engineering·Oct 1, 2026·4 min read

We built a test to find out how good duplicate-page detection actually is, including ours

We built a 19-page test file to measure how well PDF tools actually detect duplicate pages, then published our results, including what we don't catch yet.

engineering·Sep 27, 2026·18 min read

11 MCP Servers That Prove MCP Isn't Dead

MCP’s critics have a point: a bad server can make an AI workflow harder than a good CLI. But Azure, Notion, Figma, GitHub, PDFHaul, and others show why the protocol still matters. The question is whether an assistant can reach the right tools and help finish the work.

engineering·Sep 3, 2026·29 min read

How We Took Our Table Extraction Accuracy From 10% to 84%. A Case Study.

How we measured our own table extraction accuracy against a real, diverse document corpus, including the parts that didn't go well. Full methodology, real code, and the files that still aren't perfect.

engineering·Jul 24, 2026·8 min read

Free OCR for PDFs, Powered by Tesseract

PDFHaul's OCR runs on Tesseract, the open source engine behind free and unlimited scanning, without the per page fees that limit most free competitors.