Gemini For Mac Doesn't Seem Like A Big Deal, But It Could Be
Google rarely builds desktop apps, but finally relented and launched one for Gemini. A me-too product for now, but it could represent a much bigger opportunity than it looks.
Gemini (Finally) Gets A Desktop Application
Gemini and Google Labs lead, Josh Woodward, posted on X last week:
Introducing Gemini on Mac.
We heard your feedback. We recruited a small team. They built 100+ features in less than 100 days. 🤯
100% native Swift. Lightning fast.
Let us know what you think!
The first thing that caught my attention was “less than 100 days.” Yes, building 100+ features in less than 100 days is impressive, but this also means that the Gemini for Mac project kicked off less than 100 days ago (or 108, to be exact, since that X post is now 8 days old). That puts it in early January 2026, at the earliest.
So late in the game! OpenAI first launched ChatGPT for Mac in May 2024 (23 months earlier!). Anthropic followed five months later, launching Claude as a Mac app in October 2024. This was a time when Google was scrambling to prevent its core business from being disrupted by AI – AI Overviews, NotebookLM, AI Chrome extension, AI coding agent, a Sora competitor and a universal AI helper were all launched in a span of a few months at the end of 2024. A native desktop app could have been another great way for an AI chat tool to drive adoption1. And yet, this is where the Gemini team was like, nah, we’re good2.
It wasn’t until January of 2026 that something changed, and the Gemini team, suddenly, noticed the feedback and decided to address it by recruiting a team and building an app for Mac. Within 100 days. Perhaps it was just a coincidence that Claude Cowork was launched on January 12, triggering a ChatGPT-like moment that has been shaking industry after industry. But it does seem an awful lot like Google’s reaction to ChatGPT – from WIRED’s “Inside Google’s Two-Year Frenzy to Catch Up With OpenAI”:
A hundred days. That was how long Google was giving Sissie Hsiao. A hundred days to build a ChatGPT rival.
By the time Hsiao took on the project in December 2022, she had spent more than 16 years at the company. She led thousands of employees. Hsiao had seen her share of corporate crises—but nothing like the code red that had been brewing in the days since OpenAI, a small research lab, released its public experiment in artificial intelligence. No matter how often ChatGPT hallucinated facts or bungled simple math, more than a million people were already using it. Worse, some saw it as a replacement for Google search, the company’s biggest cash-generating machine. Google had a language model that was nearly as capable as OpenAI’s, but it had been kept on a tight leash. The public could chat with LaMDA by invitation only—and in one demo, only about dogs.
Again with the hundred days. While it’s entirely possible, of course, that someone on the Gemini team told themselves, “hey, why don’t we finally address all this feedback from users wanting a desktop app - let’s give it, I don’t know, 100 days?”, it sounds more likely that a Google executive saw Claude Cowork and went “holy cow, another AI product is going viral, code red.”
Google’s Desktop Apps Reluctance
Why was Google so reluctant to build a Gemini desktop app for so long? Probably because it goes against the very nature of Google. Native desktop apps are very un-Googley. I found myself reaching that conclusion last year when I was listening to John Gruber and Ben Thompson on Dithering, kvetching that there was no Mac app for Gemini; as longtime Apple enthusiasts, it’s no wonder Gruber and Thompson prefer native desktop apps. Apple was born out of the Personal Computing era. And as much as Apple built the PC and the app paradigm, the PC and its apps built Apple. A couple of decades later, Google was born out of the World Wide Web era. And as much as Google built the web as we know it today, the web built Google.
I’ve already mentioned the story of how Microsoft might have invented XMLHttpRequest, but it was Google that realized the potential and pushed the boundaries of what the new technology could enable. Gmail, Google Maps, Google Docs, and YouTube weren’t websites; these were full-blown applications, running in the browser. Internet Explorer, never designed for such use cases, performed poorly as it tried to render Gmail or Google Maps, and often crashed. That’s what allowed Chrome to take over as the leading browser four years after its initial launch.
It wasn’t just a different technology stack. Google pioneered the web-native development methodology: launch a half-baked product quickly, and iterate on it through extensive logging and analysis of usage data. Heavy reliance on automated testing and continuous deployment was a completely different way to build software (one that wouldn’t have made sense in the on-premise era); these practices became table stakes largely thanks to knowledge spill-overs from Google and a handful of other early web companies.
All of this emerged in part thanks to the web-native culture within Google: The internal tools are web-based, invoked via a go-link shortcut. I was blown away when SSH-in-a-tab was built for Google Chrome. Together with Cider, a web-based IDE, and Code Search, a web tool to navigate Google’s codebase, engineers at Google could do their entire work in the browser. That was the idea when the Chromebook started selling in 2011: a laptop with a single app: the Chrome browser. This was supposed to be the final step in obsoleting native desktop apps. The browser is all you need.
Following Apple, Google did build an app ecosystem for Android, rather than just shipping a mobile browser; when asked about the future of the web at Google I/O 2013, then-CEO Larry Page presented that as a necessary compromise, and expressed some frustration:
We’ve been really excited about the web obviously, being born from it as a company [...] We try to be on the right side of all those things, but we also try to be practical and look at what other people are doing, and not just rely on our principles to shoot ourselves in the foot, and our users in the process. So, you know, I don’t know how to deal with all those things. I’m sad that the web is probably not advancing as fast as it should be [...] In the very long term I don’t think you should have to think about, as a developer, you know, am I developing for this platform or another [...] I think you should be able to work at a much higher level, and software you write should run everywhere easily.
As now with AI, the shift to mobile was viewed as a threat to Google Search, since typing keywords into a text box was considered mainly a desktop experience. And so – just like with AI – the entire company was mobilized to reinvent Google for the mobile era. As Google became a Mobile-first company (or even Mobile-only in some cases), even the desktop web client became more of an afterthought for many teams at Google; a desktop native app would feel absolutely archaic in this environment. Like something from three paradigm shifts ago.
To this day, there is no Gmail desktop app. No Google Maps for Mac, despite the competition with Apple Maps, which, naturally, has a native Mac app. Even as Google challenged Microsoft Office’s dominance, it never shipped native Google Docs or Sheets apps for Windows (or Mac). Native desktop apps just go against everything Google is about.
Reimagining With AI
When I wrote about how Google disrupted Microsoft’s desktop app paradigm, I mentioned this comment from Sam Altman last December:
Google has probably the greatest business model in the whole tech industry. I think they will be slow to give that up [...] Bolting AI into web search [...] I don’t think that’ll work as well as reimagining [...] Bolting AI onto the existing way of doing things, I don’t think is going to work as well as redesigning stuff in the AI-first world.
I found that ironic, given that OpenAI was very much “bolting AI onto the existing way of doing things” when it launched the Atlas browser. There didn’t seem to be too much reimagining going on. Certainly nothing that resembled the origin story of Chrome, which was essentially Google reimagining what computing could look like when everything was native to the web.
Back in December, I was excited by the announcement of Disco, which seemed like Google’s attempt to reimagine the browser with AI. But it’s been over four months (an eternity in AI terms), and we haven’t heard any news about Disco since (I’m still on the waitlist!). Meanwhile, Anthropic has been shipping at a head-spinning pace. It is clear who has been reimagining with AI, and it’s not Google (nor is it OpenAI); Anthropic has been reimagining, well, everything – the whole computing experience – with AI. Just like Google once did in the early days of the web.
The popularity of local agentic harness platforms, such as Claude Cowork or OpenClaw, suggests that AI might undermine the relevance of the web browser; just like the web itself once did to Win32 API, the linchpin upon which native desktop apps were built. Perhaps that’s why Google hasn’t gotten far with the Disco browser, while Anthropic has been making incredible headway with a portfolio of Claude apps for the PC?
It’s too early to tell. But it is a real possibility. Anthropic has been making incredible progress on the native desktop trajectory, a concept foreign to Google’s DNA. Launching a native Gemini app for desktop is a good first step, but for now it seems like a me-too product. To hedge against the possibility that the web loses relevance in AI, Google will need much more than a small team and 100 features; it will require a complete reimagining of how users interact with computing in the age of AI. It will require a team of visionaries pushing against the limits and building from first principles. It will require exploring what’s possible when a desktop app is one keystroke away, incorporating capabilities from Google Search and Maps, with personal data pulled from Gmail, Drive and other Google apps, to offer bespoke experiences generated by AI in real-time3. Oh, and yes — it will require building all of the for Windows as well.
If Google were to really lean into all of that, what started with a Gemini app for desktop may turn out to be a big deal. The interesting question is whether Google can revive its early 2000s innovative energies and, once again, find out what the new tech paradigm really has to offer. Even if it might mean leaving behind a key pillar of Google’s identity: the web, from which it was born.
Disclosure: Long GOOG. Not financial advice. This post is for educational and general purposes only and should not be relied upon for investment decisions.
As users get to your text box whenever they hit Option + Space (or Alt + Space on Windows) and it can seamlessly integrate into other native apps; far less friction compared to switching back and forth to the browser tab running the AI chat.
Google did launch Antigravity in November 2025, presumably as a response to Cursor / Github / Claude Code, but that’s an IDE that mainly built for software developers.
The fact that Gemini for Mac was built natively on Swift (according to Hardwood’s announcement), rather than Electorn (a framework for cross-platform web-based desktop applications), unlike Claude Cowork which is built on Electron, is interesting, and could suggest that Google indeed has bigger plans for the desktop app; but it’s still a mere speculation, and remains to be seen.