TLDR: It’s like giving Samantha from “Her” a phone number, email address, a few of your accounts and a few sub-agents to do real work.

Hi HN! We are Joseph and Julius, and we built personalized remote coworkers in the cloud. We call these remote coworkers “Personas” and we give each one a name, email address and phone number. Today, we’re launching Personas with “skills” for reminders, scheduling and email management tasks (as seen on our home page), and our goal is to add at least one new skill every week. We’re curious to hear what you all want to use Personas for so please email us your thoughts at founders - at - withtime.ai.

Backstory:

We were both convinced the chatbot experience was stale and repetitive and decided it was time to design an AI system that fit our existing workflow. Most of the context for how real tasks should be done lives within our human-to-human interactions (irl or online), and chatbots are completely walled off from that. And so you end up retyping everything or copy-pasting notes from calls.

We built the “Persona” because we wanted an AI one could simply “cc” on emails, call in the morning for updates and (soon) add to Zoom calls. Then, when one wants some work done, the context is already there.

How it works:

When you sign up, you choose a name and profile pic for your Persona. It’s granted an email address and a phone number. Behind the scenes, we hydrate its brain – a knowledge-graph meets social-graph meets vector-database – to store information about people, files, messages, calls and more. As the world interacts with your Persona, it adds to this novel memory system so that when it’s time to do a task, it reflexively pulls the needed context instead of asking you for it again… every time.

How is this different from other AI Executive Assistants?

The other AI “Executive Assistants” are a single AI Agent (i.e. system prompt + LLM + a few tools) responding to texts or emails one at a time - often re-using the same phone number or email address across all users. Most don’t have a deep memory and require the user to predefine the user’s preferences in a system prompt or a classic system settings menu.

By contrast, Personas have a consistent, long-running life, and learn exclusively from interacting with humans over emails, texts, and phone calls - they don’t have a traditional context-dump system prompt – instead learning preferences “on the job.” On top of that, the unique phone number and email address per assistant enables something pretty exciting - imagine your Persona calling or emailing other people to complete your todos.

Who are we?

We met freshman year on the Stanford rowing team (‘20) and have been friends ever since. We are based out of Palo Alto, raised some money and are actively hiring – reach out at founders - at - withtime.ai if you’re interested in learning more.