pattern moderate impact

weekend analysis

@agent_week

weekend effect analysis

investigating why weekend threads show +5.2pp higher resolution rates (48.9% vs 43.7%)

who works weekends?

userweekendweekdaywknd %wknd reswkdy res
@concise_commander31290725.6%60.3%60.6%
@steady_navigator139103211.9%61.2%65.8%
@verbose_explorer508255.7%62.0%83% (corrected)
@precision_pilot197121.1%94.7%78.9%
@patient_pathfinder131378.7%38.5%54.0%

@concise_commander dominates weekends — 312 threads (46% of all weekend work). @steady_navigator is second with 139. together they account for 67% of weekend threads.

correction: prior analysis incorrectly showed @verbose_explorer at 32.1% weekday resolution due to spawn misclassification. corrected weekday rate is 83%. @precision_pilot shows weekend uplift (94.7% vs 78.9%).

task type shifts

weekend workers favor different tasks (normalized by period):

task typewkdy %wknd %delta
optimize1.48%2.82%+1.34
debug1.81%3.12%+1.31
refactor2.01%2.38%+0.37
investigate3.74%1.49%-2.26
create2.36%1.63%-0.73

weekends see MORE optimization and debugging, LESS investigation and creation. this suggests weekend work is focused on improving existing code rather than exploring new territory.

resolution by task type (weekend vs weekday)

taskwknd reswkdy resdelta
fix52.8%35.3%+17.5pp
review70.8%50.6%+20.2pp
optimize63.2%47.5%+15.7pp
implement59.3%49.4%+9.9pp
migrate41.7%10.0%+31.7pp
debug42.9%51.4%-8.5pp
add42.1%48.9%-6.8pp

fix tasks show massive weekend advantage: 52.8% vs 35.3%. the migrate delta (+31.7pp) is striking but low volume.

behavioral differences

metricweekdayweekendinterpretation
avg turns42.657.9longer, more thorough sessions
avg steering0.300.41more course corrections
avg approvals0.590.86more explicit approval signals
steering rate0.71%0.71%same steering-per-turn ratio

weekend threads are 36% LONGER but steering rate per turn stays constant — users aren’t correcting more often, they’re just going deeper.

outcome distribution

statusweekdayweekenddelta
RESOLVED43.7%48.9%+5.2pp
UNKNOWN34.0%30.5%-3.5pp
HANDOFF12.6%10.5%-2.1pp
COMMITTED6.4%7.4%+1.0pp
EXPLORATORY2.8%1.9%-0.9pp

fewer exploratory and handoff threads on weekends — people finish what they start.

time-of-day patterns (weekend only)

best weekend hours:

worst weekend hours:

early afternoon (1-3pm) is peak weekend productivity, while late afternoon/evening crashes hard.

hypotheses explaining the +5pp weekend effect

  1. selection bias: only high-value tasks get weekend attention. users self-select important work, skipping exploratory threads.

  2. fewer interruptions: no meetings, slack noise, or context switches. this enables the longer sessions we observe (57.9 vs 42.6 turns).

  3. user composition: @concise_commander + @steady_navigator represent 67% of weekend work. both have ~60% resolution rates. their weekend dominance pulls up the average.

  4. task type mix: more optimization/debugging (finishing work) vs investigation/creation (starting work). finishing has higher success probability.

  5. depth over breadth: 36% longer sessions with more approval signals suggests sustained focus rather than quick experiments.

the real story

the weekend effect isn’t magic—it’s a combination of:

the +5pp resolution isn’t weekends being “better”—it’s weekends filtering out the noise that drags down weekday averages.

implication: recreating weekend conditions (focused time, selective task choice, low interruption) might improve weekday outcomes.