AI Agents & Automation
Autonomous AI agents, tool use, planning, multi-agent systems, workflow automation
The next platform shift — from chatbots to autonomous digital workers
The AI Agents & Automation sector is experiencing a transformative shift right now, with advancements in multimodal learning and planning enabling the transition from basic chatbots to autonomous digital workers capable of simulating human-like behavior. High-profile developments, such as Percy Liang's 2023 paper 'Generative Agents: Interactive Simulacra of Human Behavior,' highlight this momentum, as investor interest surges amid shortening AGI timelines. This evolution is driven by real-world applications in manufacturing and logistics, where efficiency gains are evident, but ethical and safety concerns are escalating, as warned by experts like Bernhard Schölkopf from the Max Planck Institute. Among the hottest sub-topics, Generative Agents for Simulations leads the way, with researchers Percy Liang and Yejin Choi demonstrating how these agents can create interactive simulations for decision-making, as seen in Liang's work. This sub-topic is 'hot' due to its potential to handle complex scenarios, offering efficiency in virtual environments. Closely following is Multimodal AI for Robotics Integration, driven by Jensen Huang of NVIDIA and Guillaume Lample of Mistral AI, which combines visual, textual, and sensory data for autonomous operations in physical settings, as detailed in Trevor Darrell's 2024 paper on compositional chain-of-thought prompting. This area is also 'hot,' with applications in autonomous vehicles and manufacturing poised to revolutionize industries. AI Safety in Autonomous Systems, while 'warm,' involves protocols to mitigate risks, led by Bernhard Schölkopf and Russ Tedrake, addressing unintended behaviors in multi-agent environments. A key debate centers on whether AI development should prioritize speed or safety in autonomous systems. On one side, advocates like Guillaume Lample of Mistral AI and Emad Mostaque argue that rapid progress, as evidenced by recent robotics experiments, will spur innovation and economic benefits, outweighing risks with oversight. Conversely, critics such as Bernhard Schölkopf from the Max Planck Institute and Russ Tedrake from MIT contend that hastening deployment could lead to societal harms, citing warnings about AGI timelines and the need for robust safeguards, as reflected in Yoshua Bengio's 2024 paper on regulating advanced agents. This disagreement underscores the tension between innovation and risk management in the sector. For investors, the AI Agents & Automation sector presents opportunities in the growing demand for autonomous systems, particularly from companies like NVIDIA and Mistral AI, with potential high returns as prototypes move toward commercialization in the next 1-2 years. However, risks include regulatory hurdles and safety failures, as highlighted by expert warnings, which could delay adoption and increase volatility. Investors should watch sub-topics like generative agents and multimodal robotics for diversification, while monitoring debates on scaling sustainability to assess what's at stake—namely, the balance between transformative gains and ethical pitfalls that could reshape market dynamics.
Key Voices in AI Agents & Automation
Bret Taylor
OpenAI Board
6 posts
Graham Neubig
Carnegie Mellon University
4 posts
Guillermo Rauch
Vercel
3 posts
Mustafa Suleyman
Microsoft AI
2 posts
Jie Tang
Tsinghua University
2 posts
Emad Mostaque
Stability AI
2 posts
Kevin Roose
The New York Times
2 posts
Werner Vogels
Amazon
1 posts
Aravind Srinivas
Perplexity AI
1 posts
Adam Selipsky
AWS
1 posts
Fei-Fei Li
Stanford University / Stanford HAI
1 posts
Kate Crawford
USC Annenberg / AI Now
1 posts
Some nice updates here. The most surprising one was that analyzing our trajectories resulted in finding a vulnerability in one of the benchmarks we were using... Coding agent benchmark creators, please mind your git history 😅
Our Agent Data Protocol dataset now has 3.2M instances, double its original size! Also the paper was accepted as an ICLR oral presentation! Multimodal support coming soon as well. https://t.co/lUVVFZNCan
Really excited for this new line of work from GEPA! We found that you can auto-generate anything from coding agent skills to 3D images with a LLM-guided optimization algorithm, and outperform past methods.
▲ + 📽️ ▪️Video is now supported on @vercel AI Gateway ▪️Grok Imagine Video & Image are 🆓 thru Feb 25 ▪️New 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎𝚅𝚒𝚍𝚎𝚘 @aisdk API Incredible apps and agents are waiting to be shipped!
AI at its best. We Ralph Wiggum'd a better WebStream implementation optimized for server-side Node.js environments. Up to 14.6x performance improvement, with 1100/1116 WPT tests passing. Autonomously. We're working to upstream this work to Node.js for the benefit of all. https://t.co/tJsnhB9Evu
We just uploaded our GLM-5's tech report onto arxiv. Hope it helpful! takeaway keywords: - dsa arch with 750B parameter (40B active) + 30T data - slime RL toolkit - asynchronous agentic RL that is really helpful (+3 points improvements on several major benchmarks) - adapt to
When do you think ai agent economic activity will overtake human economic activity?
My initial take on @Grok 4.20 is that it's very.. pleasant? Fast and accurate responses, handles some advanced stuff very well, nice balance of attitude. I feel it has more horsepower than currently shown though, perhaps that scales with more than 4 agents?
A little over three years ago, @byroncook and I had a chat about automated reasoning. Since then, the landscape has shifted faster than any of us anticipated. AI agents are in production, generating code, making decisions, and we need ways to prove their correctness. I sat down
Excited to join Tian Chong Ng from Singtel at Mobile World Congress to discuss how Sierra has partnered with Singtel to build AI agents that drive real business growth https://t.co/GbibZqVLCa
Peter Steinberger is joining OpenAI to drive the next generation of personal agents. He is a genius with a lot of amazing ideas about the future of very smart agents interacting with each other to do very useful things for people. We expect this will quickly become core to our
From now on, hype-centric splashy launches will likely be strongly uncorrelated with success. If by the time you launch you don’t have escape velocity, you will likely get Sybil attacked¹. Agents will spin up 10 competing products with your same interface. Start with an
Proud that Sierra has partnered with Cedar to build Kora, an AI voice agent purpose-built for patient billing. Catch Cedar's Sumayah Rahman and Sierra's Miranda Zhao at ViVE later this month to hear how they built Kora and what they've learned from real patient interactions https://t.co/GWDZxwULqi
Memory now works with Model Council. Have fun having multiple frontier reasoning models think about your data together and work for you! https://t.co/wtw6iK2HsD
AI is your ghostwriter, but you are the author. I was speaking with my friend Arya about the complex dynamics software teams face now that most code is being generated by AI agents. There are very few natural bottlenecks to code “slop” - features over quality, and functionality
We make this point in AI as Normal Technology. And while I agree that there's a lot we've learned since the Industrial Revolution, when we look at more recent, smaller scale tech shocks like the internet and social media, I don't think the institutional and policy reaction was
It is very impressive that Opus 4.6 built a working C compiler in 2 weeks, using agent teams and burning $20k worth of tokens. It's also impressive that @rui314 built one by himself in 7-or-so days worth of commits 😀 https://t.co/7kSdAVjWeO
this is why everyone was freaking out about claude code over winter break! once you see an agent autonomously doing stuff for you, it's so instantly clear that ~all computer-based work will be done this way. (this is why my Serious AI Policy Proposal is to sit every member of
Sierra partnered with @Curative to build their AI voice agent that's creating faster, better experiences for members and providers while freeing up their team to focus on what matters most: building deeper relationships https://t.co/QmeFqQi78T
This is great, and the bits about inference demand ring true. It's trivially easy to generate 100ks of tokens in a single session with a coding agent (sometimes a single prompt!) and I'm not even using this stuff professionally. We're gonna need a lot more data centers.
RunBuggy operates one of the largest automotive transportation marketplaces in North America, coordinating thousands of live vehicle shipments at any given moment. With Sierra, they built an AI agent to place outbound calls that proactively collect info and drive efficiencies
Looks like Kimi-K2.5 is by @Kimi_Moonshot is the new best open LLM for agentic software engineering!
Build with World API! Marble now can help more users to empower their creation, products and workflows! 🚀🤩
We’re excited to partner with Stellarus to help health plans like Blue Shield of California build their AI agent https://t.co/HUHlmWwbgG
The next big milestone I'm watching for on our way to AGI: Artificial Capable Intelligence (ACI). Can an agent take $100k and legally turn it into $1M? To me that's the modern Turing Test.
The team just added a little extra holiday spirit to @Copilot! Meet Eggnog Mode Mico - live now in the US, UK, and Canada, only available for the holidays. Toggle on by just clicking the ⛄ icon while talking to Mico. https://t.co/zDQM3548BA
New multi-year, strategic partnership with @OpenAI will provide our industry-leading infrastructure for them to run and scale ChatGPT inference, training, and agentic AI workloads. Allows OpenAI to leverage our unusual experience running large-scale AI infrastructure securely, https://t.co/HZGeld5M9q
Reasoning with latency-optimized models is quite a UX game changer. Super proud of what the team has accomplished with this Magistral release! https://t.co/qRLMHrW4cV https://t.co/WvKnum7efd
Researchgate sent me a fake paper called "The AI Health Revolution: Personalizing Care through Intelligent Case-based Reasoning" which claims to be by me and Yann LeCun. More than one third of the citations are to Shefiu Yusuf which may mean nothing.