Control Surface for Human–LLM Decision Discipline

Helm

A system for maintaining cognitive structures over time — where LLMs propose and humans commit.

The core problem

Humans are remarkably capable — and remarkably bad at maintaining complex cognitive structures over time.

As ambition grows, individuals are forced to mentally carry:

long-horizon intent
evolving assumptions
interdependent decisions
branching possibilities
partial work and abandoned paths

These structures decay under pressure. Clarity is lost. Work fragments. Progress resets.

Helm exists to maintain those structures.

What Helm is

Helm is not a task manager.

It is not a chatbot.

It is not a productivity hack.

Helm is a control surface for decision discipline.

It tracks assumptions, decisions, and artifacts as a connected graph. When assumptions change, Helm surfaces what needs review. LLMs generate proposals; humans promote or prune.

What Helm refuses to do

Helm enforces a strict boundary between proposal and commitment:

It does not auto-apply suggestions.
It does not commit on your behalf.
It does not hide state transitions.
It does not silently mutate your decisions.

Every change to state requires explicit human action. This is not a limitation — it is the point.

Decisions as a living graph

Helm treats decisions as persistent, connected structures rather than isolated choices.

Every significant decision is represented as a node within a causal graph, linked to:

the assumptions it depends on
the work it influences
related decisions and explorations

When assumptions change, Helm surfaces the decisions and work that may be affected, allowing you to review, revise, or reaffirm them.

Adjustable structuring

Helm provides a spectrum of proposal intensity:

minimal: record and organize what you provide
moderate: surface related decisions, suggest structure
active: generate proposals, draft artifacts, flag attention points

At every level, proposals are previewed before application. Nothing changes without explicit human action.

Who Helm is for

Helm is for humans doing long-horizon work with LLMs who need:

a navigable decision space rather than chat history
visibility into what assumptions their work depends on
proposals they can review, revise, or reject
leverage without surrendering control