On May 20, 2026, TechCrunch profiled IrisGo, a startup that raised a $2.8 million seed round led by Andrew Ng’s AI Fund and is building what it calls the AI desktop buddy you never knew you needed. The product is named Iris — Siri spelled backward — and it watches your desktop activity, learns your daily workflows, and then executes them with little to no prompting.
The core promise is “demo once, and it knows.” In a hands-on demo, it learned to order a latte from Philz Coffee: entering payment details and completing the checkout, after which Iris could repeat the entire flow autonomously. Co-founder Jeffrey Lai, a former Apple engineer who helped build the Chinese-language version of Siri, is blunt about the market: “Our target audience is knowledge workers — white-collar companies.”
An Agent That Learns by Watching
IrisGo’s engine runs in three steps: Watch (gather context across windows, apps, and systems), Learn (turn observed behavior into workflows and decision logic without manual setup), and Act (run event-driven, cross-app, self-optimizing flows). Beyond learning from demonstration, it ships with a built-in skills library — email drafting, invoice processing, report building, document summarization — and the official site lists roughly 20 importable community skills spanning Email Triage, Lead Scoring, Contract Summary, and more. There is also a coding assistant positioned against OpenAI’s Codex and Anthropic’s Claude Code. The company’s own demo workflow is more telling than the latte: a Morning Brief Builder that digests Hacker News, summarizes a podcast episode, checks stock watchlists on Yahoo Finance, adds groceries on Instacart, and drafts a Gmail summary — one event-driven flow reaching across five applications.
This demonstrate-don’t-prompt contract points the same direction as Thinking Machines’ research on interaction models: the interface to AI is shifting from typing prompts to showing behavior.
On-Device Inference and Privacy Design
What IrisGo sells is “proactive” AI — anticipating needs before you express them. That requires sitting close to your desktop for long stretches, which makes privacy architecture a core product decision rather than a footnote. Most inference runs locally on your PC’s NPU and GPU; larger tasks go through a hybrid cloud path, but the company states that cloud processing “only occurs when explicitly authorized by the user and uses end-to-end encryption.” The Windows beta is available for download now, with the Mac version listed as coming soon on the official site; TechCrunch reports betas for both platforms have launched.
Funding and the AI PC Bet
The $2.8 million seed was led by Andrew Ng’s AI Fund, with NVIDIA Inception, Google for Startups, Pegatron Ventures, and Intel also in the supporter list; the Ng connection came through a shared Carnegie Mellon University network. Lai co-founded the company with Yichen Chu as COO, operating under Irixion Corp. The more interesting move is distribution: IrisGo signed a deal with Acer to preload the app on new devices and hopes to land similar agreements with other laptop makers. AI PCs need a killer app to justify the NPU, and IrisGo is betting on exactly that slot — becoming system-level digital butler infrastructure for the AI PC era rather than another chat window.
Competition and Risks
IrisGo is not alone in the “watches you work” desktop-agent lane, but it bundles three things: learning by demonstration, on-device inference, and OEM preload. The risks are just as clear. An agent that reads and buys on your behalf makes errors that cost real money — a mis-learned payment flow is far more expensive than a wrong chat reply. Trust has to be earned step by step, and the design of permissions and confirmations — moving from low-risk skills like summarizing and drafting to high-risk actions like paying and ordering — will decide whether products like this survive. There is also the platform question: Microsoft, Apple, and Google all ship their own system-level assistants, and a preloaded third-party agent lives or dies by how much surface the OS vendors leave it. And the privacy claims won’t rest on a website statement forever; independent scrutiny is coming.
Sources
- IrisGo, a startup backed by Andrew Ng, looks to become the AI desktop buddy you never knew you needed — TechCrunch
- IrisGo official site
- AI Fund portfolio
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.