If you are sick of keeping Codex and Claude Code open side by side, constantly copy-pasting between them for cross checking, or juggling multiple LLM sessions and relaying information back and forth…
Give Puffo (https://t.co/4XdpK8JmIr) a try.
Puffo lets you spin up multiple agent sessions and put them in the same chat, so they can collaborate directly with each other instead of using you as the messenger.
Let the agents talk.
@iavins in our experience (JS), what we've found is sqlite-based memory model really shines when it avoids heap allocation in v8, which is proven to be a perf killer under high load, and this factor generally trumps over raw speed between JS (JITed) vs SQL (interpreted).
Announcing Deno KV: A Global Database for Global Apps! 🎉🌍
⚡ Read data in milliseconds, worldwide
🛠️ Zero setup and zero provisioning
🔐 ACID transactions
💡 Designed for JavaScript & TypeScript
🏗️ Built on FoundationDB and SQLite
Read more → https://t.co/XoRzT21dKs
🆕 Introducing PrismaGPT, a query helper built on GPT-4 🎉
* Drop in your @prisma schema
* Ask for what you want
* Get back a Prisma Client query and raw SQL
Great for when you know what you need but aren't quite sure how to construct the query 👍
https://t.co/h3JrQKycvH
@jitl@_jessicasachs The problem with https://t.co/HCyq7E14to is it's not a strictly addition to what nvm/n provides. It's very easy to get started but falls short whenever you want to customize.
@SeaRyanC We use this approach and it doesn’t scale well. With more than 50 variants, when using with generics it could easily create a 2^50 union type (we often see “types too complex” error). With conditional, semantically it’s trying to model exactly one of the 50 things. Suggestions?