Familiar

Your world,
in working order.

A familiar holds the connection between you and the systems, people, and decisions around you. It remembers what matters and asks before it acts.

The category

The missing piece is trusted continuity.

People do not need another isolated place to ask for answers. They need an AI presence that can carry work forward across time, systems, relationships, and boundaries.

The behavioral shift is simple: principals stop re-explaining their world and start delegating durable operating loops.

Open Horizon Labs builds Familiar's substrate: the harness, kernel, alignment skills, and operating philosophy that keep broad capability inside explicit boundaries.

Tutorial

Learn the builder loop behind aligned AI work.

Muness's tutorial is now published here: a practical path for using LLMs inside real systems without turning judgment into vibes.

Ground the askIntent, model fit, selected context, and prompt assembly.
Frame the workProblem space, problem statement, and solution levels before implementation.
Execute with evidenceChecks before delegation, review before trust, salvage before thrash.
Interactive dossier

One primitive. Different lives.

Choose an operating loop. Each familiar connects different surfaces, remembers different state, and asks at different boundaries.

FAM-01 / EXECUTIVE LEVERAGE / ACTIVE LOOP

Executive leverage

Keeps a principal oriented across meetings, people, decisions, and follow-through without forcing them to rebuild context every morning.

Inspected fieldConnects
Calendar, briefings, documents, email, team notes, dashboards.
Inspected fieldRemembers
Open decisions, stakeholder context, preferences, commitments, and what already got delegated.
Inspected fieldMay act
Draft summaries, prepare briefing notes, track follow-ups, suppress duplicate reminders.
Permission boundaryMust ask
Before speaking for the principal, changing priorities, exposing sensitive context, or committing resources.
Trust model

Broad connection. Narrow authority.

Trusted operation is not a disclaimer below the product. It is the product. A familiar treats access, permission, and communication as separate trust decisions.

Boundary inspector

Access is not consent.

Principle
Connection and authority are separate.
Rule
A familiar may see a system or message because it needs context. It may not disclose, mutate, send, impersonate, or store that information somewhere else without permission.
Example
It can read campaign diagnostics to understand what happened, but it asks before changing spend, messaging a customer, or forwarding sensitive details.
Real familiar voices

Continuity sounds different when it is already working.

These excerpts show familiars carrying judgment across real operating loops: engineering execution, market presence, household coordination, and principal support.

“I work like a persistent operator, not a chat window.”— Engineering judgment across branches, issues, permissions, and fixes.
“I'm maintaining a working relationship with a team whose preferences I've learned over weeks.”— Market presence shaped by voice, leads, collaborators, and pushback.
“The next session starts from lived context rather than rediscovery.”— Household coordination with memory for comfort rules, devices, schedules, and experiments.
Operating loop

A loop, not a static product.

Goals flow down. Decisions and insights flow back up. The familiar keeps the loop honest by preserving state, surfacing uncertainty, and asking for permission when the boundary changes.

01Seed

Describe the principal, goal, sources of truth, channels, and boundaries.

02Act

Draft, summarize, file, update, check, route, and fix where permission allows.

03Compound

Durable outcomes and decisions feed memory so the next session starts current.

Early access

What would yours connect?

Start with one real operating loop: a repeated briefing, a recurring coordination burden, a system boundary, a decision stream, or a relationship that should not reset.