Subvolts take Pathway's BDH architecture offers a potential post-transformer future with built-in memory for continuous learning. Mar 11, 2026 - Developers
- Learn about a novel AI architecture that integrates memory and continuous learning, potentially offering new approaches to model design.
- Business
- Understand emerging AI research that could lead to more reliable and adaptable AI systems for real-world applications.
This interview explores a potentially groundbreaking AI architecture that could shift the field beyond current transformer models.
#326 Zuzanna Stamirowska: Inside Pathway's Post-Transformer Architecture Designed for Memory and On-the-Fly Learning
Audio streams from Craig S. Smith. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
Eye on AI features Zuzanna Stamirowska discussing Pathway's Baby Dragon Hatchling (BDH), a post-transformer architecture. BDH aims to overcome limitations in current LLMs regarding long-horizon reasoning by integrating memory and synaptic plasticity. The conversation explores its potential for continual learning, reducing hallucinations, and improving generalization over time, with implications for various applications.
This episode discusses a new AI design called Baby Dragon Hatchling. Unlike current AI that forgets easily, this design builds in memory. This could help AI learn continuously and avoid making up information.
This episode offers a deep dive into a novel AI architecture that challenges the dominance of transformers. It's for researchers and developers interested in next-generation models, particularly those focused on memory and continuous learning. Pathway's work could influence how major labs like Google Gemini and Anthropic approach future model development, especially for tasks requiring sustained context and adaptation.
- HostCraig S. Smith
- GuestZuzanna StamirowskaCEO, Pathway
- Guest: Zuzanna Stamirowska, Pathway CEO
- Episode Title: Inside Pathway's Post-Transformer Architecture Designed for Memory and On-the-Fly Learning
- Show: Eye on AI
- Length: 68 minutes
- Architecture Name: Baby Dragon Hatchling (BDH)
- Key Features: Memory, synaptic plasticity, continual learning, generalization over time
This episode is sponsored by tastytrade.
Trade stocks, options, futures, and crypto in one platform with low commissions and zero commission on stocks and crypto. Built for traders who think in probabilities, tastytrade offers advanced analytics, risk tools, and an AI-powered Search feature.
Learn more at https://tastytrade.com/
This episode dives into why Pathway's Baby Dragon Hatchling (BDH) might mark the beginning of the post-transformer era in AI. Zuzanna Stamirowska, Pathway's CEO and co‑author of BDH, explains why today's transformer-based LLMs hit a wall on long-horizon reasoning, how memory and synaptic plasticity are built directly into BDH's architecture, and what that means for continual learning, hallucinations, and "generalization over time." The conversation ranges from complexity science and brain-inspired computation to practical implications for real-world, small-data, and safety‑critical applications.
Stay Updated:
Craig Smith on X: https://x.com/craigss
Eye on A.I. on X: https://x.com/EyeOn_AI
(00:00) The Core Problem: Why Today's AI Lacks Memory
(03:16) Pathway's Mission to Bring Memory Into AI
(04:53) Zuzanna's Background in Complexity Science
(10:30) Why Transformers Reset Like "Groundhog Day"
(14:34) The Brain-Inspired Dragon Hatchling Architecture
(23:59) How the Network Learns and Builds Connections
(37:38) Performance vs Transformers on Language Tasks
(49:37) Productizing the Technology With NVIDIA and AWS
(54:23) Can Memory Solve AI Hallucinations?
- Transformer
- AI researcher
- A common type of AI model architecture that excels at processing sequential data like text, but can struggle with long-term memory.
- Long-horizon reasoning
- AI researcher
- The ability of an AI to understand and connect information over very long sequences or time periods.
- Synaptic plasticity
- Neuroscientist
- The ability of connections between brain cells to strengthen or weaken over time, crucial for learning and memory.
- Hallucinations
- AI researcher
- When an AI generates false or nonsensical information that it presents as fact.
55 min42 min51 min34 minUkraine's Secret Weapon: The Points System Winning the Drone War | Andrii Hrytseniuk
WithAndrii Hrytseniuk
1 h 05Are We Screwed If AI Works? — With Andrew Ross Sorkin
WithAndrew Ross Sorkin
24 min
44 minGreetings, Earthlings: Philip Johnston of Starcloud on Data Centers in Space
WithPhilip Johnston
Questions people ask
- Why are current transformer-based LLMs limited?
- Today's transformer-based LLMs hit a wall on long-horizon reasoning, essentially resetting like 'Groundhog Day' without retaining memory.
- What makes Pathway's Baby Dragon Hatchling different?
- BDH has memory and synaptic plasticity built directly into its architecture, enabling continual learning and better generalization over time.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Craig S. Smith.