Jacob Jackson's journey into the world of AI began early in his career. While still a computer science student at the University of Waterloo, he co-founded Tabnine, an AI-powered coding assistant that attracted nearly $60 million in venture capital. After successfully selling Tabnine to Codata in 2019—right during his final exams—Jackson took an internship at OpenAI, where he worked until 2022.
At this point, Jackson felt inspired to launch another company focused on enhancing developer workflows. “Since my time with Tabnine, tools like ChatGPT and GitHub Copilot have revolutionized how developers operate,” said Jackson. “This is a thrilling time to innovate in developer tools, given the significant advancements in technology that have sparked greater interest among developers in leveraging AI to boost their productivity.”
This motivation led to the creation of Supermaven, a new AI coding platform inspired by Tabnine but boasting several enhancements for both user experience and functionality. Supermaven's proprietary generative AI model, Babble, has a remarkable 1-million-token context window, which allows it to comprehend extensive pieces of code simultaneously. In data science, tokens are smaller bits of raw data—like the syllables “fan,” “tas,” and “tic” in the word “fantastic.”
The context window of a model defines the input data (e.g., code) it considers when producing output (e.g., additional code). A larger context window helps minimize "hallucinations," or inaccuracies, by enabling the model to draw answers from its surrounding context instead of making guesses. “Our extensive context window reduces the likelihood of errors, as it allows the model to refer back to relevant data when generating output,” Jackson explained.
Though Supermaven’s 1-million-token context window is impressive, it isn’t the largest in the AI coding landscape; startup Magic boasts a context window of 100 million tokens. However, it matches Google’s new Code Assist tool at 1 million tokens.
So what sets Supermaven apart from its competitors? Jackson claims that Babble boasts lower latency due to a “new neural architecture” developed entirely from the ground up. “Supermaven takes only 10 to 20 seconds to analyze a developer’s code repository and familiarize itself with its APIs and coding conventions,” Jackson said. “Thanks to our in-house model serving infrastructure, our tool remains highly responsive, even when processing the large prompts typically associated with extensive codebases.”
The demand for AI coding tools is rapidly expanding, with Polaris Research forecasting a market value of $27.17 billion by 2032. A significant percentage of respondents in GitHub’s latest developer survey report having adopted AI tools in one form or another. Over 1.8 million individuals and approximately 50,000 businesses are currently subscribers to GitHub Copilot.
However, Supermaven, along with other startups like Cognition, Anysphere, Poolside, Codeium, and Augment, faces critical ethical and legal challenges. Companies are often hesitant to share proprietary code with third parties; for instance, Apple reportedly prohibited employees from using Copilot last year due to worries about confidential data leaks. Some code-generating tools trained on licensed or copyrighted code have been shown to inadvertently reproduce that code, which poses a legal liability risk for developers who might unknowingly use it. Moreover, as AI systems can make errors, assistive coding tools may introduce faulty and insecure code into programming environments.
Jackson asserted that Supermaven does not utilize customer data to train its models. However, he acknowledged that the company retains data for a week to enhance the system's speed and responsiveness. Regarding copyright issues, he did not outright deny that Babble was trained on intellectual property-protected code; he indicated that the training mainly relied on publicly available code to minimize exposure to harmful content.
Despite these concerns, Supermaven's user base continues to grow, with over 35,000 developers currently utilizing the platform. A significant number have chosen premium options, with the Pro plan priced at $10 per month and the Team plan at $10 per month per user. Supermaven's annual recurring revenue hit $1 million this year, thanks to a user base that has tripled since its February launch.
This impressive momentum has attracted the attention of venture capitalists. Supermaven recently secured its first external funding, raising $12 million in a round led by Bessemer Venture Partners, along with notable angel investors, including OpenAI co-founder John Schulman and Perplexity co-founder Denis Yarats. Jackson plans to allocate these funds toward hiring additional developers (the current team consists of five people) and refining Supermaven’s text editor, which is still in beta.
“We aim to grow significantly by the year’s end,” he added. “Despite some challenges facing the tech industry as a whole, the market for coding copilots is expanding rapidly. Our growth since launch—in addition to this latest funding round—positions us well as we look forward to the upcoming year.”
Ranking
