Learn this well enough to decide for yourself
How these are written
Every guide states a capability and its limitation together. That is not caution for its own sake — the failures of this technology are not obvious from using it, because it is wrong in the same confident voice it uses when it is right. Knowing where the edges are is the literacy.
Start here. No prerequisites.
What this technology actually is
No metaphors about brains. What these systems do, mechanically, and why that explains both the impressive parts and the failures.
7 min read
What it cannot do, stated plainly.
What it cannot do
The honest list. Worth reading before you spend money on anything in this field, including on us.
6 min read
Business, nonprofit, farm and land.
AI for a small business that is already busy
Where the hours actually are, which ones are worth handing over, and which ones are the business itself.
8 min read
AI for nonprofits and community organisations
Nearly every organisation already uses this. Very few see any impact. The gap is governance, and the gap is closable.
8 min read
AI for farms, gardens and land access
What is genuinely useful on a working piece of land, and where the confident-sounding answer will cost you a season.
7 min read
Privacy, policy and accountability.
Privacy, surveillance, and what never to type
A short, usable list you can hand to your staff this afternoon, and the reasoning behind it.
6 min read
Rights, transfer and independence.
Creative work, rights and likeness
What you can control, what you cannot, and what to put in writing before you publish anything.
7 min read
Who owns what gets built
The questions to ask before you commission anything from anyone — including from us.
5 min read
The words, in plain language
Large language model
A system trained on very large amounts of text to predict likely continuations of a passage.
It has read an enormous amount and learned what usually comes next. That is genuinely useful and it is not the same as knowing anything.
Hallucination
Fluent, confident output that is factually wrong — an invented citation, a plausible statute that does not exist, a quote nobody said.
It makes things up in the same voice it uses when it is right. There is no tell. You have to check.
Training data
The material a model was built from.
If your work was on the open web, it may be in there. Removing it afterwards is not generally possible.
Prompt
The instruction and context you give a model.
What you asked for and what you told it. Most disappointing output is an under-specified request.
Data sovereignty
The principle that a community or organisation holds authority over information about it — how it is collected, held, used and returned.
Your information stays under your say-so.
CARE Principles
Collective benefit, Authority to control, Responsibility, Ethics — a framework for Indigenous data governance developed by the Global Indigenous Data Alliance.
Settle who decides before you settle which database. Developed in and for Indigenous data sovereignty contexts; cited here with credit, not adopted as ours.
Vendor lock-in
A dependency that makes leaving a supplier disproportionately costly.
You can check out any time you like. Getting your data out is the part to test before you need it.
Human in the loop
A design where a person reviews or approves before an automated output takes effect.
A named person looks before it goes out. It only counts if they have the time and the authority to say no.
Heirloom Intelligence
Our own organising synthesis: knowledge, systems and technologies deliberately preserved, activated and transmitted across generations.
We borrow the logic of seed-keeping — deciding which lines are worth saving and passing on — and apply it to knowledge and technology. It is our framing, not an inherited or ceremonial tradition.
Things people ask
Do I need to know anything about AI to start?
Is Heirloom Praxis only for Black and Afro-Indigenous people?
Are you telling people AI is good, or telling people to avoid it?
Do the tools on this site send my answers to an AI company?
Is the Heirloom Shield a certification?
How accurate is the Crop Planner?
Something missing? Tell us what you wanted to find and we will write it.
