Finding an accountability partner is a pool-shape question, not a checklist

Finding, as a present-tense participle, describes a state the pool is in, not an action the reader performs. On 2026-04-23 the pool held 52 active rows in status pending or ready, drawn from 47 distinct cities across 19 declared timezones. Whether a particular signup is currently being found is a function of whether another row exists in that pool with a circular UTC-minute distance at or below 60 from one of your session slots. The reader does not inspect candidates. The reader does not interview anyone. The reader is a row in a set, and 'finding' names whatever the set is doing to itself at the next cron tick.

Twenty-three, computed by scripts/best-pending-pairs.mjs on 2026-04-23. That script loads every waitlist_entries row in status pending or ready, enumerates every unordered pair, drops any pair further than 60 UTC-minutes apart, drops any pair that has been matched before, scores the remainder on readiness plus old-student overlap plus identical session duration plus minute-proximity, and greedily picks non-overlapping pairs from the sorted list. Twenty-three is how many the greedy picker assembled without reusing a single row. The top-scoring pair sits twenty UTC-minutes apart at identical one-hour durations; the lowest-scoring of the 23 sits fifteen UTC-minutes apart at 30-minute durations.

Almost uniformly a graduate of a 10-day residential course. Forty-nine of the 52 active rows answered 'Yes' to is_old_student, which on this site means a person who completed the minimum 10-day course in the S.N. Goenka tradition. That 49 of 52 figure is the most important demographic fact about the pool, because the sort keys reward both-old pairings above both-pending pairings. Three of the 52 rows are not old students, and they are still eligible, but they land further down the sort when a both-old pair is available. In practice, the pool selects for people who have already sat once.

47 distinct city strings and 19 distinct timezone strings for 52 rows. The coverage in April 2026 spans North America (Chicago, Minneapolis, Missoula, Berkeley, New York, Montreal, Toronto, Houston, Phoenix, Scottsdale, Livermore, San Diego, Columbus), Europe (London, Prague, Copenhagen, Amsterdam, Lisbon, Berlin, Zurich, Warsaw, Bologna, Pau, Utrecht, Warrington), Asia (Bangalore, Bengaluru, New Delhi, Ahmedabad, Dehradun, Vadodara, Mumbai), Africa (Dakar), and South America (São Paulo). The geographic breadth is a function of the product being advertised entirely on SERPs; the pool mirrors whoever the search traffic is.

The filter at src/app/api/auto-match/route.ts line 164 drops a candidate pair if timeDiff(slotA.utcMinutes, slotB.utcMinutes) exceeds 60, where timeDiff is defined two lines earlier as Math.min(|a-b|, 1440-|a-b|), the circular minute distance on a 24-hour clock. Sixty is the operator's calibration for what 'daily partnership' still means. Thirty minutes would starve the pool and produce zero pairs most ticks. Two hours would pair 06:00 sitters with 08:00 sitters and call it a daily partnership, which is not what a daily partnership is. Sixty is the distance at which two adults in different timezones can still reasonably call each other at the same moment.

Continuously from signups, every 30 minutes from the cron. The schedule '*/30 * * * *' in vercel.json calls GET /api/auto-match, which loads the pool, enumerates pairs, and writes matches where viable. Each tick can both consume rows (by matching them) and change the eligibility of survivors (by advancing the clock, which shifts 24-hour cool-off gates). A signup submitted at 09:00 is eligible at 09:30 only if more than 24 hours have passed since the row was created, so in practice the first chance for a fresh row to be matched is tomorrow's :00 or :30.

Because the WaitlistEntry schema in src/lib/db.ts does not carry any field the reader could act on. It holds timezone, frequency, session_duration, morning_time, evening_time, is_old_student, status, contact_count, and pass_count. None of that is a quality, an affinity, a lifestyle, or a goal. The matcher cannot see what the general-advice pages spend their words on. The only things the matcher sees are the shapes of the rows in the pool. So the verb 'finding' has to describe a shape-matching operation that the system does to itself, not a quality-evaluating operation the reader does to candidates.

You wait. The row stays in status pending or ready. The next tick is at most 30 minutes away. On any given tick the pool adds new rows (fresh signups aging past the 24-hour cool-off), removes rows (matched into pairs and promoted to matched status), and mutates rows (contact_count increments, retry cool-offs expire). The shape changes around you. The pool at :30 is not the pool at :00, and a signup that was unfindable at :00 may have a partner at :30. The reader's job is to leave the shape alone; every attempt to optimize the shape from the outside invalidates the symmetry the matcher depends on.

src/app/api/auto-match/route.ts calls getPriorMatchedIds on one side of every candidate pair and drops the pair if the pair has ever appeared in the matches table in any status other than cold-bilateral-decline. The forward-only guarantee is written once and then enforced on every tick. Its effect on the pool is that the viable subset shrinks monotonically over time for any given member; if you sit in the pool for long enough, you exhaust the partners your particular shape is compatible with, and the only way back into viability is for fresh rows to enter with shapes close to yours.

No. Nothing on this page is operational instruction in the Goenka tradition. The 10-day courses at dhamma.org are where technique is transmitted by authorized assistant teachers, and this site only handles the logistics of peer pairing for daily sits after a reader has completed such a course. Any question about how to sit, how to work with a difficulty, how to interpret a bodily state, how to structure a practice, or what to notice belongs in the course and with an authorized assistant teacher, not here.