Tabnine updates AI-based code-writing assistant for developers

Tabnine, which develops an AI-powered assistant for software developers that can anticipate code to be written for specific apps and fills in blanks with the appropriate snippets, launched a new version of its time-saving application programming interface on June 15.

The company, which uses a freemium-plus-service provisioning business model, also said it has eclipsed the 1 million user mark of its API in 2022.

Tel Aviv-based Tabnine’s software has a long memory for finding reusable code in 31 languages, including Python, JavaScript, Java, C#, TypeScript and other languages available today in both Visual Studio Code or IntelliJ IDEs, CEO Dror Weiss told VentureBeat.

“The use case of our product is very simple,” Weiss said. “Your developer writes code in her IDE, and she has a top line installed as part of her editor. We have a real-time look at the code that’s being created, and contextually we query the AI engine in real time; we then suggest parts like the big the next line or snippets or a block of code that the developer needs. The app makes a really snap judgment whether this is what they intended to do, and they accept the code. This happens most often with basically zero change in the process.”

‘Ninety percent of all code has already been written’

Ninety percent of all code has already been written and doesn’t need to be created from scratch, Weiss said.

Developers are now seeing 30% to 40% of their code being generated automatically using the Tabnine API without them needing to deviate, Weiss said. “No code is changed whatsoever from the standard workflow of writing code in their favorite editor, and it works for all popular ideas or code editors and all popular programming languages,” he said.

During the past year, Weiss said, Tabnine has revamped both its AI models and its approach to integrating new models into its platform. As opposed to a monolithic approach that mandates a single AI model, Tabnine is now capable of supporting any compatible AI model for providing guidance to developers, Weiss said.

From infrastructure-as-code to continuous integration, companies that have implemented devops in recent years have seen productivity improvements in their software development and release processes. But those productivity benefits have typically stopped at the developer’s keyboard, because an explosion of frameworks, tools, languages, and APIs have only made a developer’s job harder, Weiss said.

AI has big potential to help developer productivity

AI has the potential to do for developer productivity what CI has done for the release process, Weiss said.

“Every day, we get amazing feedback from developers and teams that Tabnine’s AI is now part of their flow, giving them time back to focus on higher-value activities,” Weiss said in a media advisory. “With longer and better code completions, our next-generation AI technology is a big step forward for developer productivity, further establishing Tabnine as the leader in AI-powered software development.”

The new-generation platform includes a dozen new AI models, each trained from the ground up on code (vs. models pre-trained on text and retrofitted to learn code). Each model is optimized for a specific language or domain, enabling Tabnine to contextually complete full lines and snippets of code with unprecedented accuracy, Weiss said.

Tabnine, started up in 2017, also announced it has raised $15.5 million in funding from Qualcomm Ventures, OurCrowd and Samsung NEXT Ventures, with participation from existing investors that include Khosla Ventures, Headline Ventures, Hetz Ventures and TPY Capital. With this latest infusion of capital, Tabnine has raised a total of $32 million to date.

Tabnine claims to be the only such AI-powered software development tool of its kind in the market, Weiss said.

The latest funding will be used to further accelerate development of Tabnine’s platform to add support for additional AI models and programming languages later this year, Weiss said.