PaperMachine - The AI analyst that shows its work.

Hey Product Hunt PaperMachine started from a question that kept bothering us: as AI becomes capable of doing more of the analysis itself, how do we know what actually happened between the question and the chart? Most AI tools optimize for the answer. We wanted to preserve the process. PaperMachine runs Python and R locally and turns an analysis into something you can inspect rather than simply trust. Every step can expose the code that ran, its output, and the state of the kernel. Every figure, table, or file carries provenance back to the code, execution log, conversation, and environment that produced it. If a human edits a chart, that becomes a new version rather than silently rewriting the original. That distinction matters to us. In research and serious data work, a plausible result is not enough. You need to be able to ask: Which data was used? What changed? What code ran? Under which environment? And can another person reconstruct the reasoning path? For the GPT-6 Astra Challenge, we are exploring the next step of that idea: pairing stronger long-horizon reasoning and multi-step execution with an analysis environment designed to keep the resulting work inspectable. We don't think the future of AI analysis is simply “more autonomous.” We think it has to become more autonomous without becoming less accountable. We'd especially value feedback from researchers, analysts, and builders: what would you need to see before you were willing to trust an AI-generated analysis enough to build real decisions or research on top of it? Thanks for taking a look at PaperMachine.