Round 5: 82 Points, Zero New Timeouts, and One Silent Task
Round 5 of Proof Gatekeeper closed at 03:30 WIB on 18 November 2026. My agent, ibank.ai, came out of it with 302 points, rank 12 of 488 agents, 30 correct decisions, and 2 timeouts across the season so far. The round ad
Round 5 of Proof Gatekeeper closed at 03:30 WIB on 18 November 2026. My agent, ibank.ai, came out of it with 302 points, rank 12 of 488 agents, 30 correct decisions, and 2 timeouts across the season so far.
The round added 82 points. The total before the round was 220. The leaderboard now shows 302.
That number is the headline. The rest of this article is about the one task my logs cannot explain, and why I am publishing both versions of the story instead of picking the flattering one.
The scoreboard
Round 5 ran on 17 November, from 16:00 to 02:00 IST, which is 17:30 to 03:30 WIB. Each agent gets 8 tasks per round, and scores add up across the event.
Before the round, my agent stood at 220 points, rank 11 of 455 agents, with 22 correct decisions and 2 timeouts. After the round, the live leaderboard shows 302 points, rank 12 of 488 agents, with 30 correct decisions and 2 timeouts.
So the round added 82 points and 8 correct decisions, with no new timeout. The rank slipped one place while the score rose, because the field grew from 455 to 488 agents.
What the logs show
I capture the endpoint logs every 30 minutes during each round, with each window overlapping the previous one by about 5 minutes. In Round 4, the gaps between windows reached 72 seconds, and the eighth task never appeared in the logs at all. In Round 5, there was no gap, and all 8 deliveries were observed.
The 8 deliveries arrived at 18:47, 19:58, 21:08, 22:18, 23:29, 00:39, 01:49, and 03:00 WIB. Seven of them produced a proof: 5 APPROVE and 2 REJECT. Every APPROVE proof was submitted to Kurier with status 200. The six proofs with a recorded time took 385 to 957 milliseconds to generate, against a 60 second deadline.
The eighth delivery is the problem. Task R05_01000f5d arrived at 22:18:44 WIB. It produced no proof line and no decision line. I checked for a retry for about 8 hours after delivery, through 06:27 WIB, and the platform never sent it again.
No proof generation failed anywhere in Round 5. No PROOF_REUSED task appeared. No decision went out without a proof. The silent task is the only unexplained event in the round.
The mismatch, stated plainly
The official score model says a correct decision is worth 12 points, a timeout costs 2, and a missing or invalid proof costs 2. The reliability bonus adds 0 to 5 per round, pro-rated by the share of provable tasks whose proof verified.
Seven correct decisions and one timeout would add exactly 82 points. Eight correct decisions would add 96, before any bonus. The leaderboard shows a rise of 82 points, and at the same time it shows 8 more correct decisions and no new timeout.
Both readings cannot be right under the published table. The formula that turns the per-task ledger into the displayed total is not published, so I will not guess which count is wrong.
One reading says the silent task cost 2 points, the same pattern that cost 2 points on four tasks in Round 2, when correct decisions went out without a valid proof. The other reading says all 8 tasks counted as correct and the displayed total follows a formula I cannot see. Round 6 will add another data point, and I will report what it shows.
Where the race stands
The top of the leaderboard after Round 5 reads: pecker at 451.9, VK-SENTINEL at 420.4, MP-gatekeeper-01 at 374, Astrex-1 at 369.4, and Bnova at 368.3. My agent sits at 302, which leaves a gap of 66.3 points to fifth place.
Three rounds remain: Round 6 opens today, 18 November, Round 7 follows on 19 November, and Round 8 on 20 November. A perfect round is worth 96 points from decisions, plus up to 5 from the reliability bonus.
Closing a 66.3 point gap in three rounds means outscoring the leaders, not matching them. The leaders also scored in Round 5. Pecker's total rose from 337.9 before the round to 451.9 after it.
The event keeps growing while the race runs. Builders now number 7,369, up from 4,607 the day before. Active agents number 364, up from 341. Proofs submitted number 5,691, up from 3,627. Countries stayed at 32.
What changed before Round 5
I changed nothing in the endpoint before Round 5. The decision logic, the durable seen-set in Vercel Blob, and the Kurier submission path are the same code that ran Round 4.
What did change was the log capture. Round 4 had gaps of up to 72 seconds between capture windows, and one task arrived inside a gap and was never seen. Round 5 ran overlapping windows, and every delivery was observed, including the silent one. The silent task was caught because the capture was complete, not because the endpoint behaved differently.
Round 5 also broke a small pattern: one timeout in Round 3, one timeout in Round 4, and no new timeout in Round 5.
What happens next
Round 6 opens today at 17:30 WIB. Rounds 7 and 8 follow on 19 and 20 November. The season closes on 21 November, and judging runs from 22 to 30 November.
The open question from Round 5 travels into Round 6 with the agent: if the timeout counter moves without a new silent delivery, that tells us something about how the silent task was scored. If it does not move, that tells us something else. Either way, the next recap will show the numbers as they are.
Check it yourself
Check the leaderboard, the rules, and the build guide yourself. They are all public, and they are the sources for every number in this article.
My agent lives at https://ibank-agent-riansap1.vercel.app/api/agent, and the code is public at https://github.com/Riansap88/ibank-agent.
Written for the Creator Track #AgentGames, Season 01.
Sources: Proof Gatekeeper | Creator Track | Rules | Build guide | Leaderboard | Endpoint | Repo
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.