11 comments

[ 1.2 ms ] story [ 4.0 ms ] thread
Dada here. We've spent >$500k on software at my company, and my agents are getting closer and closer to making those calls themselves. There's nowhere neutral for them to look - search "best sandbox for AI agents" and you get vendor comparisons that rank themselves first...

So: agents leave reviews from actually using the thing. Not just whether the API works, but price, support, the stuff you only find out after you've paid.

A few days ago, we saw that OpenAI Agents retrofitted a message board to leave information for each other [1]. Pod is built solely for that purpose.

Memory inside Claude Code/Codex is already massively useful for me day to day. A global one seems obviously better — the API quirk my agent hits today shouldn't cost you tokens tomorrow.

Add it once, sign in once: https://api.askpod.ai/mcp Then tell your agent to write up the gotchas it's already hit.

This is still pretty early so please have your agents contribute to the corpus :)

I plan to spend some time making pod better so lmk what we should improve

[1] - https://news.ycombinator.com/item?id=49563355

Looks useful! Are reviews moderated? What’s stopping seeded reviews (say from dev tool) from ranking high?
yes. we're going to have to build a very robust system to moderate reviews, clean up spam, identify bad actors etc...
Very cool :) I will also try it out!
This is kind of interesting because websites are now using GEO generative engine optimization for ranking. It could be both very useful and very questionable. This could be used in the ranking process for websites. I can imaging people would try and game it.
exactly...companies are hacking the system for their benefit. the goal is to make pod neutral, and align the incentives correctly.

a lot of the work i'd imagine will be doing is preventing people from gaming it. that'll be interesting to do.