How we use Salana
How we use Salana
Salana is built inside our own memory. This is what it holds and how it runs our companies.
#One memory for everything
We run several small companies from one Salana Memory. It started as a plain folder of notes and now holds every decision, customer fact, process and task we have. Every page in these docs was written from it.
Screenshot comingThe Graph page of our own memory, numbers hidden
#What is inside
| Folder | What it holds |
|---|---|
| North Star | The one number we steer by, and the score against it |
| Today | A briefing written every morning at 07:00 |
| One folder per company | Brand, customers, data, decisions, everything that company knows |
| Tasks | More than 1,600 task cards, one per piece of work done |
| Skills | About 60 written procedures agents can run |
| Automations | 18 jobs that run on a schedule |
| Proposals | Things an agent suggested keeping, waiting for a yes or no |
| Logs | What ran, when, and how it went |
| Wiki | Research and knowledge that crosses companies |
#The rules we keep
- Write it down once. A decision lives in one note. Everything else links to it.
- Every piece of work gets a card. An agent opens a task card when it starts and closes it with what it did. Nobody updates the board by hand.
- Agents propose, we decide. At the end of a chat an agent may suggest up to three things worth keeping. We press accept or reject. Nothing is stored behind our back.
- Once is a note, twice is a skill. If we ask for the same thing twice, it becomes a skill an agent can repeat.
- Plain files. No database we cannot open. If Salana disappeared tomorrow, the memory still works in any editor or AI.
#Why it works
A new chat never starts from zero. When we ask "draft this week's newsletter", the agent already knows the brand rules, the last ten sends, what sold and what did not. We stopped explaining things twice.
