Questions, answered.

Kyniskis is a perseverance trainer. It reads your body from the devices you already own and gives you one decision every day: push, hold, or recover. It is not built to help you train harder — it is built to help you never stop.

No. Bring what you already have. Kyniskis works with Apple Watch, WHOOP, Garmin and Oura — the device on your wrist is the best device for the job. There is no Kyniskis band to order and nothing to add to your other wrist.

Live today: Apple Health (including your full Apple Watch history), WHOOP, and Garmin Connect. In progress: Oura and Huawei Health. Anything that writes into Apple Health or Health Connect can also be read — even where we don't have a direct integration yet.

Day one gives you a daily plan. The training-load ratio becomes meaningful once it has roughly six weeks of history behind it, because that is what its chronic baseline measures. The early weeks run on a shorter window and sharpen as data accumulates.

You authorise read-only access and we pull it on a schedule. We never write to your health records, and we never request access we don't use.

Large histories are backfilled progressively — a decade of Apple Health does not arrive in a single request. Overnight recovery from a wearable appears after that night's sleep has been scored, which happens once you wake up.

We don't average them. Different sources are not measuring the same thing: an overnight HRV reading from one device may be a normalised RMSSD, while all-day HRV on another is scattered SDNN samples — the same night can differ sharply between them. For each signal, Kyniskis uses the source that is semantically correct for it, rather than blending numbers that mean different things.

It stays yours. You can delete your account and everything in it at any time, yourself, without contacting support.

Training Load Ratio. A single session's load is average effort to the power of 1.5, multiplied by duration in minutes. TLR is the 7-day exponentially weighted average of that daily load, divided by the 42-day average. It answers “am I overreaching?” — not “was today hard?”

It is the well-tested acute-to-chronic sweet spot. Below 0.7 you are losing fitness rather than resting it. Between 1.3 and 1.5 you are pushing into warning territory. Above 1.5, sustained, is the classic signature of overreaching.

Because TLR is a rolling ratio, not an on/off switch. One rest day only mixes about 25% of a zero into the 7-day window, while the 42-day denominator moves roughly 4.65% per day. So a single rest day typically pulls TLR down by about 20%. It takes around three consecutive rest days to fall back below 0.7. A heavy session leaves a memory about seven days long.

From overnight physiology — heart-rate variability, resting heart rate and sleep — compared against your own baseline, not a population average. That is why the same score can mean something different for you than for someone else.

Because load and readiness are two different questions. What you did cannot tell you what it cost you. Kyniskis keeps them separate — TLR decides how much, Recovery decides whether today is the day — and merges them only at the point where a decision has to be made.

Two steps. First every day gets a load number: on a WHOOP source it is that day's strain, on an Apple source it is active calories - or, when active calories are missing, steps multiplied by your own median calories-per-step. Second, TLR is the 7-day exponentially weighted average of that daily load divided by the 42-day exponentially weighted average. TLR = EWMA(7d, k=0.25) / EWMA(42d, k=0.0465) Thresholds are source-specific, because strain and active calories are not the same unit: 0.70-1.30 is optimal on a WHOOP source, 0.80-1.15 on an Apple source.

Because it lets recent days dominate without discarding the older baseline. On the 7-day side the constant is 0.25, so yesterday carries roughly a quarter of the acute number; on the 42-day side it is 0.0465, so each new day moves the chronic baseline by only 4.65% - which is exactly what makes it stable. A plain average would treat a session from 41 days ago the same as yesterday's, and would answer a different question. It also explains why one rest day pulls TLR down only around 20% rather than erasing it: a single zero enters the fast side, while the slow side barely moves.

Straight from your source when it provides one. When it does not - typically an Apple-only source - we estimate it from your own resting heart rate and HRV: score = 64 - 9.3 × z(RHR) + 2.7 × z(HRV) clamped to 5-99. Here z means how far today sits from your own 21-day average, measured in standard deviations, with a floor on the spread (6 bpm for resting heart rate, 12 ms for HRV) so that a very steady week cannot manufacture extreme scores. At least 3 days of resting heart rate are required before any estimate appears at all. Lower resting heart rate and higher HRV both push the score up; that is why the RHR coefficient is negative.

Because it is one, and we would rather label it than dress it up. Checked against same-day values from a source that reports recovery directly: correlation about 0.50, mean absolute error 15 points, and the green/amber/red band agrees roughly 70% of the time. The practical consequence is deliberate asymmetry. The estimate will happily say green, but only on two conditions at once - 75 points or more and a resting heart rate clearly below your baseline. It rarely calls red, so on a genuinely bad day it leans amber and hands you the floor workout rather than declaring you broken. The label on the card exists for the same reason.

We carry forward, with a limit and a label. If today has no resting heart rate, the estimate looks back up to 3 days for the most recent one, and the card states which date it came from. Beyond 3 days we show two dashes rather than guess, because a stale baseline presented as today's readiness is worse than an honest gap. This is also why a score can appear on a morning you have not synced yet - and why we would rather show you nothing than a number we cannot stand behind.

Weighted, and re-normalised whenever a dimension is missing. Composite = 0.5 × Recovery + 0.3 × Emotion + 0.2 × TLR If a dimension has no data that day its weight is dropped and the remaining weights are scaled back up, so a missing input lowers confidence instead of silently counting as zero. Recovery carries the most weight because it is the measurement; how you feel comes next because it is the input no wearable can capture; TLR is there to say whether the load itself is sane. If nothing is available we show two dashes rather than a number.

Two steps. First, readiness: if any dimension reads low the day is red; green requires two dimensions high at the same time; everything else is amber. Second, a decision matrix crosses readiness with your TLR band. Low TLR plus green becomes add load. Optimal TLR plus amber becomes hold. Any red becomes rest-first regardless of what the load says. Volume never overrules readiness - that ordering is the whole point of the product.

Because a binary answer is what breaks streaks. Layer 1 is 5 minutes of breathing or meditation plus 15 minutes of mobility, and it is available in every state including red. Layer 2 adds aerobic work, from amber upward. Layer 3, intensity, only on green. On a red day the plan is deliberately small and deliberately not zero, so the day still counts as done. A plan you cannot fail at is the only plan you can keep for years.

Yes, and that is intentional. Your raw measurements never change, but scores are recomputed with the current version of the algorithm, so a historical day can show a different number after an update than it did before. We prefer that to freezing an old formula for the sake of looking consistent. What the algorithm will never do is diagnose anything: it describes load and readiness, it does not detect, treat, or rule out a medical condition.

Because “I think I can handle it” is not the same as “I can actually handle it.” One tap, three states: 🔥 ready to push, 😐 neutral, 😞 don't want to train. We ask about the moment before training — not about your whole day, which nobody can summarise honestly.

Yes. 🔥 with good recovery opens the green light. 😞 with good recovery gets: “Your body is fine. Start with five minutes — and stop if you still don't want to.” The gap between how you feel and what your body says is the most useful thing we record, because it is the only thing that teaches you which of the two to trust.

Every day has a floor and a ceiling. Layer 1 — always, in any state: 5 minutes of meditation and 15 minutes of mobility. Layer 2 — from a yellow day upward: add cardio. Layer 3 — green days only: high intensity. Hold the floor, take the ceiling when you have earned it.

Because zero is the only failure that actually matters. A red day is not “do nothing” — it is “do the floor.” Fifteen minutes of mobility keeps the chain alive, and the chain is what turns training from a decision you remake every morning into an identity you keep.

Nothing breaks, and we won't show you a shattered streak. The next thing you hear is: “You skipped yesterday. That's fine. Just come back today.” Never miss twice.

Because most people don't fail by training too little within a sound programme — they fail by stopping. Personal records belong to whoever is still training in five years. That is the person this product is built for.

Trust your body first, then check the trend. Early illness, dehydration, alcohol, travel and hormonal shifts move heart rate and HRV hours — sometimes days — before you notice anything. If a single reading looks wrong, wait and look at several hours rather than acting on one sample.

That is a genuinely useful disagreement, and we show it to you rather than quietly resolving it in either direction. Over time you learn which of the two to trust in which situation — and that pattern is personal. It cannot be averaged out of a population.

No. Kyniskis supports general wellness and training decisions. It does not diagnose, does not treat, and does not replace a qualified healthcare professional. If something feels genuinely wrong, see a doctor.

No. Not to anyone, for any purpose. Kyniskis earns money from subscriptions, not from your body.

You. If you invite a coach or a training partner, they see only what the invitation grants, and you can revoke it at any time. Team and enterprise views are aggregate-only — individual data does not leave the server.

No. Your health data and your conversations are not used to train models.

Free covers the daily plan and your core scores. Pro adds full history, trend analysis and the AI coach. Coach is built for people who train others. Current pricing and entitlements are listed on the Subscription Plans page.

Yes. A coach can be invited to see an athlete's readiness and load, with red-day alerting, at exactly the scope you grant — and no wider.

Self-serve, at any time. No support ticket, no retention call. Subscription Plans → Manage.

Email jeff.qiao@konnectlabai.com. A human reads every message.