Insights
Articles, research, and whitepapers from the practitioners who lead our engagements.
8 insights
Aug 24, 2026 · 25 min read
We built re-auth, our own multi-tenant identity system. That sentence used to be a confession. It isn't anymore, and this is what actually made it tractable: a team that had done it before, a scoped problem, a correct protocol core, and a platform that removed the operations.
Aug 18, 2026 · 14 min read
A design file carries hierarchy, composition and tone. It can't carry how long something takes, what waiting feels like, or what happens when the data is missing. Our founding product designer on what changed when we stopped reviewing frames and started reviewing behaviour.
Jun 23, 2026 · 11 min read
Multi-tenant isolation often starts as a simple rule: every row carries a tenant ID, and every request remembers to filter by it. But as products grow, that boundary becomes fragile - repeated across queries, workflows, tools, automation, and AI-powered features. In this build note, we share how we moved from tenant isolation as an application convention to tenant isolation as a platform-enforced deployment pattern: one tenant, one Worker.
May 20, 2026 · 7 min read
The bottleneck in agentic AI was never the model. It is the system around it: connected data, clear ownership, durable governance, and operational state. As agent deployment becomes easier, the real advantage will belong to organizations that can move beyond demos and build the infrastructure agents need to run reliably in production.
May 7, 2026 · 10 min read
Most operational failures are not time management failures. They are state management failures. This article describes a complete reference architecture — four layers, five agent patterns, full integration topology — for teams that want to build agentic systems that carry context so their people do not have to.
May 4, 2026 · 2 min read
The new Salesforce lines up almost exactly with the thesis I have been writing toward at Reveriext. So we are starting our Salesforce practice, with everything we have learned about building AI.
Apr 25, 2026 · 6 min read
Vector search is a retrieval technique, not a database. A staff-level argument for the hybrid cascade that works in production — and for treating the vector index as a rebuildable view, never as the system of record.
Apr 17, 2026 · 6 min read
MCP is the right protocol and the wrong trust model. A staff-level argument for adopting MCP, treating every server as untrusted, and enforcing at the host layer what the spec declines to require.
No insights match these filters
Select a different format or practice, or choose All formats and All practices to view all insights.
Describe your objectives, current environment, and the support you require. We will recommend an appropriate scope and next steps.
