Former Humane engineer launches a proactive AI assistant

Our phones are bloated with apps, unread notifications, and fragmented schedules. You have a meeting in your calendar, a dinner spot buried in WhatsApp, and a friend's dietary preferences scattered across your messages.

A new AI assistant app called Poppy wants to fix that chaos. Built by San Francisco startup Second Nature Computing, the app connects your calendars, inboxes, and messages into a single feed that anticipates what you need before you even ask.

An assistant that actually pays attention

Most modern AI tools wait around for a prompt. Poppy is designed to do the heavy lifting in the background by learning what is happening in your life across multiple services.

At launch, it integrates with Apple Calendar, Google Calendar, Gmail, Outlook, iCloud Mail, Apple Health, Reminders, Contacts, WhatsApp, Uber, and Instacart. If you grant it location access, Poppy connects the dots on the fly.

For example, if your schedule has an unexpected 30-minute opening and you are sitting near a park, Poppy will ping you suggesting a quick walk. If you are scheduling brunch with a friend, it can pull their dietary restrictions from an older conversation and filter restaurant picks accordingly. It also monitors flights and pushes medication reminders.

The app was created by Sai Kambampati, a former software engineer at Humane who holds a master's degree in human-computer interaction. Kambampati says modern models finally make ambient, proactive computing practical outside of a research lab.

Workarounds and privacy questions

Giving an AI total visibility into your digital existence comes with significant friction. The biggest technical hurdle is Apple's walled garden.

To read iMessage, Poppy currently relies on a companion Mac application. Third-party access to iMessage has always been heavily restricted by Apple, meaning this workaround could easily break if the platform tightens its grip.

Privacy is the other elephant in the room. The startup says user data is encrypted at rest and that it enforces a zero-retention policy with its cloud LLM providers. But the ultimate goal is running everything on the phone itself.

Kambampati hopes that within two to three years, local hardware and small models will be capable enough to run the entire system on-device without ever touching a remote server. The four-person team just closed a $1.25 million pre-seed round led by Kindred Ventures, with backing from DeepMind's Logan Kilpatrick.

Why it matters

We think Poppy represents the right design direction for personal AI. Chatbots that require manual typing are basically search engines with better grammar. True personal assistance should be ambient, contextual, and proactive, stepping in when there is an actual gap in your schedule or context you forgot.

For users, the app offers immediate utility by gluing together disconnected services. But handing over your location, email, and private messages to a four-person startup requires immense trust, especially while the processing still relies on cloud servers.

The deeper threat is platform risk. Google and Apple already own the mobile operating systems and are building native intelligence into Android and iOS. A third-party utility relying on a Mac hack to parse text messages is inherently fragile.

Poppy gives us a compelling glimpse of what everyday AI should feel like. The question is whether it can build a loyal audience before Apple and Google simply bake the exact same proactive features directly into the home screen.