Best of the Pod: Would You Shut Down Your Most Successful Product? The Arc to Dia Story

Quick Overview

The Browser Company pivoted from their successful product Arc to build a completely new browser called Dia because they realized the rise of Large Language Models (LLMs) necessitated rewriting all software for a new personal intelligence layer that sits across all devices, moving beyond the limitations of Arc to capture the mass market.

Key Points: The decision to stop Arc and build Dia was driven by the realization that primary computer interfaces are changing due to AI, requiring a complete rewrite of software. Arc suffered from a "novelty tax," attracting early adopters but making it difficult to reach the mass market due to its complexity. Josh Hirs stated that the first two weeks after releasing Dia to private beta could not have gone better, despite acknowledging that "last year sucked" during the transition. The initial instinct was to evolve Arc into ARC 2.0, but internal issues and the need to build from scratch for AI accessibility led to the creation of a completely new product, Dia. Arc Search provided a strategic realization: the intent portion of browsing (search) would be drastically changed by AI, and the browser's role as an application platform would also evolve. The team built Dia on a completely new, high-performance architecture using Swift's structured concurrency to ensure it functioned as a fast, foundational browser, fixing performance issues present in Arc's Redux-based architecture.

Context: The discussion centers on the journey of The Browser Company, co-founded by Josh Hirs and another speaker (implied to be the CEO, Dan is the interviewer), who previously created the popular desktop web browser Arc. The interview focuses on the difficult, high-stakes decision to essentially shut down Arc, which had millions of monthly users, to start over and build a new product named Dia, driven by the sudden emergence and potential of generative AI.

Detailed Analysis

The speakers detail the difficult but necessary pivot from Arc to Dia, acknowledging that the public reaction to abandoning Arc was unexpectedly negative, especially for a company with 70 employees and millions of users. They explain that Arc hit a "novelty tax," preventing mass market adoption, and simultaneously, the explosion of AI led them to believe that all software needed rewriting for a new paradigm—a personal intelligence layer across devices. Initially, they considered an ARC 2.0, but internal resistance to fundamentally changing beloved features and the need to start from scratch to properly leverage AI led them to create Dia as a completely new product, a decision finalized around September of the previous year. A key trigger was Arc Search, which revealed that the intent portion of browsing would be drastically altered by LLMs. Josh Hirs noted that the conviction in this new direction was high internally, though met with external skepticism, even from their own families, who viewed AI as "nonsense." Architecturally, Dia was built from the ground up using Swift and structured concurrency to be fundamentally fast, unlike Arc, which utilized a complex Redux state-driven architecture optimized for rapid prototyping but suffered performance impacts. They credit the company culture, built on the value "assume you don't know," for allowing the team to embrace the blank page and prototype the future, which they conceptualized as the "Internet computer"—a system where the browser is merely the enabling technology for a miraculous personal assistant layer.