The bridge between what we mean and what machines build.
AI now writes code faster than people can. What it can’t supply is the intent: what a system is for, what it promises, what it must never do, and who decided. Structured intent is the discipline of writing that down, so machines can build from it and check against it.
A real structured intent. Tap a clause to see where it came from and how each implementation keeps it.
goal Move money between two accounts, safely and at most once per request.
Lifted from legacy/ledger.c:180. Ruled a promise by the ledger owner.
10,000 seeded random transfers. A formal proof is written but has not been run yet.
Code preserves what a machine does. Structured intent preserves why it should do it.
Author it, lift it or merge it. Then lower it.
Intent can be written for a new system, recovered from existing code, or combined from several systems. Every implementation is then generated from the same intent and reports how strongly it keeps each promise.
Why now
Each advance in programming disciplined something that had been left to chance. AI makes the next one practical, because lifting intent from code and lowering it into many technologies is finally cheap.
-
1960s–70s
Structured programming
Disciplined control
Sequence, selection and iteration replaced tangled jumps, so a program’s flow could be reasoned about.
-
1980s–90s
Object-oriented programming
Disciplined state
Data was bundled with the behavior allowed to change it, so systems could be organized around things.
-
2020s–
Structured intent
Disciplined purpose
Software is organized around what it is for: its promises, its limits, and how it trades one goal against another.
Latest news
All news-
Structure Intent manifesto and repository published
Version 0.2 of the manifesto, a draft notation for structured intent, and a worked example with Rust and Python implementations.
-
Intel details its approach to translating C into safe Rust for DARPA
Intel's IDEAS work for DARPA's TRACTOR program uses AI for translation, kept in check by equivalence testing at every step.
-
Daniel Miessler: from prompt engineering to intent engineering
Miessler urges people to stop telling AI how to do things and start describing the outcomes they want.
-
Microsoft Research: intent formalization is the grand challenge
Shuvendu Lahiri argues that turning informal intent into checkable specifications will decide whether AI makes software more reliable or just more abundant.
Machines act. People judge.
When building is nearly free, a wrong plan costs more than ever. The valuable work moves to choosing the right questions, ranking priorities and ruling on what a system should do. Structured intent is where that judgment is kept.
Where human value movesFollow the field
New research, tools and advances in structured intent, every week.