SWE-chat: Users & Trajectories

Analysis of how coding trajectories (sessions) link to users in the SWE-chat dataset. A trajectory is one coding session (sessions.parquet, one transcript file); the user is the commit author resolved via the GitHub API (user_id).

Linkage summary

5,851
total trajectories (sessions)
189
unique users
3,908
trajectories linked to a user (66.8%)
1,943
trajectories with no user_id (33.2%)

Linkage is via sessions.user_id (also denormalized onto every row of conversations.parquet: 2,692,480 turns, 187 unique users there — consistent with the sessions table). Unlinked sessions are those whose commit author could not be resolved to a GitHub identity.

Trajectories per user

20.7
mean
6
median
2 – 16
interquartile range
450
max (single user)

Distribution is heavily right-skewed: the median user has 6 trajectories, but the top 10 users account for 49.6% of all linked trajectories, and just 11 users (of 189) contribute half of them. 90th percentile: 41.2; 99th percentile: 283.0.

Distribution of trajectories per user

BucketUsers
1 trajectories40
2–5 trajectories50
6–10 trajectories28
11–25 trajectories37
26–50 trajectories19
51–100 trajectories10
101+ trajectories5

Top 25 users by trajectory count

user_idTrajectoriesRepos Total turnsOutput tokens First sessionLast sessionAgents
marcus-sa45026,0948,812,5902026-02-272026-04-08Claude Code
dayhaysoos41513,6543,789,9232026-03-062026-03-27OpenCode
cyyeh26532,3422,530,6862026-02-192026-03-06Claude Code
KeKs0r20611,7491,525,3262026-02-162026-03-13Claude Code
Soph16712,5271,343,2882026-01-052026-03-19Agent, Claude Code, Gemini CLI, Vogon Agent, unknown
zchee9631,0083,359,2562026-03-302026-04-18Claude Code, Codex
khaong9413,7071,342,7822026-01-192026-03-06Agent, Claude Code
cteyton931459193,5762026-02-132026-03-12Claude Code
gtrrz-victor7711,194237,4032026-01-132026-03-03Agent, Claude Code, unknown
jeevanpillay7738071,235,7262026-02-152026-04-08Claude Code
Nagi-ovo7618061,402,4602026-02-162026-03-29Claude Code
moebiusband73682227499,6152026-03-032026-03-30Claude Code
jdsingh1229185911,3051,072,0122026-02-272026-03-13Claude Code
schmalle551404810,6582026-02-222026-03-19Claude Code
armelhbobdad5111,8641,699,6672026-02-252026-04-17Claude Code
alishakawaguchi481282140,0102026-02-202026-03-21Claude Code
gregszero451325215,2072026-02-162026-02-23Claude Code
pc0358604427311,237,5782026-03-072026-04-09Claude Code
robouden422905304,3072026-02-112026-03-29Claude Code
135yshr4121,087739,9162026-03-022026-03-13Claude Code
tslateman412927570,8412026-02-202026-03-22Claude Code, Gemini CLI
tarasyarema4011,0881,145,8022026-03-052026-03-21Claude Code, OpenCode
Sagit-chu401229314,5982026-03-032026-03-14OpenCode
oddessentials3312,2903,830,8062026-03-262026-04-07Claude Code
hirakiuc3215271,341,2262026-03-012026-03-14Gemini CLI

Trajectories & users by agent

AgentTrajectoriesUsers
Claude Code3,192172
OpenCode50314
Codex1004
Gemini CLI448
unknown243
Agent224
Cursor143
Roger Roger Agent41
Copilot CLI22
Vogon Agent11
claude-code11
opencode11