Pathway (www.pathway.com)

@signulll Memory is key. And it is fascinating like nothing else. I would challenge if it's still art and the "no one" ;-) We did. In BDH we have coupled memory directly with the attention mechanism (through synaptic plasticity). arxiv.org/pdf/2509.26507 [btw. memory =/= text file XD]

Transformers have no memory. Every inference is Groundhog Day. @zuzanna_pathway breaks down why we built BDH to fix this on @theneurondaily. 📷 youtu.be/duw7RUif8hE

How and what tech will change our lives in 2026? The Wall Street Journal, as always, has one of the best-informed perspectives. We are excited to see Pathway and the post-transformer era being part of their predictions! wsj.com/tech/ai/tech-p…

We recently had the pleasure of speaking with Jan Chorowski, a former Mila student who is now the co-founder and CTO of Pathway, an AI-powered data processing platform. He shares his journey from academic research to entrepreneurship, reflects on his time at Mila, and talks about his vision for the future of artificial intelligence. An inspiring conversation you won’t want to miss! mila.quebec/en/news/spotli… #milaalumni

🧠 The story also broke the news on our integration with @Nvidia’s AI infrastructure and @AWS cloud and AI tech stack. We’ll be at #AWSreInvent all week if you want to meet and learn more!

Transformers got us here, but what comes next? The @WSJ & @Steve_Rosenbush covered how Pathway’s Dragon Hatchling architecture rethinks memory, reasoning, and what’s possible beyond today’s AI. wsj.com/articles/an-ai…

“Memory is key to intelligence and efficient reasoning.” @Steve_Rosenbush at The @WSJ covered how @pathway_com is rethinking AI from the ground up and our newly announced integration with @Nvidia and @AWS - not just scaling models, but evolving intelligence itself. Dragon Hatchling (BDH) represents the beginning of the Post-Transformer Era, read the full piece: wsj.com/articles/an-ai…

Check @pathway_com Baby Dragon Hatchling (BDH), that paves the way for autonomous AI.

Fun fact about #bdh_architecture - it allows you to build larger models from small ones. Like lego blocks. Thanks for highlighting this @JonKrohnLearns! youtu.be/PtJh5xqSmwU?si…

avg sidewalk in downtown palo alto 🌞

Our breakthrough post-Transformer architecture (#BDH) shows that AI can now self-evolve and reorganize itself, inspired by how our biological neurons learn and adapt. It’s a glimpse into what the next generation of intelligence should look like.

As @Forbes's @iamVictorDey writes, Pathway's BDH “may have sparked the beginning of a new era in AI — one where machines don’t just imitate the brain, but begin to think like it.” forbes.com/sites/victorde…

"The implications could be both technical and economic. Retraining large models costs companies billions each year in computing power and energy. A system that learns continuously could make AI development cheaper, faster, and more sustainable."

@adrian_pathway is exploring with @JonKrohnLearns Pathway’s new post-transformer architecture #BDH that aims to bridge the gap between transformer models and how the brain actually works. youtu.be/mfV44-mtg7c?si…

Breaking news: Adrian Kosowski speaks to @JonKrohnLearns about #machinelearning’s entry into neurology research via a biologically inspired architecture. Watch the episode here: bit.ly/sdsp929

A missing link between Transformers and the brain? 🧠 Dragon Hatchling (BDH) is a new LLM architecture based on a scale-free, biologically-inspired network of locally-interacting neuron particles. It rivals GPT2 performance, but is designed for interpretability.

@rohanpaul_ai @rohanpaul_ai side-by-side comparison of BHD and Transformer architectures.

Wow. 🧠 The paper presents Dragon Hatchling, a brain-inspired language model that matches Transformers using local neuron rules for reasoning and memory. It links brain like local rules to Transformer level performance at 10M to 1B scale. It makes internals easier to inspect because memory sits on specific neuron pairs and activations are sparse and often monosemantic. You get reliable long reasoning and clearer debugging, because the model exposes which links carry which concepts in context. The problem it tackles is long reasoning, models often fail when the task runs longer than training. The model is a graph of many simple neurons with excitatory and inhibitory circuits and simple thresholds. It stores short term memory in the strengths of connections using Hebbian learning, so links that fire together get stronger. Those local rules behave like attention but at the level of single neurons and their links. They also present BDH-GPU, a GPU friendly version that implements the same behavior as an attention model with a running state. This design produces positive sparse activations, so only a small fraction of neurons fire at a time. Many units become monosemantic, meaning one unit often maps to one clear concept even in models under 100M. The theory ties the approach to distributed computing and explains how longer reasoning can scale with model size and time. ---- Paper – arxiv. org/abs/2509.26507 Paper Title: "The Dragon Hatchling: The Missing Link between the Transformer and Models of the Brain"

Pathway just published Dragon Hatchling (BDH), and I’ve been digging into the paper. It’s a new LLM architecture where continuous learning is native—via synaptic plasticity and Hebbian updates—rather than bolted on with scaffolds. Technical breakdown ⤵️

Continuous learning doesn’t have to live inside the LLM; today it’s a system-level hack (LLMs + memory + tools). But the bitter lesson says a data‑scaled meta‑learner that learns to learn will beat our hand‑engineered scaffolds.