Solve the small problems —
someone endures them every day
The Builder stands on the road the Innovator cleared, takes the better tools it offers, and lands the
branching details of real scenarios.
On Xiaohongshu, plenty of women and humanities majors with zero technical background have shipped
excellent products — because they thought harder than anyone about one particular predicament.
Even if part of it is imitation and learning, that is completely fine. What a Builder solves is,
by definition, the small problems that press on ordinary people and ordinary organizations.
And those “small problems” are, for the specific person living them, a pain endured every single day.
ANE — AI Native Engineer
FDE is a term Palantir coined around 2010. At bottom it institutionalizes the grunt work: you have to get on site and inside the customer before you can read the real business logic and workflow, distill it, improve it, deliver a solution first and abstract a product second.
We think this role outlasts the name — long enough that it shouldn't be called FDE anymore.
| FDE | ANE | |
|---|---|---|
| Origin | Palantir, ~2010 | Proposed in this repo |
| Deployment | Must be forward-deployed | Need not be deployed, but must be present |
| Serves | Large enterprises (the ones who can pay) | Individuals / small orgs / cities |
| Backed by | In-house corporate methodology | An open WorkBench + digital public goods |
| Core skill | Domain understanding + on-site delivery | Domain understanding + orchestrating AI capability + creating new capability |
The ANE work loop
Four steps, plus a capability flywheel. Every step can bounce backwards — especially the one where you hunt for the pain.
End-to-end support
Not a course — a workbench. Wherever you jam in a real scenario, that's where we are.
AI Native selection
Which model for which scenario? A closed API or open weights you host yourself? Where the limits are, and when you should not use AI at all — that last one is often worth more than knowing how to use it.
Train your own small model
Fine-tune on the organization's own data for its own scenarios, until the model knows that organization inside out. This is a conclusion for now, and the future may break it — but today, it works.
Memory and context
How does memory actually get practiced and shipped inside context management? What are the technical options? What does each one cost you?
Agents and connectors
Scraping industry data, auto-connecting suppliers and pulling quotes, pushing new products to customers and handling after-sales — wire agents into the organization's real workflow, not into a demo.
AuraAI, as an organization that gathers ANEs, keeps accumulating and exchanging what it learns — so that more people can become Expressers, Innovators and Builders.
— The mission of this repository