Intent
Non-trivial work starts with an understandable objective and closure criterion.
Configurable software engineering process framework
An AI-native engineering method for building and operating software when humans, AI agents, and deterministic automation participate together in the delivery loop.
“Generating code has become cheaper. Understanding the problem, making sound decisions, verifying outcomes, and owning risk remain scarce.”
01 / WHY IT EXISTS
REM assumes AI can drastically accelerate artifact production, but it does not remove the need to select work, control risk, make decisions, verify outcomes, and preserve human accountability.
Non-trivial work starts with an understandable objective and closure criterion.
Agents may investigate and build; residual risk and material decisions still have a human owner.
A change is not done because the agent stopped typing. It is done when sufficient evidence exists.
Human attention is the scarce WIP. REM limits started work to favor actually finished work.
02 / KERNEL
Activities remain distinguishable even when they are not separate ceremonies. Process weight scales with risk, uncertainty, and decision permanence.

First, orient the work
Choose what deserves attention now.
Make objective, scope, constraints, and closure explicit.
Then, produce evidence
Reduce uncertainty with evidence.
Choose and record decisions that matter.
Build with humans, agents, and automation.
Prove the system distinguishes correct from incorrect behavior.
Check the real outcome when production is part of the evidence.
OBS applies when the outcome must be checked in a real environment.
03 / WHAT REM IS
REM separates its normative method specification from each project's local rules. Architecture, branching, tooling, and approvals can vary without erasing the method kernel.
One standard. Many ways to execute it.
METHOD.md defines the kernel, states, responsibilities, invariants, and conformance rules for REM 1.1.
REM provides a configurable structure instead of one rigid universal sequence.
Each project adopts REM through its Constitution and local policies while preserving the required kernel.
04 / WHAT IT GOVERNS
Constraints that protect the outcome
Where work comes from and how attention is allocated.
An explicit policy limits active and verifying work.
Separate responsibilities for humans, agents, and deterministic automation.
V0–V3 levels and evidence proportional to risk.
DEC, ADR, ARCH, Megaplan, Plan, incidents, and audits only when they actually earn their cost.
WIP, throughput, age, and cycle time instead of story points and visible activity.
05 / ADOPTION
REM 1.1 can start in one project, but installation includes all document types, the Constitution, scripts, hooks, and CI. Day-to-day work uses only the artifacts justified by its risk and scope.
The checklist and rem-doctor check adoption structure and rules that can be automated. Product outcomes require their own evidence.
Open installation checklistDecide between a multi-repo hub or one-repo installation, WIP, and policies with the human owner.
Copy the eight document types, templates, scripts, hooks, and CI without skipping components.
Record architecture, repositories, security, changelog, and local exceptions.
Give existing documentation a home and open a Megaplan when objectives depend on each other.
Run rem-doctor --adoption without errors; verify and observe the real outcome when needed.
06 / FOUNDATIONS AND EVALUATION
REM is a versioned engineering proposal combining established practices with RelataLabs' internal experience. Its effect on a project is evaluated using data from that context; sources, rules, and the evaluation approach are open to review.
Kanban, DORA, Shape Up, ADRs, and specification-driven development contribute distinct ideas. The documentation identifies what REM incorporates and how it adapts those ideas to work with agents.
Read the foundations and sourcesProcess conformance, product verification, and outcome evaluation address different questions. The guide proposes a baseline, flow metrics, and limits on interpreting change.
Read the evaluation guideAn illustrative case connects the objective, research, decision, implementation, tests, and closure. Its data and outcomes are teaching examples, allowing readers to follow the method without presenting it as a real case study.
Explore the illustrative exampleEvaluation status. The specification and its tools can be inspected and tested. Independent evaluation of its effectiveness remains open; internal experience informs its evolution.
07 / OPEN LICENSING
REM documentation and templates use CC BY 4.0. Executable tooling uses Apache-2.0. The licenses do not grant trademark rights or permission to present a modified variant as an official RelataLabs publication.
METHOD.md, documentation, templates, and written material.
Scripts, workflows, and executable REM tooling.
08 / CANONICAL SOURCE
The normative specification and its history remain on GitHub. This page is the public entry point, not a parallel copy of the standard.