Meet Macroscope: an AI tool for understanding your code base, fixing bugs (techcrunch.com)

Ex-Twitter and Periscope leaders have launched Macroscope, an "AI-powered understanding engine" for engineering teams that summarizes codebase changes, finds bugs, and answers natural-language questions about projects. Co-founded by Kayvon Beykpour, Joe Bernstein and Rob Bishop, the San Francisco startup integrates via a GitHub app (with optional Slack, Linear and JIRA connectors) to analyze repositories and produce PR summaries, bug catches, change heatmaps, and productivity insights for both engineers and product leaders. The product targets pain points around context-switching and meeting-heavy status updates by surfacing actionable, real-time summaries of what teams are actually shipping. Technically, Macroscope combines static program analysis—"code walking" using Abstract Syntax Trees (ASTs) to capture structural code context—with large language models to reason over that representation. The company says this hybrid approach yields better recall and fewer noisy comments in reviews: in an internal benchmark over 100 real-world bugs it caught 5% more issues than the nearest competitor and produced 75% fewer comments. It requires GitHub Cloud, costs $30 per active developer/month (starting at five seats), and has early customers and $40M raised ($30M Series A). For the AI/ML community, Macroscope illustrates a practical, production-ready trend: combining symbolic program analysis with LLM reasoning to deliver more precise code intelligence at org scale, while raising integration and access trade-offs around repository security and governance.