Crazy to think 10m context ONLY means emails from 1 year...too many emails!

The next big leap in language modeling is going to come from finetuning them on agent trajectories. This will lead to a big accuracy improvement in end-to-end programming (e.g. SWE-agent), controlling desktop apps (e.g. OSWorld) and web browsing (e.g. SeeAct).

'bring yourself back online'

POV: you multiplied two matrices that are above the CA legal limit

looks like we're back on github.com/trending 👀📈

...it's never been easier to build your own personal agents with long-term memory - try out the alpha build of the memgpt UI app using Docker Build Cloud is here! 🐳🧱☁️!

.Kevin Lin 林冠言's phd thesis is a sneak peak into the future of llm programming: context (state) management

An evening we'll 'never forget'!

Felicis is excited to host the Berkeley LLM Meetup with Sarah Wooders and Charles Packer from MemGPT on Monday 5/20.

Speakers include Zongheng Yang from SkyPilot and Shishir Patil from Gorilla.

Sign up: lu.ma/berkeleyllmmay…

teach an LLM to read search results and you get RAG; teach an LLM to use a search tool and you get agents with memory

gpt2-chatbot taking no prisoners in the arena 😤

great turnout at the 2nd ever berkeleyLLM meetup Databricks love to see it 😍

Welp folks, we have gpt-4 at home

can’t wait for llama3 400b but only 8k context window?? 😭

do you: 🌁 live in the Bay Area + 🧑‍💻 work on OSS/LLMs + 🥂 enjoy OSS/LLM-themed afterparties with open bars??

...then there's only once place you should be next Tuesday 👉 lu.ma/berkeleyllm

Claude 3 + MemGPT = ❤️

✨ MemGPT now officially supports Google AI Gemini ✨

More tokens = more space for in-context memory 🧠

MemGPT + 1 million tokens = 🧠🚀

✅ large context windows...
✅ RAG...
✅ structured knowledge... inserted into context...

...sounds like MemGPT 👀

It was great hosting Vinod Khosla at Berkeley AI Research for a fireside chat with Charles Packer about entrepreneurship, research, & AI!

Check out the MemGPT + LlamaIndex webinar, where we (1) chat about the MemGPT system design 📝, and (2) build a Perplexity clone using the MemGPT API and developer portal! ⚡