We all know the best data to train/eval agents should come from your own use-case, but what are the best ways to turn real-world data and workflows into agent learning experiences?
Share your ideas on this important topic at our upcoming AAAI 2027 Bridge: From Domain Data to Agentic Experience
https://t.co/ra8uATdnjf
Excited to co-organize AAAI-27 Bridge B25: From Domain Data to Agentic Experience!
How can we turn real-world data and workflows into meaningful learning experiences for AI agents?
We're bringing together researchers and practitioners across data-centric AI, agent learning, RL environments, and evaluation to explore this question!
๐ Montrรฉal | Feb 16โ17, 2027
๐ข Call for papers open! Deadline: Dec 1
We welcome papers on agent learning, benchmarks, evaluation, and real-world applications. Hope to see you there!
๐ https://t.co/KGpEMsSBTW
In terms of performance-vs-cost, @OpenAI's GPT-6.1 Sol & @GoogleDeepMind's Gemini-4 do represent big leaps in the computer vision capabilities of models in specialized scientific domains.
"Stateless Language Agents: Scaling Long-Horizon Automated Research"
Long-running AI research agents waste tokens revisiting old ideas or repeating experiments.
Stateless Language Agents (SLA) instead store research progress externally, giving each agent a fresh, focused context while an Advisor coordinates experiments.
This setup outperforms competing frameworks on software engineering and optimization tasks, matching the strongest baseline on one kernel optimization task with 93% fewer tokens.
https://t.co/RXScfTOdmh
Excited to share the latest work on Stateless Language Agents (SLAs)! ๐ Better performance doesn't have to mean more tokens. Proud of what @qizhengz_alex 's team has achieved!
AutoResearch agents keep burning tokens after they stop making progress. ๐ฐ๐ญ
From authors of ACE and Meta-Harness: Introducing Stateless Language Agents (SLAs). Stateful search, stateless agents. Advancing the performance-cost Pareto frontier of RSI.
Across 9 tasks, up to 1B tokens:
- Best final result on every task ๐
- 14.5ร fewer tokens to match the best kernel baseline ๐
- Advisor (the โcoordinatorโ) uses < 0.6% of total tokens ๐ซฃ
๐ Paper: https://t.co/umMktuZggL
NOT JUST HIGHER NUMBERS. Some more surprising findings in the thread below! ๐งต #LLM #AI #RSI #Agent
Proud of my team's contributions to Frontier-Bench (especially @ruansherry & Hui Wen Goh), helping make it the most ambitious benchmark for agentic work.
https://t.co/5hK4AIrZXH
Impressive model improvement from @AnthropicAI on this benchmark, one day after it launched!
@qizhengz_alex Congratulations!! ๐ฅณ ACE is an exciting step toward the agent research frontierโit moves beyond static context toward self-improving agents that can continuously adapt, reason, and improve in deployment.
๐ Our ACE paper is headed to #ICLR2026!
๐ Read the blog: https://t.co/HBeqTbLZSZ
This is just the beginning. We see ACE as part of a broader push toward context-driven learning and test-time adaptation in AI systems, and weโre building the ACE repo into an open research platform for exploring context-based methods and self-improving agents.
Working on similar ideas or curious to build on ACE? Weโd love to connect.
Excited to present MCP Explorer, an interactive way for learners to explore how AI agents connect to the world, at the NeurIPS 2025 Education Track! ๐ #NeurIPS25#NeurIPS
Heading to NeurIPS 2025 in San Diego? Check out the full list of papers from Stanford AI Lab - covering agents, diffusion models, robotics, reasoning benchmarks, and more. See you there! ๐ด
https://t.co/J7tF1NJAh9
Turn your imaginations into reality! ๐
๐ We're thrilled to announce our latest features:
โจ Image Generation โ Transform your words into extraordinary art, from cartoon characters to realistic portraits. Let your imagination lead the way!
๐ฌ Video Generation โ Create breathtaking videos from simple words. Whether itโs anime, films, or anything you dreamed of, the possibilities are endless!
As long as you can imagine, nothing is impossible! ๐ฅ
Unleash the full power of your creativity today!
๐ Try it now: https://t.co/we6PA2dkSS
#AI #AgenticAI #GenerativeAI
Congratulations to the Typeless team on the launch! The product is fantasticโI've been using it a ton. Amazing to see speech recognition finally ready for everyday use, nearly a decade after I started working on it at Baidu's AI Lab in 2016.
Welcome to the era of Vibe Automation! ๐
Sick of wasting hours on clunky automation setups and mini-programming headaches? ๐ฉ
With https://t.co/qCA6qGRRCI, you just type what you want in plain English, and boom! Your task is automated in just 1 minute! ๐
No hassle, no tech skills needed! Itโs that simple! ๐
Join our vibrant community and level up your automation game:
๐ https://t.co/GWv72VrTTE
#AI #AIAutomation #AgenticAI
Excited to share that ComputerX is the #1 Product of the Day on Product Hunt โ and #1 in AI of the Week! ๐
Huge thanks to our early users & supporters for the feedback, belief, and energy. Let's bring full desktop automation to life ๐ป๐ช
#ProductHunt#AI#Startup
Honored to be 1 of 10 startups (from 1,000+) to demo at #LlamaLounge17@Stanford!
We demoed our desktop AI agent that automates real tasksโfile system, terminal, browserโand a lightweight web app (free at ๐ https://t.co/FDcUYe6Sv8)
Huge thanks to @jowyang@chrisyeh + all the hosts & supporters!
#LlamaLounge #AI #Startups #AgenticAI #Founders #Stanford #DesktopAutomation