Browser Maker Ditches Custom Engine for Chromium: The Lessons From a Failed Fork
Archi4 min read
Contents
The Fork That Almost Worked#
For 26 months, a team of 14 engineers at a mid-sized browser company worked on their own rendering engine. The fork was based on an open-source project but had drifted far enough that merges from upstream took weeks. By the time the company's leadership called a halt, the engine powered less than 2 percent of the browser's 1.8 million monthly users. The rest were already on a Chromium fallback that the team had quietly shipped to stop the bleeding.
This is not a story about technical failure in isolation. It is about a decision that every browser maker with ambitions eventually faces: build your own core, or adopt someone else's and compete on everything else. The company chose to stop building. The reasons and the fallout hold lessons for any team that owns a platform layer.
The Hidden Cost of Divergence#
The engine team had a clean architecture and a growing test suite. What they did not have was a way to keep up with the pace of the web platform. New CSS features, updated JavaScript APIs, and security patches landed upstream every few weeks. Each one required manual porting because the fork had replaced the networking stack and the layout engine's handling of flexbox.
By month 18, the team was spending 60 percent of its engineering hours on merges and compatibility shims. The remaining 40 percent went to features that mattered to users, like faster tab switching and better memory handling on low-end Android phones. A competitor using stock Chromium shipped the same features in a fraction of the time.
The math became impossible to ignore. The company's leadership asked for a one-page summary: what would it cost to keep the fork alive for another year? The answer was 11 engineers and a six-month delay on the mobile redesign. The alternative was to move the remaining users to Chromium and reassign the engine team to performance work on top of it.
The Decision and the Pushback#
The CEO made the call in a Tuesday morning meeting. There was no dramatic reveal. The engineering director presented the numbers, the product lead asked two questions about user-facing regressions, and the decision was made to stop development on the fork within 90 days.
That timeline was optimistic. The migration took five months. The hardest part was not the code. It was the internal resistance from engineers who had spent two years on the engine. One senior developer left. Two others transferred to other teams. The remaining eleven stayed, but the first month after the announcement was marked by a quiet drop in commit activity that the manager had to address directly.
The company also had to tell its most technical users. A blog post explained the move in plain terms: the fork could not keep up with security patches, and the team would rather spend its time on features than on porting. The post got 340 comments, many of them negative. A few users threatened to switch browsers. The company offered a one-year window for anyone who wanted to export their data and move on.
What They Would Do Differently#
Looking back, the engineering director points to three mistakes, none of them technical.
First, the team never set a kill criterion. The original plan had milestones for performance and stability, but no explicit point at which the company would stop. The fork survived two quarterly reviews that should have ended it because no one wanted to be the person who proposed shutting it down.
Second, the fallback Chromium build was treated as a temporary measure. It was under-resourced and shipped with a generic user interface. When users switched to it, they noticed the drop in polish. That made the eventual migration harder because the fallback had a reputation problem.
Third, the team underestimated how much of the fork's value was social rather than technical. Engineers liked working on a unique engine. It was a recruiting tool. Killing it meant losing a story the company told about itself.
The Aftermath#
The company now runs a Chromium-based browser with a custom interface layer and a set of privacy features that sit on top of the engine. The eleven former engine engineers have spent the last year on memory management and tab suspension. The browser's crash rate dropped by 22 percent in the first six months after the migration, mostly because the team was finally able to take upstream fixes without manual porting.
The mobile redesign shipped four months late but with no major regressions. The company's monthly active users are flat, not growing. The leadership says that is acceptable for now. The point was to stop the drain, not to win a market share war in one quarter.
The Lesson for Platform Teams#
Any team that owns a core layer, whether it is a browser engine, a database, or a build system, faces the same trap. A fork or a custom implementation starts as a way to move faster. It becomes a liability when the upstream project's pace exceeds your ability to merge. The decision to stop is always harder than the decision to start because it means admitting that the original bet did not pay off.
The browser company's experience suggests two practical rules. Set a kill criterion before you write the first line of divergent code, and make it specific: if we cannot merge upstream within two weeks, we stop. And resource the fallback as if it is the primary product, because eventually it will be.
The company did not fail because its engineers were not good enough. It failed because it kept a fork alive past the point where the numbers made sense, and the cost was two years of engineering time that could have gone into features users could see. The decision to stop was late, but it was still the right one.
More articles