Really touched by this award from @ForbesPolska. This is what was said that evening: “It takes imagination, intelligence, and sheer audacity to go to Silicon Valley and tell the world’s leading AI labs they’ve got it wrong. Together with her co-founders, tonight’s winner has developed a new AI architecture - one that could prove far more efficient than the Transformer architecture behind models from OpenAI and Anthropic.” What we do at @pathway_com is difficult, especially on a human level. Big thanks to Krzysztof Domaradzki and the community in Poland and beyond for your support. It really means so much. To the Team at Pathway that makes the dragons fly. And to my co-founder and friend @JChorowski for accepting the award on my behalf (also, best dressed 😍😍😍).

“It would take someone with a lot of patience, budget, and conviction to push forward approaches that go beyond Transformers, approaches that don’t give you immediate gains on benchmarks. Those approaches are unfortunately cut out because of incentives in Machine Learning.” — Peter Norvig at the Open Source AI Summit in SF. And then @adrian_pathway took the stage to talk about the Post-Transformer Frontier :) For me, it was pretty incredible to see the three “big moments in AI” jamming together at a cosy meetup in the middle of a calm afternoon in the Presidio. Peter (shaped the field of AI) - Illia (co-invented Transformers) - Adrian (leads Post-Transformers). This event was an absolute gem. I genuinely had fun, took notes and photos of slides (!) The next generation of AI is inevitable. It is being built on efficiency, not brute force. The first generation of AI won by massively scaling the Transformer. At Pathway, we are here to change this. Thank you to @ilblackdragon, @ekang426, @build_planet, @near_ai, and @open_intel_fdn for bringing the community together!

What will it take to build AI infrastructure that can keep up with the latest frontier? I’m joining @mlech26l, @pilarmanchon and @readkrystalhu at @onewayvc’s One Way Summit to discuss what it will take to power the next leap in intelligence. 📅 Oct 29, 10 AM 📍 Commonwealth Club, SF

🚀 @zuzanna_pathway is building a $500M company creating the world’s first post-transformer #frontiermodel to bridge real-time memory with AI reasoning. See the @pathway_com co-founder & CEO on the #OneWaySummit main stage. 📅 Oct 28–29, SF 🔗 luma.com/onewaysummit2026

IPPT PAN and Pathway explore new opportunities for cooperation On September 23, 2026, IPPT PAN welcomed dr Zuzanna Stamirowska, CEO and co-founder of @pathway_com, and co-architect of the Dragon Hatchling (BDH) architecture. The visit provided an opportunity to present the Institute’s current work, particularly in artificial intelligence, robotics, and autonomous systems. Discussions focused on potential applications of the BDH architecture in research and engineering, covering topics such as model reasoning mechanisms, inference control, safety in agentic systems, and generalization theory. The Institute was represented by its Director, Prof. Janusz Szczepański, Deputy Directors: dr hab. inż. Łukasz Jankowski and dr hab. inż. Tomasz Mościcki, alongside Prof. Tadeusz Burczyński and dr hab. inż. Iwona Pokorska-Służalec. Follow-up discussions will continue at the working level between research teams from both organizations.

@adrian_pathway, @pathway_com's co-founder and Chief Scientific Officer, will join @ilblackdragon and others on stage to discuss putting the Post-Transformer era into practice at the Open Source AI Summit. 🕓Oct 22, 4:10 PM 📍 The House by Edge & Node, SF

In my new piece for @Forbes, I make the case that Transformers won’t be enough to get us to Superintelligence. They’ve taken AI extraordinarily far. The shift from models that retrieve to models that learn from experience will enable a wide range of applications that must reason reliably as information and constraints change, from cybersecurity incident response to real-time industrial operations. Our own bet is that the next leap in AI reasoning won’t come from making transformer-based models bigger. It will come from building AI that remembers, reasons and becomes more capable with use. These systems will bring structural efficiency to the picture. Transformers defined this era of AI. That doesn’t mean they get to define the next one.

The 12th Open Source AI Summit is near! I'm thrilled to host another OSAS on Oct 22 at @thehousesf with @near_ai and @open_intel_fdn during Open Source AI Week! 🪭 Speakers include the legendaries: Peter Norvig: Technical Staff at @Recursive_SI, Fellow at @Stanford @matthew_d_white: Former Global CTO of AI at Linux Foundation, CTO of @PyTorch @ilblackdragon: Founder of @NEARProtocol @near_ai, co-author of Attention Is All You Need @JeanKossaifi: Group Lead, @nvidia Research @adrian_pathway: Chief Scientific Officer, @pathway_com @cyntro_py: Builder, @cyberfund more to be added, register if you haven't, circle in Oct 22 2:30pm - 5:30pm! ❤️‍🔥 luma.com/buildp-ncun

Building @pathway_com is an ongoing journey. I had a great time sharing bits of it with founders, researchers, and investors. Thank you, @Mila_Quebec Ventures & Espace CDPQ for your continued support!

Our CSO @adrian_pathway took the stage at the AI Conference to explore the Post-Transformer landscape. With BDH, we’re uniting memory, reasoning, and adaptation. Call it intelligence-maxxing: more intelligence per watt, token, and dollar. Watch below 👇

OpenAI just announced a breakthrough in matrix multiplication - the critical building block of countless algorithms, including how all of the AI works. Like, if you were wondering what @nvidia's chips are doing - this is what they're doing. We could argue that slashing the theoretical efficiency exponent from 2.371339 to 2.25 for square matrices represents a tectonic shift in both computer science and the economics of artificial intelligence. Over lunch, my colleague at @pathway_com just extended the result to rectangular matrices, slashing the exponent to ω=2 for the other most used matrix shape (n × √n matrix by a √n × n matrix). Here is the paper (in replies). You're welcome. What beautiful times we live in, when you can get results like these while chewing on your burrito. Cheers to Przemysław Uznański (@PrzemekUzn25621)! 🍻 Human mathematicians can still be somewhat useful ;-) *The remarkable part is the equality: two is the smallest possible exponent, matching the lower bound imposed by the n^2-entry output. For this matrix shape, this result settles the exponent exactly, once and for all.

One of the AI trends I'm following is recurrent reasoning. The main idea is to let the model do the reasoning in its own continuous latent space and avoid the inefficiencies of CoT reasoning, which forces the model to reason one token at a time and in a discrete space. We're seeing different forms of it (perhaps the most popular being the looped transformer that got a lot of attention after the announcement of GPT-6 Astra). BDH-CQ by @pathway_com combines the benefits of latent reasoning with in-context learning, which was previously only seen in autoregressive transformer models. The result is an AI model that can reason on novel tasks based on the given examples while being very compute efficient. The model set a new Pareto frontier on ARG-AGI-1, reaching 29.5% at $0.0007 per task with 150M parameters. And the researchers behind the model tell me that the prospects for scaling the model and working on other modalities are very positive. Worth following.

🆕 Post-Transformer AI: Reasoning Without Language Pathway's CEO @zuzanna_pathway on why transformers aren't enough for true reasoning. Their BDH model reasons in latent space, learns from experience, and maintains state over long horizons - all at 150M parameters thedataexchange.media/zuzann…