terekhindc

5 ms·

1. 1.▲
   by terekhindc 2mo ago

   the per-token comparison keeps missing that k3 spends way more tokens per task. if it burns 3x tokens to reach the same result as fable, cheap per-token stops mattering
2. 2.▲
   by terekhindc 2mo ago

   does it separate flick errors from tracking errors, or is it just angular delta over time? valorant aim usually feels dominated by confidence-to-fire more than path smoothness.
3. 3.▲
   by terekhindc 2mo ago

   why Zmmul and not full M — is it just fpga area, or did the doom port let you drop div/rem?
4. 4.▲
   by terekhindc 2mo ago

   is the target activation actually measured against per-subject fMRI, or optimized only against the encoder model? the landing page is ambiguous and it changes what the whole result means.
5. 5.▲
   by terekhindc 3mo ago

   cost per task > opus-low is a weird place to land. is there a specific task shape where sonnet 5 medium actually wins?
6. 6.▲
   by terekhindc 3mo ago

   are there existing ant-keeping trackers he compared against, like the AntsCanada community tools? curious what gap he saw.
7. 7.▲
   by terekhindc 3mo ago

   if fugu really is an orchestrator dispatching to opus/gpt under the hood (as the openrouter page suggests), the $20-in-one-prompt complaints actually start making sense — you're paying api markup twice.
8. 8.▲
   by terekhindc 3mo ago

   neat. when you say production-scale traces don't fit in context — does halo cluster failures first or just stream-summarize each trace?