I see so many people obsessing over token efficiency barking up the wrong tree with bervity and compression and wotnot.
They're missing The Big Factor. Inference works *recursively*. Every token in the context is fed back through to predict the next token.
The big savings are in context lifetimes.
Jason Gorman
npub1x9…y23ah
2026-05-11 13:00:32 UTC
Author Public Key
npub1x9e8f27nz76l4hndqm69ntk47jzujxj6rv20qnt9khm8jpl98szsyy23ahSeen on
wss://relay.ditto.pubPublished at
2026-05-11 13:00:32 UTCEvent JSON
{
"id": "b6c591bdc283dcab38eded3e3e740028e482f7da498d56d36a5aab1e864487f3",
"pubkey": "317274abd317b5fade6d06f459aed5f485c91a5a1b14f04d65b5f67907e53c05",
"created_at": 1778504432,
"kind": 1,
"tags": [
[
"proxy",
"https://mastodon.cloud/users/jasongorman/statuses/116556066479785348",
"activitypub"
],
[
"client",
"Mostr",
"31990:6be38f8c63df7dbf84db7ec4a6e6fbbd8d19dca3b980efad18585c46f04b26f9:mostr",
"wss://relay.ditto.pub"
]
],
"content": "I see so many people obsessing over token efficiency barking up the wrong tree with bervity and compression and wotnot.\n\nThey're missing The Big Factor. Inference works *recursively*. Every token in the context is fed back through to predict the next token.\n\nThe big savings are in context lifetimes.",
"sig": "2a8d0b4a419adb7ab39583706618317a3554ed3e3cef3866e7a95914486c7c89b13e48ea6cb75031f49613ac105d6fda4bb87aa9935c79ae59648c6b2024c3cc"
}