Hark, a startup founded less than a year ago by serial entrepreneur Brett Adcock, has launched Hark Pro in wide release. The service is available for free with a subscription tier for heavier users.
The company frames its mission as building a user interface for AI rather than pursuing AGI or grand scientific breakthroughs. At the core of Hark Pro is a model trained specifically for computer use, and the product’s promise is essentially that it will use your computer for you.
User experience and features
Hark Pro provides a full-screen experience with a homepage-like layout: a central chat dialog and a feed populated by suggested actions the assistant might take. The product includes visual elements the company calls “panels,” small dashboards that surface information the user cares about. A custom background image adjusts to local weather and time.
Like other AI assistants, users connect Hark to their digital accounts — email, calendar, local drives, credit cards, and so on. Once connected, Hark can perform digital tasks in addition to answering conversational queries.
Abidur Chowdhury, who leads design at Hark and previously worked at Apple, demoed the product. The assistant proactively flagged expenses awaiting approval, emails that needed replies, a meeting scheduled with this reporter, and suggested booking a flight for an upcoming trip, offering to handle the booking. Chowdhury also described a case where the assistant detected a colleague’s need to renew a vehicle registration at the California DMV and completed the task through the agency’s website.
A notable UX choice is that Hark shows users a small window displaying how the agent navigates the web — an operation many assistants do invisibly. Chowdhury says surfacing these actions is intended to build user trust that the agent is executing the right steps.
Privacy positioning
Hark positions itself against big consumer platforms such as Meta and OpenAI, with the company stressing it is not in the business of selling ads or harvesting data. Chowdhury points out that to get the most out of these models, users will need to share large parts of their digital lives, which raises real cybersecurity and privacy questions. When asked about profiling concerns, he framed Hark’s approach as task-oriented: the system aims to understand users to help with tasks and remove burdens, rather than to build intrusive profiles of friends and family.
Trade-offs of a computer-trained model
Hark’s model is explicitly optimized for computer tasks, which makes it faster and cheaper for those operations, according to Chowdhury. However, it remains an open question how much general knowledge or broader capability is sacrificed compared with frontier large language models.
Chowdhury invited comparisons with competitors: “Please compare it to the others, be critical,” he said. “It might not be faster than Muse, but it will give you a better result, and it will work end-to-end instead of asking you to take over at a point.”
Hardware plans: an AI-native device in 2027
Software is only half of Hark’s stated plan. Chowdhury, known for his device design work at Apple, said the company intends to ship an AI-native device in 2027 to support its software. He declined to specify form factor details and voiced concerns about the “creep factor” of wearable cameras. Still, he promised: “The first product that we’re building is a product that’s never ever existed before.”
Chowdhury also quipped about the current state of interaction: “It’s so hilarious to me that you have AI now, you have a chatbot, and you type stuff in with a keyboard. Fifty years later and the future of computing is still a terminal.”
Hark Pro is now widely available; whether its privacy-focused positioning, computer-specialized model and planned hardware will translate into real-world advantage remains to be seen over the coming months and years.
