Agents are moving from tool use to paid work.
But paid work needs more than wallets.
It needs contracts, evidence, settlement, and reputation.
Thatโs Foundation Protocol:
the accountability layer for the agent economy.
Watch the feature overview โ
๐ซฑ Introducing ๐๐๐ฎ๐ซ๐๐ฅ ๐๐จ๐ฆ๐ฉ๐ฎ๐ญ๐๐ซs:
๐ฐ๐ก๐๐ญ ๐ข๐ ๐๐ ๐๐จ๐๐ฌ ๐ง๐จ๐ญ ๐ฃ๐ฎ๐ฌ๐ญ ๐ฎ๐ฌ๐ ๐๐จ๐ฆ๐ฉ๐ฎ๐ญ๐๐ซ๐ฌ ๐๐๐ญ๐ญ๐๐ซ, ๐๐ฎ๐ญ ๐๐๐ ๐ข๐ง๐ฌ ๐ญ๐จ ๐๐๐๐จ๐ฆ๐ ๐ญ๐ก๐ ๐ซ๐ฎ๐ง๐ง๐ข๐ง๐ ๐๐จ๐ฆ๐ฉ๐ฎ๐ญ๐๐ซ ๐ข๐ญ๐ฌ๐๐ฅ๐?
Beyond today's conventional computers, agents, and world models, Neural Computers (NCs) are new frontiers where computation, memory, and I/O move into a learned runtime state.
We ask: whether parts of runtime can move inward into the learning system itself. This is our first step toward the Completely Neural Computer (CNC): a general-purpose neural computer with stable execution, explicit reprogramming, and durable capability reuse.
Work done with Mingchen Zhuge (@MingchenZhuge), Changsheng Zhao, Haozhe Liu (@HaoZhe65347 ), Zijian Zhou (@ZijianZhou524 ), Shuming Liu (@shuming96 ), Wenyi Wang (@Wenyi_AI_Wang ), Ernie Chang (@erniecyc ), Gael Le Lan, Junjie Fei, Wenxuan Zhang, Zhipeng Cai (@cai_zhipeng ), Zechun Liu (@zechunliu ), Yunyang Xiong (@YoungXiong1 ), Yining Yang, Yuandong Tian (@tydsh ), Yangyang Shi, Vikas Chandra (@vikasc), Juergen Schmidhuber (@SchmidhuberAI)
Given a fixed budget: scale out with more agents, or scale up a single one?
We derive a sharp criterion comparing the "organization exponent" to the single-agent scaling exponent to determine when multi-agent expansion wins and when it doesn't.
๐ arXiv: https://t.co/q1pQN4MSYj
structural prototype construction and multi-objectiveโdriven optimization.
The framework remains stable and controllable under multiple continuous property constraints, and achieves clear improvements over strong baselines on both drug discovery and materials design tasks.
๐
In collaboration with Samsung, we propose a controllable, multi-stage molecular generation framework that turns molecular generation from a one-shot black box into two explicit stages.
๐ arXiv: https://t.co/UoH6hRRDTB
๐ค https://t.co/TCpOgP3yX0
Takeaway: if we want agents that truly improve over time, we need skill lifecycle managementโdebugging, freezing, and refactoring as first-class learning operators.
Glad to share our new work โEvolving Programmatic Skill Networks (PSN)โ
Arxiv: https://t.co/1Jca840qcJ
HuggingFace: https://t.co/6XDEjA7UDS
We treat skills as executable code with explicit pre/post-conditions, and the skill library is wired into a compositional evolving graph
Iโm excited to share that Iโve received an Amazon Research Award for my proposal Foundation Agents and Protocol for Collaborative Agentic AI at University of Montreal. Learn more about the program on the @AmazonScience website: https://t.co/tFxoePQYbr
#AmazonResearchAwards
Is ReAct the final for LLM Agents? ๐ค
Not anymore. ReAct is stuck step-by-step. Planning agents separate plans from actions.
Introducing ReCode: agents control their decision granularity, like humans do.
20.9% better, 78.9% cheaper, 3.7ร data efficient. The ReAct era is over.
1/5
Mila's annual supervision request process is now open to receive MSc and PhD applications for Fall 2026 admission! For more information, visit https://t.co/r01eLcY1P4
๐คCheck The Hitchhikerโs Guide to Agents HERE๐ค
Our Foundation Agents Survey V2 level up to 396 pages โ every chapter is a full-on survey itself!
๐ง Agent Framework & Components
๐ World Model & Memory
๐ Self-Evolution
๐ฅ Multi Agents
๐ก๏ธ Safety
1/4