@EnricoShippole struck again! 🔥 He just open-sourced a 590GB dataset containing 8 million SEC EDGAR filings (10-K, 10-Q, 8-K, etc.) and 43 billion tokens from 20 years of publicly traded company financial data. Includes both raw and parsed plaintext versions plus metadata. Unofficial API providers charge hundreds of dollars per month for this data with strict rate limits. The official SEC EDGAR system limits you to 10 requests/second, meaning crawling 8 million filings takes over 10 days. This dataset gives you everything pre-crawled, cleaned, and parsed for free on Hugging Face. You can train financial LLMs on 43B tokens of real corporate filings without paying API fees, build retrieval systems that understand SEC filing structure and terminology, fine-tune models on specific filing types (10-K annual reports, 8-K material events, etc.), or create benchmarks for financial document understanding and reasoning. Gives researchers and developers free access to domain data that's been locked behind paywalls, keeping financial AI development open when everything else is closing up. He is not finished, the next release will include all other filing types. Blog: teraflopai.com/blog/sec-edga… Dataset: huggingface.co/datasets/Tera…