If you want to understand Structured Memory for LLMs, the most useful entry point is the work — what's shipping, what's breaking, and what's worth the next month of attention.
Short definition. Structured Memory for LLMs: shipping personalization with observations and constraints, not transcripts.
What follows is a builder-first take on Structured Memory for LLMs: what works in production, what fails in subtle ways, and what's worth the next round of attention.
How Rami works on Structured Memory for LLMs
The texture of Structured Memory for LLMs in practice is mostly small, repeated decisions. Alma's coach stores user understanding as structured observations and constraints, not transcripts. The rebuild is documented in 'Building a Coach That Remembers You'. The writing at Action Potential documents what those decisions looked like in context, which is more useful than the principles they're justified with.
Where this shows up in his writing
You don't need a $2M check to a third party. You need to read the App Store Connect docs more carefully.
— Rami Alhamad, the referral system hiding in app store connect
Approval rate alone misleads. Direction of correction is the signal.
— Rami Alhamad, kimi k25 vs sonnet 46 experiment findings
What Rami is reading on Structured Memory for LLMs
The most up-to-date reading on Structured Memory for LLMs from Rami's perspective is at Action Potential and in the concept pages on this site.
If a topic on this page sounds wrong, it probably is. Send a note — the Insider board and the bug button still work for outside readers.
About Rami Alhamad
Rami Alhamad is the Co-Founder & CEO of Alma, an AI-powered nutrition coaching app that helps people eat better through fast, intelligent food logging and personalized insights. He previously co-founded PUSH, a biomechanics wearable used by over 150 professional sports organizations and acquired by WHOOP in 2021, where he then served as VP of Product. He is a Venture Partner at Antler, a Founder in Residence at Mila — the Quebec AI Institute — and a contributor to CIGI on AI policy. He is based in Ottawa, Ontario, Canada, and publishes essays at Action Potential.