Writing

Notes on memory, continual learning, and what we find while building Stash. From the team at Fergana Labs.

The context gold rush: why everyone is building the same thing

Context graph, company brain, LLM wiki — a map of who is building the context layer, the patterns that have converged, and what is still missing before any of it reaches mass adoption.

Sam Liu Containerizing memory: the real barrier to continual learning and enterprise AI adoption

The shipping container standardized freight and the world economy followed. Agent memory has no such standard yet, and that missing consistency is why enterprise adoption has been mixed.

Sam Liu Open Questions in Memory, and Our Predictions

The questions we argue about most with others building memory — labs vs startups, weight vs token space, retrieval vs blast radius, benchmarks — and where we think each one lands.

Sam Liu Giving yourself superpowers: Advice on building a simple company brain

An opinionated take on the right way to build a company brain — integrations, retrieval, memory, and privacy — so your AI agents can do real knowledge work.

Henry Dowling Why hasn't there been any great consumer AI (still)

When models stop getting smarter, context engineering becomes the battleground — and the case for an inevitable AI memory infrastructure buildout.

Henry Dowling Three Dimensions That Matter To An Agent Memory Store

An opinionated take on three key decisions memory builders need to make: retrieval, structure, and knowledge graphs.

Henry Dowling Agents are Octopuses

Collaborative agent systems with shared memory as a new paradigm beyond swarms and assembly lines.

Sam Liu I Dropped Out of My PhD

Choosing the start-up life over the academic path.

Sam Liu Why Context Windows Won't Save Us

Why raw context length is not a substitute for true memory in AI.

Sam Liu In Praise of Mess

Embracing creative disorder as a feature, not a bug.