https://qrdlgit.github.io/erdosproblems/index.html?status=open&prize=yes
https://github.com/qrdlgit/erdosproblems/tree/main
The goal is to use this with openai agent builder which has this: https://platform.openai.com/docs/guides/agent-builder
I've used this UI successfully to do things like iterate chat exhaustively on discrete data inputs. In the past I've coded up complex agentic flows via API, but have found this to be more potent and accessible. Using transform, you can directly access json elements. eg: state.problems[state.counter].number I persist results using a custom mcp point, but there are other options provided by the builder.
Some use cases:
- Literature search: People have done this already, but with the agent builder above they can fine tune their prompts and customize how it is performed, and they can re-iterate over sets as they see fit.
- Generate a rubric / consistent analysis for listed problems
- Perform an ELO style analysis along some dimension(s)
- When new advancements in erdos related math are made or discovered, perhaps iterate through all of them to see their relevance
- Many use cases with comments, imagination is the limit
- OEIS integration.
..
Is there an equivalent yaml / API for the erdosproblem formulations or should I scrape from the tfbloom website?
https://qrdlgit.github.io/erdosproblems/index.html?status=open&prize=yes
https://github.com/qrdlgit/erdosproblems/tree/main
The goal is to use this with openai agent builder which has this: https://platform.openai.com/docs/guides/agent-builder
I've used this UI successfully to do things like iterate chat exhaustively on discrete data inputs. In the past I've coded up complex agentic flows via API, but have found this to be more potent and accessible. Using transform, you can directly access json elements. eg:
state.problems[state.counter].numberI persist results using a custom mcp point, but there are other options provided by the builder.Some use cases:
..
Is there an equivalent yaml / API for the erdosproblem formulations or should I scrape from the tfbloom website?