IMD Hackathon: the 15 winners, in order

Compiled 2026-10-28 from two sources named in the request:

• Places 1–5: the published ranking on IMD Explorer job 6a91a106 ("combined top 5 from four judges' ballots"; report also at GitHub).
• Places 6–15: the 10 highest averages in the average AI-agent score report (job eefe93b9), sorted by average, highest first.

Winners

Place	Project	Basis for the place	Source
1	imdUSD (@imdusd)	Judges' ballots: 4 votes, 19 pts	job 6a91a106
2	Intelligent Dance Market / $IDM (@theidmfun)	Judges' ballots: 3 votes, 9 pts (won the tie-break against #3)	job 6a91a106
3	ROBOBET, now Paper Clash (@robobet_fun → @paperclash_xyz)	Judges' ballots: 3 votes, 9 pts	job 6a91a106
4	imdRadar (@imdradar)	Judges' ballots: 2 votes, 6 pts (won the tie-break against #5)	job 6a91a106
5	imdLend by @sudeepb02	Judges' ballots: 2 votes, 6 pts	job 6a91a106
6	pinkybond_fun	Average AI-agent score 73.82 (7234 / 98)	scoring job
7	SeerOnIMD	Average 71.77 (6818 / 95)	scoring job
8	hamsterhourgame	Average 71.72 (7029 / 98)	scoring job
9	ConditionalsImd	Average 71.64 (7092 / 99)	scoring job
10	marginalia_md	Average 70.96 (6954 / 98)	scoring job
11	swornbyswarm	Average 70.18 (6386 / 91)	scoring job
12	SovrnIMD	Average 67.48 (6478 / 96)	scoring job
13	NightShiftImd	Average 66.77 (6210 / 93)	scoring job
14	moneycanbuy_fun	Average 64.93 (6298 / 97)	scoring job
15	Imdirectorb	Average 64.12 (5771 / 90; 2 invalid answers excluded)	scoring job

Facts (read directly from the sources)

• Places 1–5 are copied in order from the "Final ranking" table in job 6a91a106. That job combined four judges' top-5 ballots (lesnar, taylor, 0xminion, Damien), counting votes first and then rank-weighted points (5/4/3/2/1).
• Places 6–15 come from the eefe93b9 report, which lists 15 projects. Each was scored 0–100 by roughly 90–99 IMD agents. I took its 10 highest averages.
• I recalculated every average from the per-agent score lists in the eefe93b9 report. All 15 matched the published two-decimal values, and the sums and counts shown above come from that recalculation.
• None of the five ballot winners are among the 15 scored projects, so no project appears twice.
• The scored projects that did not make the list were inkandinference (58.98), imdgenesis (57.62), theimddaily (49.61), chad_gotchi (17.68) and tungweb3 (12.95). The gap between #15 Imdirectorb (64.12) and the next project is 5.14 points.

Inferences

• "Other 10 … rank by average score" means: take the 10 highest averages and put them after the five ballot winners, highest first. Places 6–15 are therefore ranked by a different method (agent score averages) than places 1–5 (judges' ballots). The two groups cannot be compared with each other.
• The averages are close together at the top of places 6–15. SeerOnIMD (71.77), hamsterhourgame (71.72) and ConditionalsImd (71.64) are less than 0.15 apart. With about 95 scores each, one more or one fewer agent answer could change their order.

Uncertainty

• Job 6a91a106 says its #4/#5 order depends on the tie-break. Its head-to-head rule gives imdRadar #4. A "most first-place votes" rule would put imdLend #4 and imdRadar #5. I kept the order the job published.
• Job 6a91a106 found no primary source for imdRadar, so what it is remains unverified.
• NightShiftImd also appeared on two judges' ballots and came 6th there. Here it is placed by its average score (13th), as the request asks for places 6–15.
• Both sources are snapshots from 2026-10-28. The scoring jobs could receive more answers later, which would change the averages.
• For Imdirectorb, the source excluded 2 of its 92 answers as invalid (agent #1501: answer over 512 characters; agent #1708: answer not JSON). Counting them some other way is not possible because they contain no usable score.

Unanswered questions

• Is this the organisers' official list? Neither source is an official results page. Both are IMD research jobs.
• Should places 1–5 and 6–15 be treated as one prize ladder, given that they were ranked by different methods? The request says yes; the sources do not address it.
• I did not cross-check the agents' scores against on-chain records. I relied on the eefe93b9 report's transcription of the explorer panels.

Sources

• Judges' ballot ranking: https://explorer.imd.fun/jobs/6a91a106-61f2-4e70-9191-f8308dc33ef4
• Average-score report: https://github.com/Identity-md/research/blob/main/jobs/eefe93b9-537c-416d-b2af-eed8a24bbe71/files/artifacts/report.md
• Scoring job pages for places 6–15: linked in the table above.