Kayvon Beykpour on Periscope's origin story, Twitter acquisition, and new venture Macroscope — X-ray vision for engineering teams

Summary

Kayvon Beykpour, co-founder of Periscope and former head of product at Twitter, has launched Macroscope, an AI tool that gives engineering leaders automatic visibility into what their teams are building — pulling from the codebase rather than meetings, spreadsheets, or status updates.

From Bounty to Periscope

Periscope began as a prototype called Bounty — a reverse marketplace where users dropped a pin on a map and crowd-sourced photos of what was happening there. Beykpour describes it as an early attempt to build a "teleportation device." When the team recognized the liquidity problem with static, already-stale photos, they pivoted to live video and added floating hearts as an infinite-expression mechanic. The product never had a formal launch. Twitter saw a beta of roughly 30 users, and the acquisition happened before Periscope shipped publicly.

Growth after the acquisition was fast — zero to 100 million users in around 18 months — but Beykpour is candid about what went wrong. Periscope was live-only, while Instagram and Facebook treated live as a feature layered on top of asynchronous social graphs. By the time Periscope built async connection features, competitors had already replicated everything. The Twitter integration that was supposed to differentiate the product took too long to materialize.

What Macroscope does

Beykpour founded Macroscope in July 2023, seeded by Thrive Capital, Adverb Ventures, and GV. The company announced a Series A led by Lightspeed in September 2025.

The pitch is that for any software company, the codebase is the source of truth — and LLMs are now capable enough to read it and answer questions that previously required meetings, Jira queries, or pestering engineers. Macroscope integrates with GitHub, Jira, Linear, and Slack, and surfaces answers to questions like what changed this week, who shipped it, how a system works, and what is currently in front of which users.

The product has two components. The first is AI code review, a more crowded space where Macroscope competes on bug-detection benchmarks — specifically, what percentage of bugs it catches in a pull request before they reach production. The second is what Beykpour calls the "status" layer: automated visibility into anything happening across the product development process. He describes this as largely greenfield, with no established product category to displace.

The buyer is split roughly evenly between CTOs or heads of engineering and CEOs. Beykpour argues the value proposition for code review sells itself — catching a single production incident justifies the cost intuitively — and that the harder question customers ask is not whether the tool is valuable but whether it actually works as advertised.

The agentic coding tailwind

Beykpour argues the AI coding wave makes Macroscope more necessary, not less. If companies are producing ten times more code with fewer humans writing or reviewing it, understanding what was built, by whom, and why becomes harder — not easier. He frames Macroscope as an "AI air traffic control system" designed to keep human leaders accountable for outputs even as agents do more of the production. His expectation is that today's use case — what are my engineers building with AI assistance — gradually shifts so that in five years the primary question is what have my agents produced, with humans accounting for a smaller share of output. The underlying question stays the same.

On the roadmap, Beykpour points to automating the Slack demo post engineers write when they ship a feature — running the code automatically and generating a visual of what just merged into staging — as one near-term direction. Generated dashboards and richer visual status outputs are also planned, though the team is still sequencing where to start.