SUBNET TRACKER · All subnets
SN124 · Swarm
Every Bittensor Daily mention of SN124 Swarm, newest first. Price changes are snapshots from each edition.
24h change −1.47%
- What happened
- SN124 (Swarm)'s Sentinel competition evaluation queue backed up to 38 submissions, with each evaluation taking about five hours, pushing the estimated clearing time past a week. The team attributed the backlog to validators that had not updated and to limited compute capacity. It declined to reduce seed counts, skip evaluations, or stop the contest, citing the 0.5 TAO submission fee paid by participants.
- Community
- Miners repeatedly raised fairness concerns about the backlog and the dominant champion's continued submissions, with some calling for the contest to be paused. Several proposed a pre-stop or screener-phase mechanism to reject clearly non-competitive submissions before full evaluation.
- Official
- The team reported evaluation throughput rising from about 278 seeds/hour to about 358, then about 537 seeds/hour, with a target of about 800 seeds/hour, and announced that all validators had come back online. It said it would consider the proposed pre-stop mechanism.
24h change −1.49%
- What happened
- SN124 (Swarm) opened submissions for a new challenge called "Swarm Sentinel," quickly drawing more than 39 submissions and overwhelming the evaluation backend, with the team noting the queue reached over 30 models within minutes. The team confirmed a UI bug that failed to apply a 5x scoring penalty for missed or false alarms and said it is being fixed to show real scores.
- Community
- Miners debated whether the rapid submission queue was fair, with some calling the process unfair or disappointed in the team, while others proposed halting evaluation to prioritize top submissions given the 39-item queue would take days to clear. A moderator separately warned that channel spam and bickering could be an organized attempt to bury an exploit report and urged the team to check for automated cheating.
- Official
- The team said the UI-displayed score was a bug not applying the 5x penalty and is being corrected, and stated it will test the current evaluation system, expand compute once validators are onboarded, and will not ban the 39 submissions since each paid 0.5 TAO to register.