Andrew Ng has released a new document extraction service utilizing OCR models, which faced criticism during testing by users. The service exhibited issues with accuracy, particularly in complex financial documents. Users reported high rates of hallucinations and fabricated values, reminiscent of challenges faced with older models like GPT. The tool's capabilities in parsing were deemed inconsistent and underwhelming, leading to questions about its reliability. Comparisons with other tools like Claude highlighted mixed performance in document processing, sparking discussions on the credibility of claims made regarding the tool's "agentic" capabilities. Critics pointed out that while Ng is well-respected in the AI community for his educational contributions, the real-world effectiveness of his products remains questionable.