time series analysis
analysis of 4,656 threads spanning 2025-05-12 to 2026-01-08 (~8 months)
busiest hours
peak activity 10am-8pm, with hour 19 (7pm) as absolute peak at 322 threads.
| hour | threads | avg turns |
|---|---|---|
| 19 | 322 | 24.0 |
| 16 | 293 | 57.5 |
| 17 | 280 | 31.8 |
| 10 | 272 | 37.3 |
| 18 | 262 | 35.8 |
| 15 | 261 | 51.8 |
lowest activity: 2-6am (60-94 threads/hour)
busiest days
| date | dow | threads |
|---|---|---|
| 2026-01-08 | wednesday | 303 |
| 2026-01-07 | tuesday | 296 |
| 2025-12-19 | friday | 252 |
| 2026-01-06 | monday | 179 |
| 2025-12-10 | wednesday | 110 |
last week of data shows massive spike—likely represents team scaling or usage surge.
day of week patterns
| day | threads | % of total |
|---|---|---|
| wednesday | 965 | 20.7% |
| thursday | 915 | 19.7% |
| friday | 738 | 15.9% |
| monday | 644 | 13.8% |
| tuesday | 721 | 15.5% |
| saturday | 389 | 8.4% |
| sunday | 284 | 6.1% |
midweek peak: wed/thu account for 40% of all threads.
weekend vs weekday
| period | threads | resolved | resolution % | avg turns |
|---|---|---|---|---|
| weekday | 3983 | 1741 | 43.7% | ~40 |
| weekend | 673 | 329 | 48.9% | ~50 |
weekend threads have HIGHER resolution rate (+5.2pp) despite lower volume—possibly:
- more focused work (less interruption)
- self-selected important tasks
- fewer exploratory/speculative threads
time-of-day vs outcome
| time block | threads | resolved | resolution % | avg turns |
|---|---|---|---|---|
| late_night (2-5) | 313 | 189 | 60.4% | 38.1 |
| morning (6-9) | 609 | 363 | 59.6% | 43.6 |
| midday (10-13) | 1027 | 493 | 48.0% | 48.0 |
| night (22-1) | 567 | 267 | 47.1% | 58.5 |
| afternoon (14-17) | 1080 | 467 | 43.2% | 48.9 |
| evening (18-21) | 1060 | 291 | 27.5% | 33.0 |
key finding: evening threads (6-9pm) have dramatically lower resolution rates (27.5%) despite being the busiest period. late night and early morning show BEST outcomes.
possible explanations:
- evening fatigue → lower quality prompts or follow-through
- evening = more exploratory “what if” threads
- morning = fresh focus, clear intent
- late night = dedicated deep work sessions
monthly trend
| month | threads | resolution % |
|---|---|---|
| 2025-05 | 24 | 75.0% |
| 2025-06 | 297 | 35.4% |
| 2025-07 | 344 | 54.4% |
| 2025-08 | 288 | 52.4% |
| 2025-09 | 254 | 59.1% |
| 2025-10 | 296 | 52.7% |
| 2025-11 | 496 | 51.0% |
| 2025-12 | 1620 | 46.0% |
| 2026-01 | 1037 | 29.3% |
volume surge: dec-jan shows 2.7k threads (58% of total dataset) but resolution rates dropping—likely recency bias (recent threads haven’t resolved yet) or scaling effects (more users = more varied quality).
weekly cadence
notable spikes:
- W48-W01 (late nov through early jan): 2,657 threads—major usage acceleration
- W01 (2026): 859 threads in single week
suggests either:
- team adoption wave
- project deadline crunch
- seasonal work pattern
summary
- best productivity windows: early morning (6-9am) and late night (2-5am) have highest resolution rates (~60%)
- avoid evening sessions: 6-9pm shows only 27.5% resolution—worst time block
- midweek dominance: wed/thu are workhorses (40% of threads)
- weekend quality premium: fewer threads but better outcomes
- recent surge: last 6 weeks represent majority of usage, with jan 2026 showing massive scale-up