01 / AGENT DESIGN
Agent design
How we draw the boundary between agent autonomy and human approval. Tool-use schemas. Prompt scaffolds that survive contact with a Slack thread of fifty people.
OPENRIFT // BLOG // FIELD NOTES
Notes from
the rift.
Long-form writing on what we're learning as we deploy agents across SEA. Case studies, post-mortems, and the engineering behind bespoke deploys. Nothing live yet — we're shipping before we publish.
What goes here.
Four lanes02 / SEA SMB OPS
SEA SMB ops
Case studies from real Singapore, Jakarta, Manila, KL clients. What we found in discovery. What we built. What broke in week three.
03 / INTEGRATION PLAYBOOKS
Integration playbooks
How we wire agents into Stripe, HubSpot, Xero, BigCommerce, GitHub, and the long tail of regional tools. The OAuth gotchas no docs warn you about.
04 / POST-MORTEMS
Post-mortems
When agents go wrong, we write about it. Public retros on workflows that failed, scope we mis-read, and how the fix shipped.
On the slate.
ETA — drafts in flight2026-06Case study
How a Singapore B2B SaaS cut 14 hours of weekly billing reconciliation to zero.
First public deploy. What we found in week-one discovery, the agent we built, the Stripe + Xero reconciliation flow now running on its own.
In drafting
2026-07Playbook
Building agents that live in WhatsApp without becoming chatbots.
Why WhatsApp Business API is the wrong abstraction for agent work, and the harness we built to make it tolerable.
In drafting
2026-07Engineering
The approval boundary — how Openrift agents ask before acting.
Designing the trust ladder. Tool-call gating. Audit logs. Why we chose synchronous confirms over async retries for high-stakes actions.
In drafting
// BACK TO MAIN
The work is the writing.
We're heads-down on deploys. When we have something worth saying, it lands here first.