Our first initiative is AI Memory.
Every agent that reaches production hits the same question: what it may remember, and who answers for it.
That layer is being set now.
20
Operating companies at one table
8 / 8
Layers of the memory stack covered
Sept 24
First working session, San Francisco
Eight layers, all covered.
01
Programmable compute
The silicon underneath, and how far it bends to the workload.
02
Systems memory
The memory wall. A physics problem before a product problem.
03
Foundation models
The models, and what they hold in context at inference.
04
Graph and context
Structure over retrieval. Retrieval without structure is search.
05
Agent memory
What an agent may remember, and who owns it.
06
Harness and orchestration
The runtime that calls tools and carries state.
07
Enterprise deployment
Into production, inside real systems of record.
08
Governance and trust
What was retained, where it lives, who is accountable.
Most of the debate lives in one layer. The other seven decide whether any of it holds in production — across retrieval, context, privacy, security, evaluation, enterprise knowledge, correction, and deletion.
The Technical Council. 20 companies, one layer.
Practitioners who run these systems in production — not observers. Four of them facilitate the working session.
FacilitatorPhilip Rathle
Chief Technology Officer
Neo4j
Zhamak Dehghani
Founder and CEO
Nextdata
Charles Fan
Co-Founder and CEO
MemVerge
Taranjeet Singh
Co-Founder and CEO
Mem0
FacilitatorArun Rao
Machine Learning
NVIDIA
Manuel Meyer
Co-Founder and COO
Qdrant
Roxane Fischer
Co-Founder and CEO
Anyshift
Ganesh Padmanabhan
Co-Founder and CEO
Autonomize AI
Shail Khiyara
CEO
Swarm Engineering
Jiquan Ngiam
Co-Founder and CEO
MintMCP
FacilitatorVasilije Markovic
Founder and CEO
Cognee
Michael J.J. Tiffany
Co-Founder and CEO
Fulcra Dynamics
Ben Kearns
Co-Founder and CEO
Ferrosa AI
Shawkat Kabbara
Founder and CEO
Papr
FacilitatorAli Töre
SVP and GM, Conversational AI
RingCentral
Amandeep Khurana
Enterprise Agentic Harness
Anthropic
Suresh Pillai
VP of Engineering
TriNet
Daniel Chalef
Founder
Zep AI
Navrina Singh
Founder and CEO
Credo AI
One seat. Six returns.
01
Set the blueprint.
Help write the industry blueprint for the layer you run.
02
Skip the rework.
Your production constraints built in before the approach is set.
03
Ship first.
Early access to the reference architecture and benchmark.
04
Integrate once.
One shared interchange, not custom glue for every partner.
05
Prove it.
Benchmark results and a public record that buyers can trust.
06
Stay ahead.
Our outputs feed ISO, IEEE and NIST before the rules are set.
The rooms we have built before.
