I think the 4 I listed are the big ones. There are lots of smaller businesses thinking about this, but none have traction to me. Maybe I can dig up some emails from founders again.
Thanks, I see what you mean now. I kind of missed that point because I had a different idea of what counts as a LLM agent startup which is more Adept and Lindy and not so much ChatGPT and Langchain. But I think your take is fair too. I think another one for the list could be Inflection.ai.
Love your point on the contrast between Adept and Lindy.
I see lots of companies chasing this idea like AirOps / Fixie / Relevanceai in this space trying to create agents. The reality is that everyone has created agents for decades using rules-based programming. It's tedious, but works, and automates many mundane tasks. However, when something changes, or you need to deal with high variety, rules-based programming doesn't work. We will see if there is a good compromise between the two.
Yes, agreed that agents are not a new idea and that the weaknesses of rule-based agents are clear. But I do wonder if LLMs might be the replacement for rules, much like how supervised learning substitutes rules via a function learned from data.
I'm curious. What are some startups based around the idea of LLM as agents? If you are able to share?
I think the 4 I listed are the big ones. There are lots of smaller businesses thinking about this, but none have traction to me. Maybe I can dig up some emails from founders again.
Thanks, I see what you mean now. I kind of missed that point because I had a different idea of what counts as a LLM agent startup which is more Adept and Lindy and not so much ChatGPT and Langchain. But I think your take is fair too. I think another one for the list could be Inflection.ai.
Love your point on the contrast between Adept and Lindy.
I see lots of companies chasing this idea like AirOps / Fixie / Relevanceai in this space trying to create agents. The reality is that everyone has created agents for decades using rules-based programming. It's tedious, but works, and automates many mundane tasks. However, when something changes, or you need to deal with high variety, rules-based programming doesn't work. We will see if there is a good compromise between the two.
Yes, agreed that agents are not a new idea and that the weaknesses of rule-based agents are clear. But I do wonder if LLMs might be the replacement for rules, much like how supervised learning substitutes rules via a function learned from data.