APREX · EST. 2026
You're already behind
While you're still opening apps, switching tabs and googling your way to answers, the world's biggest tech companies have already decided: that way of using the internet is finished. Done. Dead. They're building something new right now — without asking you, without asking publishers, and definitely without asking their competitors.
This isn't a prediction. This is June 2026.
What is intent-first computing?
In the old model, you navigate. You open a website, search for something, filter results, compare options, click through. You do the work.
In the new model, you express a goal — and an AI agent does the rest.
> "Find me a used electric car under £25,000, less than five years old, within 50 miles of London, and book a viewing this weekend."
That's intent-first computing. You describe what you want. The agent navigates, filters, compares and acts on your behalf. Apps become background infrastructure — something you never see or touch.
The three giants and what they're building
| Platform | Agent/Product | Core Function | Status (June 2026) |
|---|---|---|---|
| Gemini Spark | Personal AI agent on dedicated VM, 24/7, integrated with Gmail/Drive/Docs | Launched to AI Ultra users in the US | |
| AI Mode + Daily Brief | Agentic search with query fan-out + personal morning digest | Rolling out from 19 May 2026 | |
| OpenAI | ChatGPT Atlas | AI-powered browser | Available at chatgpt.com/atlas |
| OpenAI | ChatGPT Shopping | Product discovery + "Buy it in ChatGPT" | Live, merchant info at chatgpt.com/merchants/ |
| Perplexity | Comet | AI browser with built-in agent integration | Available at perplexity.ai/comet |
| Open source | Browser Use | Open source browser agent | Active on GitHub |
Google: Your new digital assistant lives in your cloud
At Google I/O on 19 May 2026, Google unveiled what could be the biggest transformation in the company's history since the search engine itself.
Gemini Spark is not a chatbot. It's a personal AI agent that lives on a dedicated Google Cloud VM, around the clock, every day of the week. It's connected to your Gmail, Google Drive, Docs and Calendar — reading, organising and acting on your behalf. According to Google, Spark runs on Gemini 3.5 with what they call the Antigravity harness, with support for the Model Context Protocol (MCP) arriving later this summer. For now, only Google AI Ultra subscribers in the US have access (blog.google).
AI Mode in Google Search introduces something Google calls query fan-out: a single request from you automatically triggers multiple parallel searches in the background. The search engine is no longer trying to understand your keyword — it's trying to understand your intention.
Google Daily Brief delivers a personalised morning summary based on your Gmail, Calendar and Tasks. From 19 May, this is available to Google AI Plus, Pro and Ultra users.
And Chrome is becoming something far greater than a browser. At I/O, Google launched what they call the "agentic web" — new web APIs that allow AI agents to interact with web pages in structured ways (developer.chrome.com/blog/chrome-at-io26).
OpenAI: The browser is now a shop
OpenAI isn't watching from the sidelines.
ChatGPT Atlas is OpenAI's answer to the future of browsing — an AI-powered browser experience available at chatgpt.com/atlas. Where Google builds the agent into search, OpenAI builds it into the browser experience itself.
But the really big shift is what's happening with commerce.
ChatGPT Shopping lets you discover products directly inside a ChatGPT conversation. You describe what you need, and ChatGPT presents relevant products. Merchants can register at chatgpt.com/merchants/ and upload product feeds according to a spec published at developers.openai.com.
Then there's the Agentic Commerce Protocol — a collaboration between OpenAI and Stripe for AI-powered commerce. The result is called Instant Checkout, which lets you buy products directly inside a ChatGPT conversation. OpenAI calls it "Buy it in ChatGPT" (stripe.com/newsroom and openai.com/index/buy-it-in-chatgpt/). You don't need to leave the conversation. You don't need to visit a shop. The agent handles the transaction.
FACT BOX: What is MCP?
Model Context Protocol (MCP) is an open standard that lets AI agents connect to external tools, databases and services in a structured way. Think of it as USB for AI — a universal connector that lets agents plug into everything from email to enterprise systems. Google Gemini Spark will get MCP support during the summer of 2026.
Perplexity and the open field
Perplexity is an underdog that refuses to act like one. Perplexity Comet is the company's AI browser with built-in agent integration (perplexity.ai/comet). Where Google and OpenAI have billions in resources and existing user bases, Perplexity is betting that users tired of big tech will want an alternative.
In the open market, Browser Use is growing — an open source project on GitHub (browser-use.com) that lets developers build their own browser agents. It's the technical foundation for an entire generation of new agent applications that haven't been built yet.
> PULLQUOTE: "The web is shifting from places you visit to tasks you get done. Nobody is asking whether that's okay with you."
Who loses?
When AI agents handle search, shopping and information gathering, someone is left holding the bill.
Publishers and news outlets are already at war with Google. In the UK, the Competition and Markets Authority (CMA) is pushing for regulation of Google's AI summaries in search results — summaries that give users answers without clicking through to articles (The Guardian, 3 June 2026). When the agent reads the newspaper for you and summarises it, who pays the journalist?
Online retailers and e-commerce must rethink everything. If customers shop through ChatGPT conversations and never visit your website, you lose control of the user experience, brand building and the direct customer relationship.
Traditional apps — from travel apps to product comparison services — risk becoming irrelevant if agents do their job better and faster.
TIMELINE: The agent revolution in 2026
- 19 May 2026: Google I/O. Gemini Spark, AI Mode, Daily Brief and agentic Chrome launched.
- May 2026: OpenAI launches ChatGPT Atlas, Shopping and Instant Checkout with Stripe.
- June 2026: Perplexity Comet available. CMA pressure on Google in the UK.
- Summer 2026: Google Gemini Spark gets MCP support. Agentic web APIs roll out more broadly.
- Ahead: The battle over who owns your "intent" — and the money that follows — escalates.
What does this mean for you?
You'll soon stop thinking about which app to open. You'll think about what you want to achieve — and say it aloud or type it. The agent handles the rest.
That sounds fantastic. And in many ways it is. But it also means that one player — Google, OpenAI or Perplexity — sits between you and the entire rest of the internet. They decide which answers you get, which products you see, and which services exist for you.
That's a lot of power to hand to one assistant.
BOTTOM LINE
None of the major players are asking for permission. Google, OpenAI and Perplexity are each building their version of the future internet — a world where you express goals, not navigate to apps. The technology is real, the products are live, and the effects on publishers, retailers and competitors are already measurable. The big question is no longer whether agents take over, but whose agent you end up trusting — and what you give up in return.
Source assessment: This article has been verified against 8 open primary sources (blog.google, developer.chrome.com, openai.com, stripe.com/newsroom, chatgpt.com, perplexity.ai, browser-use.com, theguardian.com) and 0 independent analyses. All factual claims are based on publicly available documentation from the companies in question.