OAI deep research > Sonnet 3.7 reasoning
its not even close

4 diff usecases, all software/blockchain related
1. researching pectra upgrades and diff smart wallet approaches
2. proof of concept clicker came to demonstrate blockspeeds
3. strategy doc around grants program
4. post-everywhere (twitter, wc) personal app to replace yup

tried claude on 1, 3 and 4 w/ frustrating results. I think claude got the app architecture more correct for #4, but openAI was better at finding a few workaronds for API constraints on WC and Twitter.

comparing against some different types of tasks, deep research came out ahead w/ much more usable outputs w/ minimal edits on all 4 (tho slight edge for 3.7 Sonnet on initial architecture sketch)

Founder, designer, & product guy at /basepaint and /eigenlayer

comparing against some different types of tasks, deep research came out ahead w/ much more usable outputs w/ minimal edits on all 4 (tho slight edge for 3.7 Sonnet on initial architecture sketch)

https://warpcast.com/zherring/0x0ca7de9a

Not exactly but getting a summary for setup to throw in Cursors composer is 🔥

does deep research code? i don't think i realized that