The next Transformer moment for AI - read in Forbes

Our groundbreaking post-transformer BDH architecture has launched

October 15, 20254 min read

"Pathway 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." — read in Forbes

Hello,

At a time when we face a shortage of energy to power AI, where data is running short and companies are investing billions in incremental model enhancements, the world needs a paradigm shift.

For well over a year, Pathway has been building in stealth with the unwavering support of our investors. Glimpses of our work were publicly visible, such as our "Future of Deep Learning" event in San Francisco with Lukasz Kaiser, and our increased focus on LiveAI™. The keen-eyed observer could even spot the synapses in our logo and realize that the name Pathway is a direct reference to neural pathways. From the outset, we have been making machines think and adapt in real time, just like humans. Today, I'm excited to announce the launch of our AI breakthrough.

With warm regards,
Zuzanna

BDH architecture: enter the new era in AI

The BDH research paper on arXiv

On October 1, Pathway launched BDH, a groundbreaking post-transformer architecture that reimagines AI as a dynamic, brain-like system. BDH is capable of continuous learning and adaptation without the constraints of traditional Transformer-models.

Built around a core neuroscience principle called Hebbian learning, BDH proposes a future where AI performance is decoupled from exponential increases in energy and cost. In a world where computational needs are insatiable, BDH is a more sustainable and commercially viable path forward.

Why it matters: Unlike traditional Transformer-based models, BDH learns continuously — with little data, no massive retraining cycles, no exploding compute costs. It offers efficiency by design.

"The implications (of BDH) could be both technical and economic. A system that learns continuously could make AI development cheaper, faster, and more sustainable" — Victor Dey, Forbes

Overwhelming market response

As the world seeks to unlock next-generation intelligence, BDH has seen explosive interest.

By the numbers:

  • Reaching Forbes and Semafor
  • Delivering a full tech deep dive in John Krohn's Super Data Science Podcast
  • Organically sparking widespread discussions across developer and social communities
  • #1 paper of the day on Hugging Face within hours, trending ever since with 450+ upvotes
  • 3k stars on GitHub for our #BDH Repo in 1 week
  • Amplifying quickly across top AI voices, like Rohan Paul (40k views), Daily Papers (60k views), The Turing Post (90k subscribers) and McKinsey's Sohrab Rahimi (20k followers), with team video interviews racking up 200k views within a week

BDH isn't just a research breakthrough — it's a movement rapidly gaining traction across the AI ecosystem.

Dragon Hatchling on the Super Data Science Podcast

BDH trending as #1 paper of the day on Hugging Face

Technology deep dive: beyond the Transformer

The paradigm shift in AI can be explained by a distinctly new architecture that scales brilliantly on GPU. The patent has more than 100 claims.

Inside the BDH architecture

Next steps: the first enterprise-native LLM intelligence

The BDH innovation opens the door to AI that's not just more efficient, but smarter, more adaptable, and better suited for large-scale deployment with dynamic data. Today, we are laser-focused on putting a model based on BDH in the hands of eager enterprise customers. Strategic partners: AWS and Nvidia have rolled up their sleeves to help Pathway accelerate.

Get involved

If you like what we have created, we would love your support socializing BDH with top AI brains in your network. We will begin to offer select design partnerships in Q4 2025, and are eager to connect with interested parties.

If you are active in this space, or curious about post-Transformer, enterprise-native LLMs, we'd love to engage. You can book a meeting, or stop by to visit us in Palo Alto!