Runs

Each time A858's program started, it created one System.Random seeded with the machine's uptime and drew one GUID count per scheduled post. A run is one such app start. Fixed-N batches, tests and other groups are listed too, so every filler post belongs somewhere. How runs are identified →

• seed
• probable
• fixed
• open
• short
• test
• single
• setup
• pre-automation
• companion
• implied boot

Runs by type

When the runs were started

A run's seed is the machine's uptime when the program started, so the seed and the boot time give the start time. For the 77 seeded runs with a known start time, the start falls in a window of minutes, at most half a day. For the other 53 runs, it is estimated from the posting schedule: within one posting interval before the first post.

Theory: the operator starts each run by hand, on a weekday, in working hours on the US Pacific clock, and the program runs until it is stopped or the machine reboots. The starts fit that pattern.

• Working hours. 54 of the 77 seeded starts fall between 08:00 and 18:00 Pacific. A uniform clock would give 32 (one-sided binomial p = 4e-07). Counting the estimated starts too, at the middle of their windows: 95 of 130, against 54 (p = 4e-13).
• Weekdays. 7 of the 77 seeded starts fall on a Saturday or Sunday, against 22 if the day of the week did not matter (one-sided binomial p = 3e-05). Counting the estimated starts too: 9 of 130, against 37 (p = 7e-10).
• Not automatic. Starts come at arbitrary uptimes, usually days after the machine booted and days after the previous run stopped. Restarts without a reboot often change the count range or the interval, so the settings look chosen each time.
• Caveats. The clock is taken to be US Pacific, with daylight saving applied. Estimated starts are weaker evidence than seeded ones: they assume the first post follows the start within one interval. The pattern is a theory that fits the data, not proof of who started the runs.

All runs

Interval: the median gap between a run's posts (for groups, the first 11 posts, since one group can hold several app starts). Single posts have no interval.

seed

Seed run: the count of every post is reproduced by one long-lived System.Random seeded with the machine's uptime.

probable

Probable seed run: the seed rests on a narrow launch window, not on many matching posts, so a chance fit cannot be ruled out. Each one has its own page with the search.

fixed

Fixed-N batch: one app start that posted the same count every time, so there is no seed to find.

open

A run of the program whose counts no tested generator reproduces. Its counts are plotted by draw, numbered from the first post (an assumption).

short

An app start too short to pin down a seed. Some have candidate seeds, listed on their pages; their counts are plotted by draw with the draws numbered from the first post (an assumption).

test

Operator test posts.

single

A single post on its own (usually a test start).

setup

Account-setup posts.

pre-automation

Posts from before the scheduled app.

companion

Posts by the companion account, 9CB9D65F54ED858A, on its own subreddit, r/9CB9D65F54ED858A. A seed reproduces its counts, as in a seed run, but it is not part of the A858 counts; each post has a page of its own.