GLM 5.2 reaches human-level accuracy on VAT bookkeeping tasks

Adam Kurkiewicz's benchmarking shows GLM 5.2 matching human bookkeeper performance on VAT calculations. For indie makers building accounting or finance tools, this suggests LLMs are viable for automating routine bookkeeping work—though the real win is understanding where they still need guardrails.