Introducing deep-research - my own open source implementation of OpenAI's new Deep Research agent. Get the same capability without paying $200. You can even tweak the behavior of the agent with adjustable breadth and depth. Run it for 5 min or 5 hours, it'll auto adjust.

Internally, the agent will take the user input, break it down into different sub research threads that it'll run in parallel, and recursively iterate based on new learnings, spawns new research threads, and collect new knowledge until it reaches the necessary breadth and depth.

It's a pretty simple architecture, but o3 doesn't need much guardrails. Just give it the right tools, and let it follow its curiosity. Repo here:

Here's a report I ran on nvidia's new RTX 5000 series announcement, this is with breadth=3 and depth=2, took ~5 min.

The expected return of Greg Brockman to OpenAI will test how well he can coexist with CEO Sam Altman and others at the startup he co-founded.

This is great and love @aomniapp . Gpt Researcher introduced deep research years ago super recommend -

@aomniapp thank you! does gpt researcher also do recursive refinement?

Massive especially if you’re trying to find everything about your prospect

polished version coming to @aomniapp 🍓

This is a great hack David cost ploptimization is vital , instead of giving $200 it’s better to utilize open source capabilities

Awesome work. Can you update this so that the content of report cites the specific reference that it is sourced from? At the paragraph level would be sufficient.

The productionized version of this on @aomniapp will. Would love to learn about your use case, mind sharing on dm?