Anonymous founder Signull launches Sky, an agentic AI home screen replacing the static iPhone app grid
Apr 20, 2026 with Signull
Key Points
- Sky replaces the static iPhone home screen with an AI-generated feed powered by 22 background agents that draft emails, prioritize tasks, and surface relevant information without user prompts.
- The startup faces steep infrastructure costs running continuous agentic inference and plans contextual, location-aware advertising embedded in the feed as its revenue model, analogous to Facebook's News Feed.
- Signull argues Apple's broadcast software model prevents nondeterministic personalization at scale, positioning Sky's widget-based approach as faster to market than building inside LLM app stores.
Summary
Read full transcript →Sky
Signull, an anonymous technologist with a substantial following on X, is launching Sky — an iPhone app that replaces the static home screen grid with an AI-generated, continuously updating personal feed, delivered through two iOS widgets.
The core argument is that the iPhone home screen hasn't meaningfully changed in twenty years. Sky installs a medium "wild card" widget, which surfaces what the app judges most relevant at a given moment, and a large "for you" widget running a scrollable feed. Twenty-two background agents run continuously, processing signals like incoming email to draft replies, bucket tasks, and rank items by importance — without the user ever opening a prompt.
“The iPhone home screen is twenty years old. Two decades and it's just static icons. We're building kind of the new iteration — the Facebook News Feed 2.0 — that's entirely AI generated about your life, highly personal, and lives directly on your home screen. We have 22 agents that work continuously to generate content for that feed.”
The product is deliberately built within Apple's existing widget APIs rather than against them. That's both a constraint and a positioning choice. Apple's software is inherently broadcast — written once, runs identically for everyone. Sky's experience is nondeterministic by design, meaning every user gets a genuinely different feed. Signull argues Apple, managing several billion users, can't easily make that transition; a small team can move faster by embracing the nondeterminism rather than fighting it.
Monetization and distribution are the two open questions. Signull is candid that background agentic inference is "nontrivially expensive" — unlike ChatGPT or Claude, where cost is triggered by user requests, Sky's agents run constantly on the user's behalf. The company has raised some capital, amount undisclosed. On revenue, Signull points to advertising as the likely model — specifically contextual, location-aware ads integrated naturally into the feed rather than interruptive units. The analogy he reaches for is the Facebook News Feed, reimagined as fully AI-generated and personal, living one glance away on your home screen.
Getting users to install two widgets is the practical friction point, and Signull acknowledges it. The bet is that the feed paradigm is already deeply familiar, and the value delivered — intelligence surfaced without asking for it — is sufficient to clear that bar.
On building inside LLM ecosystems like the ChatGPT app store, Signull is skeptical. Unless a developer brings genuinely proprietary data, he doesn't see a durable incentive structure for third-party apps living inside an LLM. The comparison he draws is to flashlight apps on the original iPhone — useful for a moment, then absorbed.
Sky is early-stage and in an experimental phase. No user numbers or launch date were disclosed.
Every deal, every interview. 5 minutes.
TBPN Digest delivers summaries of the latest fundraises, interviews and tech news from TBPN, every weekday.