It's experimental but you can use steer conversation to influence the model mid-turn
LLMS
-
Token Undertrain and Variance in Language Model Training
By
–
Many papers! The tokens were significantly undertrained already last year, and the trend towards even fewer tokens likely just makes the variance worse! NanoGPT's small competition hides variance problems by gathering statistics over many seeds, IMHO just switch to medium size.
-
Claude users have an insane competitive edge over ChatGPT
By
–
Claude users have an insane competitive edge over ChatGPT users.
-
Apple Plans to Turn Siri into AI Chatbot ‘Campos’
By
–
BREAKING : Apple plans to turn Siri into a full featured Campos AI app according to Bloomberg. “The new Siri experience is an AI chatbot code-named “Campos” that will be embedded deeply into iPhone, iPad, and Mac operating systems and will replace the current Siri interface.”
-
Comet browser agent powered by Opus 4.5 on Perplexity Max
By
–
Comet browser agent is now powered by Opus 4.5 on Perplexity Max plans. Testing time
-
Opus 4.5 becomes default browser agent model for Comet
By
–
Opus 4.5 is just absolutely insane as an agent orchestrator. So, we’re making it the default model for the browser agent on Comet for all Max subscribers.
-
Claude adoption surge among founders in 18 months
By
–
I attended a YC event in summer 2024 and the room was polled on who uses Claude. About 10% of hands went up. Fast forward to now, when I talk to founders, not only are they all using Claude, but many tell me Claude is what lets their team of 2 build like a team of 10 and that
-
Claude’s Constitution Released and Recommended
By
–
Claude's constitution is out! Highly recommend taking a look
-
Vibe coding loop: app that codes another vibe coding app
By
–
Vibe code a vibe coding app that vibe codes another vibe coding app Vibe coding loop