed officials from OpenAI LP, VMware Inc., and Snyk Ltd. also invested.

The GPT-4 large language model from OpenAI serves as the foundation for the AI model TestGPT that CodiumAI has developed. Its responsibility is to help developers evaluate their code interactively by creating tests as they go.

Testing is an essential step in the coding process, as CodiumAI describes. The business asserts that software of questionable quality cost US companies a combined USD 2.08 trillion in 2020 alone. Errors in software code can be costly.

Given those expenses, it stands to reason that preventing mistakes is preferable to fixing them after they have already occurred during the software development lifecycle. Regular and thorough testing is necessary to avoid errors, but for developers, this can take a lot of time. Furthermore, it’s a flawed procedure because they frequently are unsure of the proper things to screen for.

Dedy Kredo, Co-founder and Chief Product Officer at CodiumAI, said, “Developers have a love-hate relationship with writing tests for their code. They love the quality it brings but hate spending time on this. CodiumAI is here to change that, offering free auto-generated software test suite suggestions to developers, starting with Python, JavaScript and TypeScript.”

As the developers work, TestGPT analyzes the freshly written code, docstring, and comments and then recommends which tests should be run to ensure the functionality and integrity of that code. To maintain the integrity of their code, the developer only needs to accept and commit those tests when TestGPT recommends them.

Itamar Friedman, Co-founder and Chief Executive, claimed that his business operates at the nexus of cutting-edge innovations like ChatGPT and coding helpers like GitHub Copilot. It has modified ChatGPT’s generative AI powers for testing code logic.

The tool from CodiumAI is currently in beta and accessible as an extension for well-known integrated development environments (IDEs) like PyCharm, VSCode, and WebStorm. It will eventually integrate with more IDEs and more programming languages. It has been installed by thousands of developers since its January release as a closed alpha, according to the firm.

TLV Partners’ Brian Sack said he’s backing CodiumAI because it tackles a real problem for developers. There is no practical method for evaluating code logic, even though they have access to tools for testing performance, identifying software vulnerabilities, and generating code.

“It is not a coincidence as it is a very challenging task, almost like getting into the mind of the one who wrote the code. CodiumAI is a creative, targeted, and focused use of generative AI and an application that provides value immediately,” said Sack.