famile

Companions for the daily work of staying well — metabolic care, recovery, and practice.

Ask Mira Try ninety seconds

Sukari

Calm enough to live with.

Sukari turns the day’s biomarker patterns into one doable action — and tells your care team only what shifted.

WHAT'S HERE

Care and practice, when useful.

Metabolic Care

Sukari

A daily companion for metabolic care that has to last.

One thing today. Biomarker patterns collapse into a single, doable action — never a wall of metrics to interpret first.
Try it before it counts. A short practice run primes the decision before it's live, so adherence is built by repetition, not willpower alone.
Only what shifted. Care teams see what diverged and why — a note, not another screen to watch.

Details

Open Sukari →

1 action

Daily focus

90s

Practice before live

By exception

Care-team signal

Recovery Intelligence

Orbura

Recovery intelligence for people, and the teams around them.

Signal, not noise. Recovery intelligence drawn from the lifestyle and biometric streams you already own — distilled, not dumped.
Scales with the unit of care. From one person to a whole squad's load management, Orbura grows with whoever's responsible.
In the clinician's language. Coaches and clinicians receive context they can act on — no translation layer, no second system.

Details

Open Orbura →

Days ahead

Early warning

Live

Plan adaptivity

1 → squad

Unit of care

Practice Continuity

Ardum

The shape of your practice — intention before inventory.

Intention before inventory. Begin with what you are trying to make space for — not a catalog of places and filters.
One decision at a time. Mira asks only for judgment the system cannot supply. Everything else stays in her care.
Earned agency. Monitoring, holds, coordination, and booking require explicit, scoped consent — never implied by memory.

Details

Open Ardum →

Episode

Primary object

One at a time

Decisions asked

Secondary

Booking

The bench

Open research, built in public — not care apps.live ledger →

Cellular Modelling

signal in
Basal expression for an unseen cellular context. 300 CRISPRi targets to predict. No perturbation data from this context.
observed
The 2025 atlas overlaps only a handful of these targets. Priors transfer where real signatures exist; elsewhere the model must fall back honestly.
reasoning
The nearest honest baseline — real control-cell resampling — sets the floor. Layer a per-target perturbation prior with log1p transport, and measure headroom to the ceiling before adding complexity.
one thing today
Run the Replogle-prior build: real per-target signatures for 272 of 300 targets, transport on, submit.
to the Observatory
Run published as it landed: cell-eval score, rank, audit flags, literature evidence, briefing. The deliberately-failing probe run stays on the wall beside it.

Kytos

How an unseen cell will answer perturbation, read from its quiet state.

Basal only. Predictions are made from the unperturbed state alone — no peeking at the perturbation data for the context being tested.
The Observatory. Every run publishes its metrics, audit flags, and briefing as it lands. Failures stay on the wall beside the rest.
Ceiling as compass. Headroom to the measured ceiling decides the next experiment, not rank.

Details

Open Kytos →source ↗

360k

Cells

300

Targets

Every one

Runs shared

Agentic Science

paper in
An unseen paper arrives by arXiv ID. No prior knowledge of its numbers is allowed into the audit.
observed
Six testable claims, ordered by importance. Each success criterion is derived only from the claim's own statement — no thresholds added after the fact.
reasoning
Write an audit script per claim, run it, iterate. Where reproduction can't reach the paper's number, record the gap rather than adjusting the bar.
verdict
Mark each claim supported, falsified, or inconclusive — receipts attached.
judge before human
A judge grades the assembled logbook's trustworthiness before anyone reads it. Every LLM call and tool result sits in the append-only trace.

Lemma

An auditor of scientific claims that refuses to trust itself.

Claims, not vibes. Up to six testable claims are extracted per paper, ordered by importance, each with a success criterion drawn only from the claim's own statement.
Honest audits. Numerical audits are written and run per claim. Where reproduction can't reach the paper's number, the gap is recorded — not the threshold.
Judge before human. A judge grades the assembled logbook's trustworthiness first; every call and result sits in an append-only trace.

Details

Open Lemma →

4 stages

Pipeline

≤ 6

Claims / paper

Append-only

Trace

NOTES

Measure. Attend. Return.

What is noticed with care can be met with care. Most systems fail by rushing past the signal.

Contemplative practice and clinical observation agree: the body often speaks before the crisis. Listening is empirical work.

Adherence research favors small, repeatable acts over walls of advice. Complexity is usually the enemy of return.

Outcomes compound in ordinary days. Drama is optional; showing up again is not.

A question, if you have one.

Ask Mira