The founders who previously sold their livestreaming video startup Periscope to Twitter are back with a new venture. This time, their startup is focused on artificial intelligence.
On Wednesday, former Twitter head of product Kayvon Beykpour announced the launch of Macroscope, an AI system designed for developers and product leaders. The tool summarizes updates to a codebase and catches bugs, among other functions.
The startup was co-founded in July 2023 by Beykpour, who is now the CEO, along with his childhood friend Joe Bernstein. Both were previously at Periscope and their earlier enterprise startup, Terriblyclever, which was sold to Blackboard in 2009. They are joined by co-founder Rob Bishop, who sold his computer vision and machine learning company, Magic Pony Technology, to Twitter in 2016.
The company describes its product as an AI-powered understanding engine built to save engineers time. It is the type of tool the founders have stated they wish they had when building their earlier companies.
According to Beykpour, engineers today use a variety of tools like JIRA, Linear, and spreadsheets to track work, and they spend too much time in meetings instead of building. Macroscope is designed to solve this problem.
Beykpour explained that he personally experienced this pain at every company he worked at, from the startups he built to large public companies like Twitter. He noted that trying to understand what everyone was doing, especially in an organization with thousands of engineers, was a significant and frustrating part of his job as head of product.
To address this issue, Macroscope’s customers first install its GitHub app, which grants access to the code base. They can then optionally install other integrations like a Slack app, Linear app, or JIRA app. The software then analyzes the code and tracks what is changing.
This process involves code walking, which uses the Abstract Syntax Tree, a structural representation of programming code, to gather important context about how the code base works. This knowledge is then used in conjunction with large language models.
Once operational, engineers can use Macroscope to discover bugs in their pull requests, summarize their PRs, get a summary of codebase changes, and ask code research-based questions. Product leaders can use the software to get real-time summaries of product updates, productivity insights, and answers to natural language questions about the product or development activity. This helps them determine how engineering resources are being allocated.
Beykpour notes that users can ask natural language questions regardless of their technical ability. This is useful for someone trying to learn about the code base without distracting a senior engineer. He adds that a CEO wanting to know what was accomplished in a week can ask Macroscope instead of interrupting teammates, which is a much more expensive distraction.
While no product offers a direct competitor to everything Macroscope does, it competes in the code review space with tools like CodeRabbit, Cursor Bugbot, Graphite Diamond, and Greptile. The company stated that in its own internal benchmark of over 100 real-world bugs, its product caught five percent more bugs than the next-best tool and generated seventy-five percent fewer comments.
The software costs thirty dollars per active developer per month, starting at five seats. It offers enterprise pricing and custom integrations for larger businesses and requires the use of GitHub Cloud. A number of startups and larger firms have been using the product ahead of its launch, including XMTP, Things, UnitedMasters, Bilt, Class.com, Seed.com, ParkHub, and A24 Labs.
The San Francisco-based startup has a team of twenty people and is backed by thirty million dollars in Series A funding. The round was closed in July and led by Michael Mignano at Lightspeed. Other investors include Adverb, Thrive Capital, and Google Ventures. To date, Macroscope has raised a total of forty million dollars.