I was running a block at half-marathon race pace. The coach kept nagging me about being over or under the target, without helping me settle into the pace.
That made me think that sometimes you just need to get to run.
A correction needs to help.
That run gave me a clearer standard for the coaching: an instruction should help me do the session. Calling out another deviation is only useful if it helps me respond.
The changes below address the prescription and the way Klivet interprets a run. They are part of that work. Whether the live coach has learned when to leave me alone still needs its own verdict on a run.
When the easier part stopped being easier
One defect was in an alternating tempo session. The steady sections between harder efforts were eligible to become goal-race-pace work. The part intended to provide relief could therefore become another demanding section.
That path has been corrected. Steady float sections keep their role, and a race goal can no longer pull training paces arbitrarily far beyond the app’s estimate of current fitness.
A target should help you train towards it.
This does not make the estimate infallible. It gives the prescription a boundary, and keeps the recovery within a session from being silently rewritten by the goal.
The name of the run matters
Another change addressed runs being labelled Tempo from heart-rate evidence without supporting pace evidence. The classifier now checks pace before keeping certain quality-session labels, when suitable pace data is available.
That check has exceptions, including hill repetitions and trail runs. It is a specific correction to the classifier, rather than a claim that pace can explain every run.
The weekly plan had a simpler problem too: its long-run distance had been calculated but was not being used to allocate the week. The long-run slot now receives that allocation, with the remaining distance spread across the other runs.
What you say reaches the next decision
You can now record why a run felt off from the watch as well as the phone. Recent reports of feeling ill or having a niggle can feed into the next readiness decision when today’s check-in has no answer.
That report is one input. It does not diagnose an injury or automatically stop training. A fresh check-in takes precedence over yesterday’s note.
What changed
- Recovery sections retain their purposeSteady floats are excluded from goal-pace substitution.
- Goals meet current fitnessTraining pace calculations limit how far an ambitious target can pull the prescription.
- Run labels ask for corroborationApplicable quality labels need pace evidence alongside heart rate.
- The long run gets its allocationThe plan uses the long-run distance it already calculated.
- Your feedback carries forwardA recent illness or niggle report can inform the next decision.
The next test is another race-pace block. I want a target I can trust, useful help when I need it, and room to settle into the running. That is the experience these changes have to earn.
Klivet’s readiness and fitness estimates remain estimates. They cannot diagnose illness or injury.