🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

It's too bad that "compound systems" lost out to "agents" to describe systems built by composing LLM calls Now we're stuck calling everything agents and having to provide a definition every time we use the word

AI agents are the future—but they're stuck with the infra of the past. Legacy infrastructure is blocking their full potential. We've raised a $21M Series A to unblock them. Here's exactly what agents really need to excel: ⚡ A computer they can quickly boot up 📁 Ability to

We have Notion AI at home 📝

this is really cool, will sargent from the @letta_ai discord wrote this cool AI agent chef tersesystems.com/blog/2025/07/2…

if other startups aren't shamelessly cloning your website/product you're ngmi 📈

Charles Packer Satwik Singh I stole the Letta website. I think it's way better now

if you have an iphone 13 mini (tiny phone gang) do not i repeat do not upgrade ios for liquid glass i turned a perfectly usable phone into a 15fps brick 😿

the world famous bluesky bot is now cross-platform shared agent memory across bluesky & x

i almost forgot language models could hallucinate this hard💀

WFH (work from home) -> WFP (work from prison)

Maziyar PANAHI let me finish my beer first and ill call my boys to shoot the new model

America needs to take open models more seriously. This summer the early lead in open model adoption of the US via Llama has been overtaken by Chinese models. With The American Truly Open Models (ATOM) Project we're looking to build support and express the urgency of this issue.

i can't believe openai is going to drop GPT-5 before google drops token streaming for tool calls in the gemini api 😭

(SF 911): Sir, what is your emergency? (Me): cough…I’ve been stabbed… (SF 911): Where? (Me): at the intersection of art…cough…and technology…ugh… (SF 911): And where is that, sir? (Me): the arena…aagghhh…you can’t miss it…

this is what it felt like to go from gpt-3.5 to gpt-4

server-side context-management means you can build an agent that beats claude code in <200 lines of code (and all open source)

gpt-4 funeral is going to do numbers 😿

love the comparison, i have some fond memories of reading and attempting to repro GQN back during my phd (for an occlusion prediction paper)... and i only graduated ~1 yr ago what a time to be alive