Recovery-Aware Planning: When Your Calendar Knows Your Sleep
I have worn a Whoop continuously for four years. I have a calendar that I optimize obsessively. For most of those four years, neither system knew the other existed.
I have worn a Whoop continuously for four years. I have a calendar that I optimize obsessively. For most of those four years, neither system knew the other existed.
Tuesday morning my Whoop would say recovery is 31% — the third bad night in a row. My calendar would say "9:30am: deep work block — finalize Q2 forecasting model." The deep work block did not care that I was running on cognitive fumes. It got pushed to 11am, where it competed with a meeting, and eventually slipped to Friday. By Friday I would be guilty about the model, and the recovery score had not improved because the slipped work was now dread, which is its own kind of strain.
The fix is obvious in retrospect: the calendar should know the recovery score. The deep work block should land where the body has the bandwidth, not where the calendar has a hole. This is what we mean by recovery-aware planning, and it is what Elyra's morning brief does.
The Two Systems That Don't Talk
Most productivity advice assumes a body that is uniform across the week. "Schedule deep work in the morning." "Block your calendar for two-hour focus sessions." This advice is fine in the average. It is wrong on Tuesday, after a 11pm finish on Monday, when your HRV is in the basement.
Health-tracking advice has the opposite problem. Whoop will tell you your recovery is 31% and recommend an easy day. It does not know that you have a board prep at 2pm that cannot move. The recommendation is correct in isolation and useless in context.
The two systems are each half of the answer. The calendar without recovery is a body-blind dictator. Recovery without the calendar is a context-blind suggestion. Neither alone changes how you actually plan a Tuesday.
What Elyra Does Differently
Elyra is a local-first personal agent that ingests both sides — your calendar (Google, iCloud, or Outlook) and your Whoop data — and writes a morning brief at 7am that takes both into account.
The brief is not generic. On Tuesday after a bad night it might say:
Recovery is 31%, sleep was 5h 47m, your HRV is 12% below baseline. Today's calendar has the QBR at 2pm — that is a fixed-cost cognitive block. The deep work you flagged for "forecasting model" should not happen today. Two options: defer it to Thursday morning when recovery should rebound, or do thirty minutes of structural outlining now while caffeine is fresh and queue the substantive work for Thursday.
Then it actually moves the deep work block on the calendar.
This is not magic. It is the agent reading two streams of context and writing a plan that respects both. The reason no off-the-shelf calendar app does this is that no off-the-shelf calendar app has access to your Whoop data plus your goals plus the running history of which slipped work needs a slot.
What "Recovery-Aware" Doesn't Mean
I want to head off two reasonable objections.
This is not "do less when you are tired." That is the easy version, and it is wrong. Sometimes the right answer on a low-recovery day is to do more low-cognitive work — clearing inbox, structural editing, calls that do not require novel thinking. The agent's job is not to reduce volume on bad days. It is to change the composition of the day to match the cognitive budget.
This is not pseudoscience about HRV. The signal is noisy. The body is not a deterministic system. Elyra does not pretend to predict your day from a single biometric. What it does is detect outlier days — the worst-recovery 15% — and adjust the schedule for those specifically. On normal-recovery days the brief is mostly unchanged from a calendar-only plan.
The high-leverage move is not optimizing every day. It is not wasting deep work on a body that cannot do deep work today.
The Trend That Matters More Than The Day
Daily recovery is interesting. Weekly drift is more interesting.
Elyra runs a weekly drift review every Sunday — looking at the last 90 days of recovery, sleep duration, and strain, and comparing them to the prior 90. The brief flags trends:
Sleep average dropped from 7h 12m to 6h 48m over the last six weeks. HRV is trending down. The calendar shows three consecutive weeks of double-booked Wednesdays. Suggest: protect Wednesdays from 2–6pm starting next week, and move strength training from Thursday morning to Saturday morning to give a recovery buffer.
This is the part of personal AI that has changed how I actually work. The daily brief is helpful. The weekly drift review is the difference between catching a problem and being three months into a problem.
The Privacy Constraint That Made This Possible
I want to be specific about why this works as a local-first product and would not work as a cloud product.
Whoop data is sensitive. So is your calendar. So is the agent's running model of when you sleep, what you eat, and which weeks went badly. None of this is data I would casually pipe to a cloud productivity SaaS — even one whose privacy policy I trusted on day one.
Elyra runs locally. The Whoop sync writes to local SQLite. The calendar sync writes to local SQLite. The morning brief is written by Claude, with the focused prompt for that single morning, and the response lands in a local markdown file. There is no cloud index of my sleep history. There is no embedding of my schedule on a vector database I do not control.
The product would not exist if the privacy model were anything else. I would not feed four years of biometric data into a cloud agent. Most people who think carefully about this would not either.
What To Try If You Want To Test This Yourself
If you have a Whoop or Oura ring and a calendar you actually keep, the experiment is straightforward:
For two weeks, write a one-line note each morning: today's recovery score, today's planned deep work block, and what actually happened to it. At the end of two weeks, look at the days where the deep work moved or did not happen.
There will be a pattern. The deep work that slipped will correlate strongly with low recovery in the prior 24 hours.
Once you see the pattern, the question is whether you want to do this manually every morning — or have an agent that does it before 7am.