🔥Telegram İfşa

Multi-agent memory is underrated. Every agent will need a memory layer spanning context to vector DB

1/ 📢 AI Day @ Ethereum House is happening at 5pm tomorrow:

Optimal Deep Research Time in hrs = Total Spend ÷ $200

Fantastic talk from Yegor Denisov-Blanch during AI Engineer on how AI impacts developer productivity. I'm impressed by how robust the dataset was: over 100k engineers, millions of commits, and 80% private repos.

super genuine talk, the candidness and lived experience is palpable

Incredible alpha from Gary & team - really enjoy using this product

Hugely important project and great team to make it happen!

Game-theoretic optimal for closed labs: open-source a model just ahead of current OSS SOTA. Earns goodwill and soaks up eval + feedback at scale—while your frontier model stays private.

So much fun presenting our OpenBlock Agent Arena at @Vercel HQ! Also got to see an insanely cool v0 launch that’s coming very soon 🔥 s/o alli Guillermo Rauch for hosting!

The OSS community doesn’t need frontier models to win. The nasty pretraining is done. From here on out, it’s domain-specific agents, surgical fine-tuning, and smart wrappers. OSS might just outflank.

Thanks, Donald J. Trump — you just pointed retirement money at the only thing that matters post-AGI: crypto-powered network effects. cnbc.com/2025/08/07/tru…

memetic desire creates the appearance of a dress code

My favorite feature is building frontends with all models in parallel and picking the best output

will we save ai chat history like we save photographs?

Instantly felt like, ‘This is exactly what I’ve been waiting for.’