buffaly
@buffaly_ai
Followers
38
Following
96
Statuses
277
Inventor of ProtoScript, Semantic Database, and Buffaly NLU. Heavy into graph based programming. Perfect AI agents. Beware the Llamas
Orlando, FL
Joined January 2025
Check out our white paper on Intelligence Factory. We've solved * Incremental learning * AI without hallucination * Controllable and predictable agents It's not theoretical. We use them everyday to process thousands of claims for our partners. Real AI, real results.
0
0
0
What's worse, they don't even have to reject the claims. We see hundreds of claims a day that just sit in a queue. In this modern age, where "AGI is just around the corner" insurance companies can't figure out how to process claims without a human calling to remind them! Ask a medical biller. They know what I'm talking about. Fight back with AI.
@unusual_whales If you didn't already know, United Health Care has a policy whereby their AI and automation claims receipt software automatically rejects roughly 30% of the claims. No reason.....just instant rejection. So sad.........We can fight back with AI.
0
0
1
Did you know that a significant portion of the time, insurance billers have to call to find out the status of a claim? It can be 30 minutes of their time to find out why $24 wasn't paid. Insurance companies do this shit on purpose. They don't want to pay. They put an automated system that the small practice can't afford to work around. We're fighting back! Our software recovers claims, cuts processing time, and generally cuts through all the bullshit that keeps doctors from getting paid!!!
Big healthcare is SWALLOWING medical practices because the medical insurance industry is making it impossible for them to exist. We have a solution for that.
0
0
1
Our AI assistant is helping fight automated insurance denials. We're helping doctors paid for services so they can stay in business.
Intelligence Factory has a one click solution for your over-the-phone input needs.
0
0
1
@SullyOmarr How are you doing search across a large body of documents with this chunk size? How well does it work? From my experience you can't embed chunks that big.
1
0
2
It's the same reason every agent demo is to create a to-do list application from scratch. That's easy. Make a simple modification to my 10-year-old code base without introducing any new bugs.
People start asking questions. I recommend the CEOs to think about a new shiny rattle.
0
0
5
@omarsar0 Why would we ocr PDF? Unless it's a scan of a 1977 paper, why not just parse PDF to markdown for free?
1
0
3
@roninhahn @tszzl Operator was just testing their infrastructure for the deep research crawler. I don't believe it was meant to be a product.
0
0
0
Practically AGI I tell you...
o3-mini is the first LLM released that consistently gets this tic-tac-toe question correct. The summarized CoT is pretty unhinged but you can see on the right that by the end it figures it out.
0
1
3
@deedydas I sent 16 pages of deep research generated text, and one DB record with history to an llm today and it was 400,000 tokens. How can 6,000 pages of PDF be 1 million tokens?
1
0
7
@rosstaylor90 I don't understand. They said in their paper they don't use supervised fine tuning. It specifically says no supervised fine-tuning. Can you clarify? Am I missing something here?
1
0
1
Where does it get the rewards from? How do you have relative advantage for correct or incorrect mathematical questions. The deepseek paper's implementation of grpo is a bunch of lies. Propaganda. The truth will come out
0
0
2
@omarsar0 Serious question, is it just another prompting technique? Or is there a actual advance presented in this paper?
0
0
0
@bevenky @cursor_ai Seem suspect. With the release of canvas and o1, our entire company has moved off of cursor. I'm sure others are doing the same. I'd like to see some real numbers from them before buying into this hype
0
0
0