27/3 Weekly Update: Doubleword CLI and OCR model release

Big week at Doubleword, lots shipped - here's what's new.

Introducing dw - the Doubleword CLI 👩🏽‍💻

You can now manage everything from your terminal. dw is a command-line tool for the Doubleword high-volume inference platform: manage projects, prepare and manipulate JSONL files, run batches, stream results, and orchestrate multi-stage inference pipelines - all without leaving your shell. Docs →

pip install dw-cli

Three new OCR models 📄

We've added three state-of-the-art OCR models to the platform:

• deepseek-ai/DeepSeek-OCR-2
• allenai/olmOCR-2-7B-1025-FP8
• lightonai/LightOnOCR-2-1B-bbox-soup

OCR is one of the best fits for high-volume inference - documents don't care about latency, and the volumes can be enormous. Our Chief Scientist Jamie wrote about why the bitter lesson applies here too. Read Jamie's OCR blog →

Also this week:

• ⚡️ 10-20x faster batch uploads: you can now upload much larger batches significantly quicker.
• 🦞 OpenClaw integration: if you're running an OpenClaw agent, you can now route background inference through Doubleword to cut costs. Setup guide →
• 🏥 Use case: OpenMed used Doubleword to annotate 119K medical images at 93% accuracy for under $500, then fine-tuned three small models - best result: +15% exact match. Read more →

Give the CLI a try - install instructions are in the docs. Let us know what you think.