3un01a

> 3un01a has entered the chat

@taishiyade SFあたりですと、毎日何かしらイベントがたくさんあるので、Lumaとかのサイトでそう言うミートアップなどに参加するものお勧めです！ luma.com/sf

サンフランシスコ、オススメのやることありますか？

It’s always fun when you get Honcho to better understand you

I made a short 90s style anime featuring the two most popular AI agents. Complete with an opening and outro Hermes vs OpenClaw I use both agents in my workflow and I really like how both of them have very distinct mascots, that's how the idea came to life! Made in @runwayml @NousResearch @openclaw

Jensen has always been the “head Honcho” all along 🫡

Importantly, a month of 🫡Honcho-powered Hermes usage Perfect for supplementing the local markdown system Memory is a reasoning task... and Chuck's just beginning to see the benefits 🚀

Everybody is talking about recursive self-improvement (RSI) and meta learning. Here is my old 2020 talk about this [1]. It has aged well. Example: humans still define the starts & ends of trials of many modern meta learners. My RSI systems since 1994 LEARN to (re)define them [2]! [1] Meta Learning Machines in a Single Lifelong Trial (talk for workshops at ICML 2020 and NeurIPS 2021, based on earlier talks since 1994). Abstract: the most widely used machine learning algorithms were designed by humans and thus are hindered by our cognitive biases and limitations. Can we also construct meta learning algorithms that can learn better learning algorithms so that our self-improving AIs have no limits other than those inherited from computability and physics? This question has been a main driver of my research since I wrote a thesis on it in 1987 [2]. Here I summarize our work on meta reinforcement learning with self-modifying policies in a single lifelong trial (since 1994), and mathematically optimal meta-learning through the self-referential Gödel Machine (since 2003). Many additional publications on meta-learning since 1987 can be found in the RSI overview [2]. [2] J. Schmidhuber (AI Blog, 2020-2025). 1/3 century anniversary of first publication on recursive self-improvement (RSI) and meta learning machines that learn to learn (1987). For its cover I drew a robot that bootstraps itself. 1992-: gradient descent-based neural meta learning. 1994-: meta reinforcement learning with self-modifying policies. 1997: meta RL plus artificial curiosity and intrinsic motivation. 2002-: asymptotically optimal meta learning for curriculum learning. 2003-: mathematically optimal Gödel Machine. 2020-: new stuff!

the moment you realize the toast reference comes from doraemon's anki (memorization) bread... iykyk

had a blast checking out where all the magic happens at @itsalltruffles

Why Nothing Feels Interesting Anymore Fun, Reward, and What It Means for the Mind to Move open.substack.com/pub/leafsartem…

New Honcho package! @vercel-ai-sdk meet Honcho. Add persistent user memory to any Vercel AI SDK app. AI doesn't need to start from zero. 📦 npm install @honcho-ai/vercel-ai-sdk 🛠️ npx skills add plastic-labs/vercel-ai-sdk

@thrialectics congrats!! 👏

hmmm… cs.toronto.edu/~bonner/course…

Our paper "Statistical Impossibility and Possibility of Aligning LLMs with Human Preferences" was just accepted to the Annals of Statistics. It's known that reward models can't represent Condorcet cycles. We make this *quantitative*, and prove a possibility result for Nash-based alignment. 📄 arxiv.org/abs/2503.10990 w/ Kaizhao, @DrQiLong , Zhekun, and @JiancongXiao14 1/n

I think what people tend to not realize is that **anything** can be a substrate of memory and that there exists a hierarchical structure to how it can exist at different levels. From the parametric space of the model, to the agents internal memory component within the harness, to the agents external environment - including the surrounding artifacts as well as other agents and entities that are within that environment itself, and even collectively within agent communities themselves in multi-agent settings. Every layer of the hierarchy has carries some fragment of the memory representation and that it’s very hard to attribute the memory to a single point in the system. Smells a bit like “distributed representations”… 🤔

I swear the museum of tomorrow has Frutiger Aero vibes

Silly Tavern x Honcho Integration! Give SillyTavern a memory upgrade 🧠 Drop-in extension that wires Honcho into ST so your characters actually remember you across sessions, chats, and personas

And that’s a wrap at ICLR in Rio 🇧🇷! Great meeting many of you at the conference! Gonna miss the views and the food 🥹

maids in Japan are giving lectures on linear algebra