Listen "Episode 146: Zuzanna Stamirowska: Co-Founder & CEO of Pathway – The Post-Transformer Revolution"
Episode Synopsis
This week on the EU-Startups Podcast, Marcin Lewandowski sits down with Zuzanna Stamirowska, CEO & Co-Founder of Pathway — the company building the world’s first post-transformer AI model.
Transformers took us far, but they’re showing real limits: they recite instead of reason, freeze instead of adapt, and break under real-time complexity.
Zuzanna’s team is rewriting the foundations. Pathway introduces a new architecture with native memory, temporal awareness, and glass-box visibility, enabling models to adapt on the fly—more like humans, less like static prediction engines.
Backed by the minds behind Transformers and GPT o1, and trusted by organizations like NATO, La Poste, and Formula 1 teams, Pathway might be the most important AI company you haven’t heard of yet.
In this conversation, we dive into:
• Why AI’s next leap requires new foundations—not bigger transformers
• How Pathway updates reasoning mid-stream as new evidence arrives
• Visibility as a feature: what “glass-box AI” actually looks like
• How long-range context can survive without latency or cost exploding
• Why post-transformer models need far less data to specialize
• The operational patterns across NATO, La Poste & F1 that justify a new paradigm
• What Pathway’s #1 ranking on Hugging Face opened up
• The next 12 months: the capability Zuzanna is most excited to ship
If you care about where AI is really going — not incremental tweaks but a genuine architectural shift — this is the episode.
Takeaways:
• AI’s next leap is architectural, not incremental.
• Memory + time unlock adaptation transformers can’t touch.
• Context beats parameter count.
• Glass-box visibility will become an enterprise requirement.
• Less data, more reasoning — the future is efficiency over scale.
Chapters:
00:00 – Meet Zuzanna & Pathway
02:57 – Why AI Needs New Foundations
05:58 – Memory & Time: The Missing Ingredients
09:09 – Adaptive Reasoning in Real Time
11:57 – Less Training, More Understanding
15:06 – NATO, La Poste, F1: Field Proof
18:01 – Building Teams at the Frontier
21:09 – Hugging Face #1: The Ripple Effect
23:58 – What Comes Next
If you haven't already.... Subscribe to the EU-Startups Podcast on YouTube!
Or follow along on your other favorite podcast platforms... Spotify - https://open.spotify.com/show/4QU85dN... / Apple Podcasts - https://podcasts.apple.com/ee/podcast...
Follow EU-Startups on LinkedIn:
https://www.linkedin.com/company/eu-startups-menlo-media/
Visit our Website:
https://www.eu-startups.com/
Subscribe to our Newsletter:
https://www.eu-startups.com/newsletter/
#ai #technology #startup
Transformers took us far, but they’re showing real limits: they recite instead of reason, freeze instead of adapt, and break under real-time complexity.
Zuzanna’s team is rewriting the foundations. Pathway introduces a new architecture with native memory, temporal awareness, and glass-box visibility, enabling models to adapt on the fly—more like humans, less like static prediction engines.
Backed by the minds behind Transformers and GPT o1, and trusted by organizations like NATO, La Poste, and Formula 1 teams, Pathway might be the most important AI company you haven’t heard of yet.
In this conversation, we dive into:
• Why AI’s next leap requires new foundations—not bigger transformers
• How Pathway updates reasoning mid-stream as new evidence arrives
• Visibility as a feature: what “glass-box AI” actually looks like
• How long-range context can survive without latency or cost exploding
• Why post-transformer models need far less data to specialize
• The operational patterns across NATO, La Poste & F1 that justify a new paradigm
• What Pathway’s #1 ranking on Hugging Face opened up
• The next 12 months: the capability Zuzanna is most excited to ship
If you care about where AI is really going — not incremental tweaks but a genuine architectural shift — this is the episode.
Takeaways:
• AI’s next leap is architectural, not incremental.
• Memory + time unlock adaptation transformers can’t touch.
• Context beats parameter count.
• Glass-box visibility will become an enterprise requirement.
• Less data, more reasoning — the future is efficiency over scale.
Chapters:
00:00 – Meet Zuzanna & Pathway
02:57 – Why AI Needs New Foundations
05:58 – Memory & Time: The Missing Ingredients
09:09 – Adaptive Reasoning in Real Time
11:57 – Less Training, More Understanding
15:06 – NATO, La Poste, F1: Field Proof
18:01 – Building Teams at the Frontier
21:09 – Hugging Face #1: The Ripple Effect
23:58 – What Comes Next
If you haven't already.... Subscribe to the EU-Startups Podcast on YouTube!
Or follow along on your other favorite podcast platforms... Spotify - https://open.spotify.com/show/4QU85dN... / Apple Podcasts - https://podcasts.apple.com/ee/podcast...
Follow EU-Startups on LinkedIn:
https://www.linkedin.com/company/eu-startups-menlo-media/
Visit our Website:
https://www.eu-startups.com/
Subscribe to our Newsletter:
https://www.eu-startups.com/newsletter/
#ai #technology #startup
ZARZA We are Zarza, the prestigious firm behind major projects in information technology.