# Loop & Graph — the method we install now

How we organize the work today is called Loop & Graph — versioned, not eternal. When a stronger method ships, Helixfield uses that.

How we organize the work today. A loop is a repeating job. A graph is a shared list of facts. Every turn is gather, do, write down, save. A named person accepts important facts.

## Gather the facts

Before the helper answers or acts, it pulls the facts you already agreed on. Hunches stay labeled as hunches. Guessed numbers stay out.

## Do the job

Draft, route, or look something up — with those facts in view. If a needed fact is missing, it asks. It does not invent a number, a client, or a permission.

## Write down what changed

After the turn, it lists new facts, decisions, and risks. Each item is marked proposed, accepted, or ask.

## Save it

Those notes go back into the shared list. Only a person can mark an important fact as final. The next turn is smarter because this one closed.

## Questions

### What is Loop & Graph?

How we organize the work today. A loop is a repeating job. A graph is a shared list of facts. Every turn is gather, do, write down, save. A named person accepts important facts.

### What does pack mean?

Before the helper answers or acts, it pulls the facts you already agreed on. Hunches stay labeled. Guessed numbers stay out.

### Why do business chatbots lose context?

Because chat history is not company knowledge. Useful facts live in people’s heads and in tools that do not talk. A shared list is the memory. Chat history fades.

### Who can accept a fact?

Only a named person. Helpers may propose. A model guess cannot replace an accepted fact. Anything that leaves the company stays a draft until someone says send.

### Is Loop & Graph Helixfield’s forever method?

No. It is what we use now because it is useful and teachable. When a stronger method ships, we use that. The survey also picks the AI tools from access, data rules, cost, and your preference — not from a partnership.