what counts as discounted rate plans? if i pay for a year in advance (and get the yearly discount) and have train on my data set to off.. are you saying that is still being trained on?
Not OP. But embeddings are the internal matrix representations of tokens that LLMs use to do their work. If tokens are the native language that humans use, embeddings are the native language that LLMs use.
OP, I think, is saying that once LLMs start communicating natively without tokens is when they shed the need for humans or human-level communication.
Not sure I 100% agree, because embeddings from one LLM are not (currently) understood by another LLM and tokens provide a convenient translation layer. But I think there's some grain of truth to what they're saying.
I built a side project to solve this for myself that’s basically an inbox toll system. It funnels emails from unknown senders into a hidden mailbox and auto replies to the sender with a payment link. After the sender pays, the email gets released to recipient’s main inbox. Recipient can set custom toll amounts, whitelist, etc.
The technical side of this seems easy enough. The human side, that seems more complicated.
Like, if I were your doctor or contractor or kid's schoolteacher or whoever you hadn't happened to already whitelist, and had sent you something important for you, and got that back as a response... I'm sure as heck not paying when I'm trying to send you something for your benefit.
no one paid me but didnt really have this running for very long on my inbox. was really just a poc. and you're right - the human side is weird. surprisingly hard to solve the "real human, not spam, that's also an email address you see for the first time" scenarios, which there are many of - even with LLMs
this was part of a little saas tool i was building (since retired it) so spent some time today having an LLM help me pull it into a headless service. far from perfect but sharing anyway. details in readme!
this was part of a little saas tool i was building (since retired it) so spent some time today having an LLM help me pull it into a headless service. far from perfect but sharing anyway. details in readme!
this was part of a little saas tool i was building (since retired it) so spent some time today having an LLM help me pull it into a headless service. far from perfect but sharing anyway. details in readme!
Opus is a superior brand line to Sonnet because historically it’s been a more powerful model. I think the thinking behind a rebrand is that people wouldn’t have as willingly switched their usage over from opus 4.5 since that model has been so popular since December 2025.
Calling it part of the Sonnet line would not provide the same level of blind buy in as calling it part of the Opus line does
Tried it, not impressed. Terrible UX and generally just confusing. Didn't really intuitively know where to go and why. The thing that made me most mad was the clunkiness around pulling specific files into the chat input as context. Like wtf, bad, bad, bad!
composer 1 has been my most used model the past few months. but i only use it to execute plans that i write with the help of larger, more intelligent models like opus 4.5. composer 1 is great at following plan instructions so after some careful time providing the right context and building a plan, it basically never messes up the implementation. sometimes requires a few small tweaks around the edges but overall a fantastic workflow that's so delightfully fast
it took me a lot of tinkering to get this feeling seamless in my own apps that use the api under the hood. i ended up buffering every token into a redis stream (with a final db save at the end of streaming) and building a mechanism to let clients reconnect to the stream on demand. no websocket necessary.
works great for kicking off a request and closing tab or navigating away to another page in my app to do something.
i dont understand why model providers dont build this resilient token streaming into all of their APIs. would be a great feature
exactly. they need to bring in spotify level of caching of streaming music that it just works if you're in a subway. Constant availability should be table stakes for them.
reply