Antonio Montano ☼
@AntoMon
Followers
716
Following
649
Media
530
Statuses
44K
Digital transformer
Milan
Joined March 2009
From S&OP to Continuous Orchestration: Envisioning the Hybrid Enterprise of Humans and AI Agents – From calendar-driven cycles to signal-driven enterprise adaptation #enterprisearchitecture
https://t.co/hL0jR70cSJ
4m4.it
This essay explores the evolution of enterprise planning from traditional Sales and Operations Planning (S&OP) toward a paradigm of continuous orchestration, where humans and AI agents act as...
1
1
0
GEPA can augment even RL-tuned models! @IntelligenceArc team uses GEPA in their ATLAS framework, to augment an already powerful and RL-tuned teacher model, achieving +142% student performance improvement when guided by the improved teacher!
3
8
47
🚨 We just proved that in medical AI, quality beats quantity. Our 2M image dataset outperforms 15M+ datasets. Introducing Open-PMC🧵
3
7
67
🧵 As AI labs race to scale RL, one question matters: when should you stop pre-training and start RL? We trained 5 Qwen models (0.6B→14B) with RL on GSM8K and found something wild: Small models see EMERGENCE-LIKE jumps. Large models see diminishing returns. The scaling law?
37
99
694
This a great experiment! It's not quite a RL scaling law, in the sense of what I've collected from a few frontier labs. Lots of things are similar, but their methods for establishing these relationships seems a bit different. First, they definitely use a set of base models sort
🧵 As AI labs race to scale RL, one question matters: when should you stop pre-training and start RL? We trained 5 Qwen models (0.6B→14B) with RL on GSM8K and found something wild: Small models see EMERGENCE-LIKE jumps. Large models see diminishing returns. The scaling law?
6
13
219
New APPLE paper says a small base model plus fetched memories can act like a bigger one. With about 10% extra fetched parameters, a 160M model matches models over 2x its size. Packing all facts into fixed weights wastes memory and compute because each query needs very little.
11
33
279
Our recent CCDD paper on discrete language modeling is out: 📚Coevolutionary Continuous Discrete Diffusion: Make Your Diffusion Language Model a Latent Reasoner https://t.co/ChoCrMuIs3
arxiv.org
Diffusion language models, especially masked discrete diffusion models, have achieved great success recently. While there are some theoretical and primary empirical results showing the advantages...
1
22
58
Market research firms are cooked 😳 PyMC Labs + Colgate just published something wild. They got GPT-4o and Gemini to predict purchase intent at 90% reliability compared to actual human surveys. Zero focus groups. No survey panels. Just prompting. The method is called Semantic
46
230
1K
[AI] LLMs Reproduce Human Purchase Intent via Semantic Similarity Elicitation of Likert Ratings B F. Maier, U Aslak, L Fiaschi, N Rismal... [PyMC Labs] (2025) https://t.co/mtIRoEVnYD
0
2
6
AMD's SVP of AI Vamsi Boppana says the company's AI software, designed with input from OpenAI, helped secure the multi-billion dollar deal with OpenAI (@richardjnieva / Forbes) https://t.co/8ooxh7YFtU
https://t.co/QuWG63mPFN 📫 Subscribe:
techmeme.com
Richard Nieva / Forbes: AMD's SVP of AI Vamsi Boppana says the company's AI software, designed with input from OpenAI, helped secure the multi-billion dollar deal with OpenAI
0
1
6
• Each humanoid robot requires 0.9 kg of rare-earth metals, comparable to that of an electric vehicle, including 28 rotary and linear actuators for joints and 12 actuators for the hands. • The humanoid robot revolution could trigger $800 billion worth of demand for key
54
207
864
🚨 Paper Alert 🚨 ➡️Paper Title: Human3R: Everyone Everywhere All at Once 🌟Few pointers from the paper 🎯Authors of this paper presented “Human3R”, a unified, feed-forward framework for online 4D human-scene reconstruction, in the world frame, from casually captured monocular
0
2
34
AI-Driven Power Demand Surge in the U.S. Brings Gas Turbines to the Forefront: How GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries Are Responding The U.S. “power crunch” triggered by AI data centers has thrust the traditional industrial equipment — medium and large
16
78
394
My first tutorial on creative task optimization with DSPy and GEPA is online! Not everything went as expected, but it’s a solid start toward better results. https://t.co/3UaaUCCcCB
Finished my first tutorial on improving the quality of smaller LLMs for creative tasks using a teacher + GEPA. I'll publish it on Monday. What style of signature do you prefer? I usually go with classes, but for this tutorial I decided to use inline
3
32
205
Mark Zuckerberg's metaverse chief is urging employees to adopt AI across every workflow as part of a broader shift inside the company.
wired.com
Mark Zuckerberg's metaverse chief is urging employees to adopt AI across every workflow as part of a broader shift inside the company.
0
2
1
New paper we're excited to get online! Taming Imperfect Process Verifiers: A Sampling Perspective on Backtracking. A totally new framework based on ~backtracking~ for using process verifiers to guide inference, w/ connections to approximate counting/sampling in theoretical CS.
6
25
151
true. @orf_bnw was the OG attention is all you need for everything guy. https://t.co/gOEo0esozY
@kchonyc hahaha you got me. i don't remember when we took that photo, you? Linda? man life is weird- i wish i was known for “attention is universal and can be shared across tasks” but oh well, “orhan fix it” is what got me more followers :-P
0
4
41
OmniRetarget: Interaction-Preserving Data Generation for Humanoid Whole-Body Loco-Manipulation and Scene Interaction Project: https://t.co/X8p5jesI8X Paper: https://t.co/qkaGwJjWSY New Paper from Amazon FAR proves that by leveraging high quality data from a
0
18
77
Analysis: in 2025, tech companies have raised about $157B in the US bond markets, up 70% from last year, as debt seeps into every corner of the AI economy (@edludlow / Bloomberg) https://t.co/iMhDgkpl5D
https://t.co/Zqnlj8kxHC 📫 Subscribe:
techmeme.com
Edward Ludlow / Bloomberg: Analysis: in 2025, tech companies have raised about $157B in the US bond markets, up 70% from last year, as debt seeps into every corner of the AI economy
1
4
8
Great recap of security risks associated with LLM-based agents. The literature keeps growing, but these are key papers worth reading. Analysis of 150+ papers finds that there is a shift from monolithic to planner-executor and multi-agent architectures. Multi-agent security is
11
60
249
Your LVLM says: “There’s a cat on the table.” But… there’s no cat in the image. Not even a whisker. This is object hallucination — one of the most persistent reliability failures in multi-modal language models. Our new #NeurIPS2025 paper introduces GLSim, a simple but
2
28
124