Alex Built the Cursor for Xcode, Then OpenAI Hired the Team
By the end of 2024, AI had rewired almost every code editor in common use. Copilot lived inside VS Code. Cursor had forked VS Code outright and turned that fork into a multi-billion-dollar business. JetBrains shipped an assistant across its IDE line.
The big exception was Apple's ecosystem. If you built iOS or Mac apps for a living, your day happened in Xcode, and Xcode had none of this. Millions of Swift developers were copying code into a chatbot in another window and pasting the answer back, one snippet at a time, like it was still 2022.
Alex was built to close that gap. Founded in 2024 by Daniel Edrisian and backed by Y Combinator's Fall 2024 batch, the company described itself in four words: the Cursor for Xcode. Its product, Alex Sidebar, was a native Mac app that docked alongside Xcode and gave Apple developers what everyone else already had: chat that understood the open project, inline edits, code search in plain English, and quick fixes for build errors.
Ten months after its public launch, OpenAI hired the whole team to work on Codex, its coding agent. For a three-person company, that is one of the cleanest arcs in recent developer tools: find the platform everyone else skipped, build the missing tool well enough that the biggest AI lab in the world wants the people who made it, and land there before your second birthday as a company.
The story is worth retelling a year later, because the playbook still works.
The editor nobody could fork
To see why Xcode went years without a real AI assistant, look at how the other editors got theirs.
Cursor's founding move was a fork. VS Code is open source, so a startup could take the whole editor, rebuild the core interactions around an AI model, and ship it as a new product. Copilot had it even easier: VS Code has a deep extension API, and Microsoft owns both sides. Every AI coding startup of that era leaned on one of those two doors being open.
Xcode has neither door. It is closed source, so there is nothing to fork. And since Apple locked down third-party plugins years ago, its official extension surface is too narrow to host anything like an AI assistant. The standard playbook was simply unavailable. Most founders looked at that wall, noted that Apple developers are a smaller market than "everyone who uses VS Code," and moved on.
Edrisian's answer was to stop trying to get inside Xcode and sit next to it instead. Alex Sidebar ran as its own macOS app, attached itself to the Xcode window, read the project, and applied changes straight into the code. From the developer's chair it felt like Cursor had been grafted onto Apple's toolchain, down to the muscle memory: Cmd+L to chat about the selected code, Cmd+K for an inline edit, Cmd+P to search the codebase in natural language, then apply the change without leaving the editor. Anyone who had touched Cursor already knew how to use it. That was the point.
It helped that the founder had spent his career inside this ecosystem. Edrisian built the iOS app for ElevenLabs, was a founding engineer at FlutterFlow (YC W21), and interned at Apple, Google, and Meta before starting Alex. By the time he sat down to build a tool for Apple developers, the platform's quirks were familiar ground.
What three people shipped
The feature list grew fast for a team that never passed three people.
The core was the chat and edit loop, powered by whichever frontier model the developer preferred. Alex was model-agnostic from the start, with Claude and OpenAI's models both on tap, which meant the product got better every time any lab shipped an upgrade. On top of that came inline completions with tab-to-accept, drag-and-drop image to code for turning a design mockup into SwiftUI, and one-click resolution of build errors.
The deeper work was the agentic side. Alex could migrate code to newer Swift versions, suggest architecture changes, and refactor its way through cascading errors rather than fixing one line at a time. By the 3.0 release in spring 2025, it hooked into compilation itself, so the assistant could build the project, read what broke, and iterate. That build-test-fix loop is what separates a chat window from something that actually does the work, and it is much harder to deliver when the IDE gives you no official way in.
Developers noticed. More than 1,200 signed up in the first week after the YC launch, with hundreds using it daily almost immediately. Over the following months the user base grew to include mobile engineers at Duolingo, Notion, LinkedIn, Robinhood, and ByteDance. For a product aimed at professional iOS developers, names like those are the whole game: these are people who ship apps used by hundreds of millions, working inside the most demanding mobile organizations in the industry, choosing a sidebar built by three people in San Francisco.
Investors noticed too. In August 2025 the company announced a $5.3 million seed round, a bet that the wedge could widen into the default AI layer for Apple development.
Apple moved, and so did OpenAI
Then the platform owner woke up. At WWDC in June 2025, Apple announced that Xcode 26 would build in support for ChatGPT and other AI models directly, no third-party tool required. The capability Alex had spent ten months proving that Apple developers wanted was now headed into the IDE itself.
Founders in developer tools know this moment. When the platform absorbs your category, it validates the thesis and threatens the business in the same breath. What happened next says a lot about how much the thesis was worth: in September 2025, Edrisian announced that the entire Alex team was joining OpenAI's Codex division, the group building the company's coding agent.
The logic on OpenAI's side is easy to reconstruct. Codex's ambition is to work everywhere developers work, and Apple's ecosystem is the hardest major environment to reach, for exactly the reasons Alex had already solved. The three people who had shipped a production AI assistant into a closed IDE, kept it fast, and won professional users at Duolingo and Robinhood had more experience with that specific problem than anyone else alive. Rather than rebuild that knowledge in-house, OpenAI hired it.
The wind-down was handled the way you would want. New downloads ended on October 1, 2025, existing users kept working, and the team was public and clear about what would and would not continue. The app that people had built businesses on kept running while its makers moved on to build the successor inside OpenAI.
The ten-month playbook
Strip the story to its structure and it reads like a repeatable set of moves for anyone building developer tools in the AI era.
Pick the corner the giants skipped, not the one they crowded. Every AI editor startup piled into VS Code because the door was open. Alex chose the ecosystem where the door was welded shut, which meant no competition and a user base with real money at stake. The App Store economy runs through Xcode. The developers stuck there were underserved precisely because serving them was inconvenient.
Borrow an interface people already know. Copying Cursor's keybindings was not laziness, it was positioning. "The Cursor for Xcode" told an iOS developer everything about the product in one phrase, and the identical shortcuts meant the pitch and the onboarding were the same thing.
Ship the harness, not the model. Alex never trained a foundation model. It built the connective tissue between frontier models and a hostile IDE, and let the labs' progress compound its product for free.
Move like the clock is running, because it is. A wedge product in a platform's shadow always has one. Apple was eventually going to put AI in Xcode; the only question was whether Alex would matter by then. Three people went from launch to must-have tool for engineers at some of the biggest apps in the world in under a year, and when the clock struck, they had built enough that the outcome was a hire by OpenAI rather than a quiet shutdown.
That last point deserves the emphasis, because founders often read platform risk as a reason not to build. Alex is the counterexample. The risk was real and it arrived on schedule. The team outran it anyway, and the speed itself is what created the exit.
Where it landed
Edrisian has since left OpenAI to start his next company, Blackstar Computers, this time in AI hardware. The Alex app is frozen where the team left it, still running for the developers who had it installed.
As an Orange Collective portfolio company, Alex is the fund's clearest sprint story: a Fall 2024 YC company, a public launch that November, a $5.3 million seed the following August, and a team inside OpenAI by September. Somewhere in the AI features now landing in every Apple developer's editor is the trail of a sidebar that three people bolted onto an IDE never designed to let them in.
