No live demo, no marketing screenshots dressed as a product — a faithful, anonymized sample of the actual instrument. The local dashboard, the insights you act on, meal logging that cites its sources, and a training week that flexes to your recovery every figure here is sample data. Yours is read on your own machine.
Coaching, not medical care — no longevity or anti-aging promise; the name is the long view, not a claim on your lifespan. Ilfirin runs on your own computer, reasoned by Claude (a frontier AI model) through the Claude plan you already pay for — so "no new subscription," not "free": you still need a Claude plan and an always-on machine. Privacy by architecture — no Ilfirin cloud, account, or database. Your records stay in files you own; during a chat, the readings Claude needs pass through Anthropic. Reads today: Fitbit · Pixel · Withings · Garmin (community) · Oura · Whoop · Apple Watch (via the free Health Auto Export app). Consumer CGM patches (Stelo, Lingo) aren't supported yet — a Dexcom/Libre bridge can push glucose in. Numbers below are a deliberately anonymized sample — not anyone's real records.
The real local dashboard, reproduced faithfully: one most-actionable headline, then all five domains as dense metric cards, each with its own trend. It surfaces the gaps — a flagged marker, a biological age it won't compute without the inputs, a strength session the wearable can't see — rather than a wall of green checks. Read-only; you act in chat.
Read-only. Every action happens in chat. Anonymized sample — no composite score, by design.
Ilfirin reads the trend, states it plainly, and names the single thing to do about it — each insight paired with the study behind it and an honest evidence tier. Sample signals, from the same anonymized user as the dashboard. It surfaces the signal and proposes the adjustment; it does not moralize.
Resting HR is +3 bpm above your 30-day baseline for 5 days running — an early sign of overreach, before it shows as fatigue.
SuggestedTake a deload week: cut Zone 2 volume about 30% and hold strength steady. Recheck once RHR settles back to baseline.
VO₂ max rose +1.2 this month — strong evidence of real cardiovascular adaptation. It's the single best-validated longevity signal, with no real ceiling.
SuggestedKeep the 4×4 interval cadence exactly where it is. Don't add volume — the current dose is working.
Your last panel put LDL cholesterol at 118 mg/dL — above the <100 target. LDL (and ApoB, which you haven't drawn yet) is the causal driver of heart disease, so a flagged panel is worth confirming.
SuggestedGet the fuller panel your screening list already flags — ApoB + Lp(a) — and review it with your doctor. Not a number to coach around on your own.
Your bedtime varied by 78 min last week — up from 52 min. Sleep regularity predicts mortality independent of how long you sleep.
SuggestedAnchor a consistent bedtime — pick a window and hold it, even on weekends. Duration's fine; it's the variance to tighten.
A photo is the recommended way to log a meal — identifying food is exactly what image models do well, and that's most of what the diet-quality score reads. What a photo can't do — size a portion, see the oil in the pan — Ilfirin asks in one question, and shows its estimate as a range until you answer. No false precision, no silently-logged guess.
Most of the diet-quality score is identity-driven — exactly the part a photo is good at. The few things it can't read, it asks about or marks "unknown." It never imputes a number it doesn't have.
Photos are great at what the food is. Bad at how much. So Ilfirin asks the one question that closes the gap — and holds the estimate as a range until you answer.
Prefer to type, or logging a meal you already ate? Say it the messy way you'd say it out loud — Ilfirin decomposes the sentence into real foods with real macros, flags any guess, and cites every match instead of writing fiction to your record.
AHEI-2010 · validated dietary index (Chiuve 2012, J Nutr)
Built from the meals you actually logged — a validated quality score, not a calorie scoreboard. Components shown, never collapsed into a single grade you can't interrogate.
Most apps log the model's guess as truth. Ilfirin asks first — and cites its source.
A worked example for a balanced, longevity-leaning week. The plan flexes to your goals and your recovery: when readiness slips, the hard days soften before you ever feel buried. This is the default template; yours adapts to your equipment, travel, and the capabilities you want to keep.
Thursday's swim converts to 4×4 intervals only when your recovery clears four checks — otherwise it stays easy. The point is to push hard when your body can absorb it, and not when it can't. One amber flag and it drops back to easy that week, automatically.
Today's session, laid out for the gym floor — sets, reps, load, and a form video for the main lift. Tap a dot as you finish a set. No logging app to wrestle; tell the chat "did Strength B, all sets" afterward and it's recorded.
The main lifts link the reference video — pull it up if a cue slips. No memorizing tempo notes between sets.
One dot per set, glanceable across the room. The progress count keeps the tally for you.
"Did Strength B, all sets at 25 lb" lands the whole session — reps, load, and RPE — without a single form.
Representative numbers from a sample longevity-block plan. Yours adapt to your equipment, readiness, and benchmarks. Coaching, not medical care.
Everything you just saw runs locally, reading your own data. Tell us a little about your devices and your setup — we'll provision your own instance and send a personalized installer with a guided setup — no command-line expertise needed. It's an early, invite-based beta, and access is limited while we keep it small.
Sample data throughout — during a chat the readings Claude needs pass through Anthropic; your records otherwise stay in files you own. No new subscription beyond the Claude plan you already pay for, plus an always-on machine. An early beta, and not pretending to be more.
