Whoever is making these chat templates for models: please st...

npub12262qa4uhw7u8gdwlgmntqtv7aye8vdcmvszkqwgs0zchel6mz7s6cgrkj

hex

42aa604bf76b7ad76b737853b7a17664675675050b72d707b4230071ec43ba3f

nevent

nevent1qqsy92nqf0mkk7khddehs5ah59mxge6kw5zskukhq76zxqr3a3pm50cprpmhxue69uhhyetvv9ujuem4d36kwatvw5hx6mm9qgs99d9qw67th0wr5xh05de4s9k0wjvnkxudkgptq8yg83vtulad30gnmf8em

Kind-1 (TextNote)

2026-05-01T12:50:17Z

Whoever is making these chat templates for models: please stop.

ChatML is already shit. Then model labs are now bolting on reasoning via non-special tokens, then trying to make tool calls by adding some absurd XML-like <tool_call> prefix to assistant messages to start tool calls, etc.

Special tokens were created for a reason and now even the competent labs are doing this type of garbage

Raw JSON

{
  "kind": 1,
  "id": "42aa604bf76b7ad76b737853b7a17664675675050b72d707b4230071ec43ba3f",
  "pubkey": "52b4a076bcbbbdc3a1aefa3735816cf74993b1b8db202b01c883c58be7fad8bd",
  "created_at": 1777639817,
  "tags": [],
  "content": "Whoever is making these chat templates for models: please stop.\n\nChatML is already shit. Then model labs are now bolting on reasoning via *non-special* tokens, then trying to make tool calls by adding some absurd XML-like \u003ctool_call\u003e prefix to assistant messages to start tool calls, etc. \n\nSpecial tokens were created for a reason and now even the competent labs are doing this type of garbage",
  "sig": "f9ea54b712b6507a14129b588ee58dc487cc953ba337f12b7b5c225f731294f3e45dfc4c5c68368787bdfe158a01698ba1791336c4cc17d8c77a788deedf39f3"
}