# Nine Commits, post 2 — every run behind the table in index.mdx.
#
# 42 runs of one question against gpt-5-mini at commit v2-hands:
#   "Which file in this project sets the model name the agent uses, and what is it set to?"
# The only variable is TOOL_DESCRIPTIONS. Everything else is held fixed.
#
# TRANSCRIBED BY HAND from each run's stderr line and the outcome the recorder
# derived for it. It is a weaker artifact than a machine-written log — the rows
# were read off a terminal by a person — and this file says so rather than
# letting you find out. From v2 the CLI appends its own row to
# agent/traces/runs.tsv, so the next post's tally is generated, not typed.
#
# Rows are in BOOKKEEPING ORDER, not run order. The two runs published as
# traces are last, at rows 41 and 42; they were chronologically first. The
# blocks above them are, in order: 5 precise, 5 thin, 15 precise, 15 thin.
#
# tools_called is what the model actually called, comma separated: one run
# (row 23) called both in a single step, which prepareStep does not prevent
# because it caps steps rather than calls within a step.
#
# outcome is derived by agent/src/recorder.ts against
#   TRACE_EXPECT="agent/src/config.ts,gpt-5-mini"
# success = the final answer carried both facts, partial = one, failure = none.

row	descriptions	tools_called	outcome	note
1	precise	search_files	failure	
2	precise	list_files	partial	
3	precise	list_files	partial	
4	precise	search_files	success	
5	precise	list_files	partial	
6	thin	list_files	partial	
7	thin	list_files	partial	
8	thin	list_files	partial	
9	thin	list_files	partial	
10	thin	list_files	partial	
11	precise	search_files	success	
12	precise	search_files	failure	
13	precise	search_files	failure	
14	precise	search_files	partial	
15	precise	search_files	failure	
16	precise	search_files	failure	
17	precise	list_files	partial	
18	precise	list_files	partial	
19	precise	list_files	partial	
20	precise	search_files	success	
21	precise	list_files	partial	
22	precise	search_files	failure	
23	precise	search_files, list_files	success	
24	precise	list_files	partial	
25	precise	search_files	failure	
26	thin	list_files	partial	
27	thin	list_files	partial	
28	thin	list_files	partial	
29	thin	list_files	partial	
30	thin	list_files	partial	
31	thin	list_files	partial	
32	thin	list_files	partial	
33	thin	list_files	partial	
34	thin	list_files	partial	
35	thin	list_files	partial	
36	thin	search_files	failure	
37	thin	list_files	partial	
38	thin	list_files	partial	
39	thin	list_files	partial	
40	thin	list_files	partial	
41	precise	list_files	partial	shipped trace-a-precise.json
42	thin	search_files	failure	shipped trace-b-thin.json