Charles Packer
@charlespacker
Followers
3K
Following
4K
Media
111
Statuses
792
CEO at @Letta_AI // creator of @MemGPT // AI PhD @berkeley_ai @ucbrise @BerkeleySky
SF
Joined March 2014
đź’¤ sleep-time compute: make your machines think while they sleep -> over the past several months we (at @Letta_AI) have been exploring how effectively utilize "sleep time" to scale compute. the concept of "sleep-time compute" is deeply tied to memory -.
arxiv.org
Scaling test-time compute has emerged as a key ingredient for enabling large language models (LLMs) to solve difficult problems, but comes with high latency and inference cost. We introduce...
We're excited to release our latest paper, “Sleep-time Compute: Beyond Inference Scaling at Test-Time”, a collaboration with @sea_snell from UC Berkeley and @Letta_AI advisors / UC Berkeley faculty Ion Stoica and @profjoeyg .
2
11
93
hyped to see the open LLM vtuber community embrace @Letta_AI đź’ś. the anime vtuber stack of today is the enterprise AI stack of tomorrow
0
1
10
RT @theinfrapod: 🧠Forget bigger models—AI needs memory. On this week’s Infra Pod, @tnachen & @ianlivingstone talk with @charlespacker….
0
2
0
RT @tnachen: Memory is more important than models - said @charlespacker in his hot take. Come listen to why in our pod!.
0
1
0
memory is bound to become far more valuable than the model. with enough advances in self-improvement, a single agent will carry the same memory forward through many model generations. memory compounds in value, model weights depreciate.
FROM THE LIVE SHOW: @charlespacker, co-founder & CEO of @Letta_AI, on why memory may matter more than the model. “Even for frontier labs like Google or OpenAI, memory is actually more valuable than the model. An agent might run for 10 years, swapping through GPT-5, 6, 7, 8…
2
3
11
who’s going to be the first ai startup to hit 100M ARR in one-month-free course subscriptions.
app guys are turning into course bois . makes you wonder how bad the profits are from these million-dollar mrr apps really are
0
0
4
if your cofounder isn’t a clanker glazer ur ngmi.
types of new SF guys . - clanker glazer.- “can i get an intro” man.- forward deployed mormon engineer.- creatine suppository influencer.
0
0
4
hopping on @MatthewBerman 's stream in minute to chat memory memory memory 👇.
We're going LIVE. Guests:.* @GregKamradt from Arc Prize.* @charlespacker from Letta.* Allen Liu from @Skywork_ai . Join us!.
0
0
2
RT @johncoogan: Ender, that wasn’t an RL environment with a verifiable reward, those were real Amazon orders you placed.
0
122
0
RT @YadKonrad: fuckin Opus 4.1 man, hahaha .this is like the worst type of coworker, makes a 3 pager to get buy-in on doing X, then with on….
0
1
0
RT @Letta_AI: A new article by @WIRED covers "sleeptime compute", interviewing @charlespacker @ClemDelangue (@huggingface) @hwchase17 (@Lan….
0
3
0
RT @lisbonai_: We all get annoyed when LLMs forget what've already mentioned. Why doesn't it keep that memory?. Meet @Letta_AI - the start….
0
4
0
we’re so over it’s so back.
1
0
2
cohere branding continues to crush. on-prem llm never looked this cool.
Big news today: we’ve raised $500M to grow @cohere, and have added some incredible new leaders to our team!. We’re fortunate to build Cohere to be the world’s best choice for enterprises alongside some of the best investors, partners, and customers in the business.
2
0
16
if you want to see this IRL:
jobs.ashbyhq.com
Letta Jobs
ex faang engineer joining a startup and meeting his cracked zoomer coworker pair programming
0
0
8
RT @giffmana: Update on this: the reason Microsoft (and probably Amazon) were so much worse at serving gpt-oss is that they ignored reasoni….
0
22
0
many such cases.
Gonna be straight with you guys. gpt-5 is nowhere near as good in Cursor as it was when I was using it a few weeks ago. Actively working with OpenAI and Cursor to figure out what's up.
0
0
2
RT @Letta_AI: Stateful agent meetup prep at Mindspace! See everyone soon. I'll show off Deep Thought.
0
1
0