sometime

Social intelligence

The average person spends more time looking at screens than into someone’s eyes. Our time is harvested, packaged, and sold to the highest bidders.

Today, someone can live entirely from their home without setting foot outside. They can work remotely, get groceries delivered from Amazon, order meals through DoorDash, and be entertained by an infinite stream of content. Dating has turned into a humiliation ritual of swiping through hundreds of options. People spend less time with friends than ever before. Friction and serendipity have been removed from daily life.

Soon, powerful AI will surpass human cognition and dexterity. If all goes right, the result is a world of material abundance. In this new world, what remains scarce? Human connection.

Loneliness has its own straight lines on graphs...

Time spent alone since 2010 0% +20% +40% 2010 2015 2020 2025 15-29 +37% 30-44 +32% 45-59 +11% 60+ +8% Source: BLS American Time Use Survey
Searching for connection “how to meet people” 100 0 2010 2015 2020 2025 “meet new people” 100 0 2010 2015 2020 2025 “where to make friends” 100 0 2010 2015 2020 2025 “feel lonely” 100 0 2010 2015 2020 2025 Source: Google Trends, US, trailing 12-month average

Some will choose the dopamine machine and AI companions. Many more will crave the belonging and purpose found only in community. Sometime is building social intelligence to increase human connection.

The graph

Repeat time makes ties

50 leisure hours to a casual friend, 200+ to a close one; friendships live on repeat time together Hall 2019
you write

What this graph is: the digital twin of your social world, and why repeat time is the number sometime optimizes.

The product

The app

The agent drafting a dinner plan in chat
you write

The lead moment: the agent turning "dinner with Ox this Friday" into a plan. What is happening here and why it matters.

you write

How a plan flows end to end: ask, draft, invite, RSVP.

Flows
Library
The shape of it

How it works

you write

The overview: what happens from opening the app to standing somewhere with your people. The agent, the graph, the plan.

The user path

you arrive you the twin deepens every loop adds context onboard straight into a community community rhythm the weekly thing you go to meet people ties form in person circle grows new friendships to read sometime proposes plans, parties, intros plans happen tables get booked 1 2 3 4 5 6 the business lives here booking yield: OpenTable, Resy, SevenRooms, Tock

The context engine

WHO the graph relationships and their strength mutual friends, group chemistry who's in, who's drifting communities you belong to WHAT taste what you love doing, and with whom the group's shared taste and energy plans that worked before dealbreakers WHEN rhythms open windows across calendars weekly rhythms, the usual nights how far ahead the group plans WHERE place home bases and realistic radius central to the group venues and events in reach sometime reads all four together
you write

What these two figures mean in your words: the loop the product runs, and why the decision takes all four inputs at once.

Plans

intimate

you write

The default unit: a few people, a time that works, a spot. How the agent drafts one.

Parties

large gatherings

you write

Cover art, web invites, RSVP without the app. How a party spreads.

Communities

group identity

you write

The weekly rhythm you belong to, and where new people enter your graph.

Positions

What I believe

you write

Suggested topic: intent beats availability. One tap of "I want to do something" is worth more than a shared calendar.

you write

Suggested topic: the agent coordinates so people can connect. AI belongs in the logistics, never in the friendship.

you write

Suggested topic: who and what together. Recommending a place is easy; recommending the place AND the people is the actual problem.

you write

Suggested topic: events are social context. "Who from my circle is going" matters more than the listing.

you write

Suggested topic: software with weather. Why the app lives under a real sky that tracks the sun.

Under the hood

The system

Seven products in one app

Parties

covers, web invites, RSVP

a Partiful, a Luma

Messaging

groups, media, reactions

a full chat app

The agent

plans, venues, events

an AI social agent

Communities

spaces, subgroups, event chats

a community platform

Context

the graph, taste, memory

a context engine

Profile

identity, photos, people

a social profile

Calendar

windows, blocks, availability

a calendar
you write

The scope, in your words: what it meant to build all seven at once, and why the app needed every one of them.

0agent tools
0database migrations
0commits
The agent
venues, events, the graph, calendars, plans
The social graph
relationship embeddings, friend-of-friend reach
Group co-planning
a second agent living inside group chats
Growth surface
web invites, SMS, App Clip
The sky engine
solar-anchored palettes, blended in OKLab
The stack
Expo React Native, FastAPI, Supabase, LiteLLM
you write

A paragraph on how you actually built this: your process, the agents, what the numbers above mean to you.

The journey

The year

May 2026

you write

Left Dodge and Cox. One line on the bet you were making.

Summer

you write

Split time: expert equity-research data labeling to fund it, Sometime the rest. What grading models taught you.

The tangent

you write

The AI safety detour, and why it made you better at driving agents.

Sometime

you write

What building all of this solo taught you. The big takeaway.

Now

you write

Where you're pointed, and why.