Kayvon Beykpour on Periscope's origin story and his new AI engineering visibility tool Macroscope

Summary

{
  "long_summary": "Kayvon Beykpour, co-founder of Periscope and former head of product at Twitter, has emerged from stealth with Macroscope, an AI-powered engineering visibility platform backed by Thrive Capital, GV, and Lightspeed Venture Partners. The company started in July 2023.\n\n**What Macroscope does**\n\nMacroscope integrates with GitHub, Jira, Linear, and Slack to give engineering and product leaders automated answers to questions that currently eat enormous amounts of human time: what shipped this week, who built it, what bugs were introduced, how is the product evolving. Beykpour's core argument is that the source of truth for any software company is the codebase, and LLMs are now capable enough to read that source of truth and surface it automatically rather than forcing engineers through status meetings, update emails, and Slack threads.\n\nThe product has two distinct components today. The first is AI code review, a more competitive space where Macroscope has released a benchmark measuring what percentage of bugs its tool catches in pull requests. The second is what Beykpour calls the \"status\" layer — automated visibility into what teams are building — which he describes as genuinely greenfield, with no direct product competitor to displace.\n\nThe buyer is either the CTO or the CEO roughly half the time each. Beykpour says customers rarely question whether the problem is worth solving; the sales question is whether Macroscope actually works as described. His implicit pitch to engineering leaders is that catching a single production incident through AI code review more than justifies the cost.\n\n**The agent-era thesis**\n\nBeykpour argues the problem Macroscope solves gets harder, not easier, as AI coding agents proliferate. If companies are shipping ten times more code with fewer humans reviewing it, the need for an automated understanding layer compounds. Today the product is roughly answering \"what are my humans building, assisted by AI.\" In five years, Beykpour expects the question to be \"what have all my agents produced\" — but says the underlying problem is identical either way.\n\n**Roadmap signals**\n\nBeykpour points to two near-term directions: integrating with tools like Figma or LaunchDarkly to explain not just what changed in the codebase but why and who can see it; and automating the informal demo loop engineers already do manually, where someone merges a feature and records a quick Slack video to show it working. Macroscope wants to run the code automatically and surface that visual update without the engineer doing it.\n\n**Periscope origins**\n\nThe earlier part of the conversation covers Periscope's founding arc. The original prototype, called Bounty, was a reverse marketplace for real-time photos — users dropped pins on a map and offered prompts, and strangers responded with images. The team concluded static photos felt stale by definition and pivoted to live video. The floating hearts mechanic, an infinite tap-to-react gesture rather than a single like button, was the moment the small beta cohort clicked. A corp dev employee at Twitter was among those 30 beta users, which surfaced the product to Jack Dorsey and then-CEO Dick Costolo. Twitter acquired Periscope before it publicly launched.\n\nBeykpour's candid retrospective on Periscope: growing from zero to 100 million users in roughly 18 months was real, but building a live-only standalone social network was structurally untenable. Instagram and Facebook built all the same features faster inside platforms that already had asynchronous social graphs, and Periscope's integration with Twitter took too long to materialize. The lesson he draws is that live video as a standalone product can't compete with live as a feature inside a broader social network."
}