Engineering Blog

Notes from building event-first agents in a distributed AI platform

A technical blog about the parts that usually stay hidden: timeline-first context, uncached operational state, multi-channel generation, sparse workspaces, isolated execution, and why some apparently simpler choices stop fitting once the system becomes real.

Series focus React v2 internals, workspace engineering, and agent-runtime design

Current lead article Memory is not one bucket: how React remembers

Palette reference Using the website color schema

Featured article

The newest post explains the full memory picture behind React: timeline memory, attention area, turn-log reconstruction, workspace continuity, logical namespaces, and why visible context is only one layer of remembering.

Short answer: React does not remember through one bucket. It remembers through cooperating surfaces: timeline, SOURCES POOL, ANNOUNCE, turn log, workspace, hosted artifacts, and searchable index records, all tied together by logical paths.

Timeline memory

Durable event memory lives in the timeline, while the model consumes only a shaped visible slice of that history.

Attention memory

SOURCES POOL and ANNOUNCE form the always-on-top attention area where React expects to find fresh sources and must-see signals.

Path-based reopening

Logical paths let the runtime compress, hide, summarize, and relocate memory without losing the route back to it.

Read the newest post Read the cache post

Editorial direction

The point of these posts is not to advertise complexity for its own sake. The point is to explain why the system ended up with some apparently unusual choices once the requirements became: multi-user, multi-tenant, streaming, auditable, isolated, and cluster-friendly.

The blog should feel technical, opinionated, and grounded in the actual runtime rather than generic “AI architecture” slogans. The first two posts set the tone: one on event-first orchestration, one on the fixed attention board, and one on how the runtime shapes the cache instead of leaving it to chance.