it doesnt kill prompt engineering, but by god it simplifies tweaking model behavior from a vague art for wordcels toward a science for tensor rotators. more in ainews
@swyx
-
Anthropic’s Progress Toward Productizable AI Interpretability
By
–
IMO @AnthropicAI is very close to making a breakthrough in productizable interpretability. For ~4 years all we've had to really control LLMs is temperature/top_p and logit bias. We recently got `seed` and constrained structured output, with `interactive=false` on the way. But
-
Hyperwrite Assistant Limitations: Recording Requirements Instead of Direct Instructions
By
–
perhaps @mattshumer_ can correct me but i dont think hyperwrite assistant does what i ask. just like adept workflows, it requires me to record what to do first (exactly the thing i'm trying to skip) instead of just giving it instructions and then checking in on it every now and
-
Training Time vs Committee Discussions on Voice Selection
By
–
this will take like 2 hrs of training, 2000 hours of committees on what voice to train
-
Login complexity with Devin agent workflow preferences
By
–
yea i thought about it but its just tricky to do login on devin, i would normally prefer to directly enter my login and get to the right screen before i hit go on the agent but i get what u mean and yeah will do
-
Consider Llama 3 1M Context Instead of 8B Context
By
–
have you considered starting with @markatgradient
's 1m context llama 3 instead of just the 8b context? -
Limited access to AI tools and shortage of alternatives
By
–
i mean.. do you have adept access? nobody here actually has adept access so how can you say that? ya will try hyperwrite. but surprised there are so few options
-
Looking for one-off automation to fill out repetitive forms
By
–
doesnt look like what i'm looking for – i want a oneoff automation just filling out a form that i've filled out before
-
State of Browser AI Agents for Form Filling Tasks
By
–
what is the SOTA of browser AI agents these days? – I have a crappy government website form to do that takes >60mins to fill out
– i've already filled it out once before. have the end result in a PDF
– need to fill it out again with 99% the same info in the same order
– want to -
Running Evals: Scripts, Model Swapping Libraries, and Tools
By
–
how do you run these evals? is it just a bunch of scripts? do you use any libraries for swapping models? any other libraries useful in your eval?