context window analysis
estimated tokens derived from char_len / 4 — rough approximation.
token distribution
| bucket | threads | % of total |
|---|---|---|
| empty (no messages) | 376 | 8% |
| <10k | 2,961 | 64% |
| 10-25k | 978 | 21% |
| 25-50k | 291 | 6% |
| 50-100k | 42 | 0.9% |
| >100k | 8 | 0.2% |
most threads stay well under context limits. only ~1% push past 50k tokens.
outcome rates by token bucket
| bucket | n | resolved% | committed% | frustrated% | handoff% | unknown% |
|---|---|---|---|---|---|---|
| <10k | 2,961 | 38.6 | 7.3 | 0.1 | 13.8 | 35.9 |
| 10-25k | 978 | 69.1 | 7.2 | 0.6 | 13.5 | 9.5 |
| 25-50k | 291 | 73.2 | 5.8 | 1.4 | 10.3 | 8.6 |
| 50-100k | 42 | 81.0 | 4.8 | 0.0 | 4.8 | 9.5 |
| >100k | 8 | 62.5 | 12.5 | 12.5 | 0.0 | 0.0 |
observations
-
resolution rate INCREASES with thread length up to 100k — longer threads correlate with deeper, successful work (81% at 50-100k vs 38.6% at <10k)
-
frustration spikes at >100k — 12.5% frustrated (1 of 8 threads) vs near-zero elsewhere. context pressure starts hurting.
-
short threads have high UNKNOWN rates — 35.9% at <10k suggests quick lookups or abandoned exploratory threads
-
handoffs decrease at scale — longer threads tend to complete in-place rather than spawning
threads likely hitting context limits
8 threads estimated at >100k tokens:
| thread | title | turns | status | est_tokens |
|---|---|---|---|---|
| T-0ef9…afaa | Minecraft resource pack CIT converter | 1623 | PENDING | 272k |
| T-048b…665e | Debugging migration script for book pack | 988 | RESOLVED | 172k |
| T-019b…33c1 | Untitled | 1 | FRUSTRATED | 146k |
| T-6113…1381 | Investigate trace link issue | 170 | RESOLVED | 128k |
| T-b428…773d | Create implementation for project plan | 594 | RESOLVED | 126k |
| T-2e58…f98 | Map rc-menu dependencies | 330 | RESOLVED | 122k |
| T-939a…1534 | Enhance search_modal aggregation | 455 | COMMITTED | 110k |
| T-c66d…68a | Review S3 background ingest | 615 | RESOLVED | 105k |
the FRUSTRATED >100k thread
T-019b88a4-5dc7-7079-a2c7-a68d5d8a33c1 — single turn, 146k tokens. user pasted entire CI job output into one message. not a context window exhaustion from conversation — input was already overwhelming.
steering patterns by token bucket
| bucket | steering per 10k tokens | total steering |
|---|---|---|
| <10k | 0.33 | 0.1 |
| 10-25k | 0.42 | 0.7 |
| 25-50k | 0.39 | 1.2 |
| 50-100k | 0.35 | 2.2 |
| >100k | 0.30 | 3.9 |
steering rate per 10k tokens stays roughly constant (~0.3-0.4). longer threads accumulate more steering but not disproportionately — users don’t steer MORE when context is long.
FRUSTRATED threads by token count
14 total FRUSTRATED threads:
- 1 at 146k (CI log dump — immediate frustration)
- 1 at 43k (Effect race conditions)
- 1 at 31k (scoped context isolation)
- 11 at <30k tokens
most frustration happens UNDER context limits. frustration correlates more with problem difficulty than context exhaustion.
conclusions
- context limits rarely hit in practice — <1% of threads exceed 50k tokens
- when limits ARE hit, resolution still common — 6/8 threads >100k resolved or committed
- the single >100k frustrated thread was user error — pasting 146k tokens of logs in one message
- frustration is problem-bound, not context-bound — difficult debugging tasks at normal token counts
- longer threads = deeper engagement = better outcomes — selection effect: hard problems that need more turns get more effort