Why Existing Research Tools Weren't Enough
Before building Archimedes, I did the obvious thing: I looked around for tools that already solved the problem.
And to be fair, there are a lot of good ones out there.
I tried the usual suspects and explored tools that were good at pieces of the workflow:
- paper search
- literature discovery
- summarization
- note organization
- citation management
The problem was not that these tools were bad.
The problem was that none of them matched the exact shape of the workflow I wanted.
I wanted something more end-to-end.
Something that could start from a research question, pull relevant papers from sources like arXiv and OpenAlex,
analyze them with an LLM, and then produce a direct answer with paper-level evidence.
A lot of tools stop at discovery.
Some stop at summaries.
Some stop at note-taking.
Some are good for collecting information, but not for turning it into a structured report.
That gap mattered to me.
I did not just want more information.
I wanted synthesis.
I wanted a result I could hand to myself or someone else without manually doing all the glue work.
That was the key realization:
the value was not in any single step.
The value was in chaining the steps together into a workflow that felt natural.
So Archimedes became a research assistant instead of just a search tool.
It searches, analyzes, aggregates evidence, and exports the final answer as a PDF.
That is the part I kept coming back to.
It was also the point where the project stopped being "maybe I should build this" and became
"apparently I need to build this because nothing else quite fits."
In the next post, I will talk about the first version that made that idea real:
a simple CLI tool that worked for me before it worked for anyone else.
Top comments (1)
nice. interested. url repo?