About this library
A reference library of named, reusable design patterns for the trust layer between humans and autonomous machines — vehicles, surgical and industrial robots, drones and warehouse systems. It is for practitioners, and it is also a learning tool written in public.
On this page
What a pattern means here
A pattern is a named, reusable solution to a problem that keeps recurring — the same sense the term carries in software design. It is not a component, a guideline or a piece of advice. It has a name you can say out loud in a design review, a mechanism you can draw, and a set of conditions under which it applies and another under which it does not.
What narrows this library is the second half: autonomy with physical consequences. Most AI pattern libraries cover AI inside software, where a bad output is a bad output. Here the machine moves, cuts, lifts or drives, so a miscalibrated handoff is measured in seconds and metres. That is why every pattern carries a risk and a reversibility, and why friction is sometimes the feature.
Anatomy of a pattern page
- 01IntentOne sentence: what the machine does, and what it buys the human.
- 02ProblemThe competing forces the pattern resolves, not just the symptom.
- 03The mechanismThe sequence, step by step, with the key moment marked.
- 04Without / withThe same three beats, played twice, so the difference is visible.
- 05When to use itThe conditions that justify it — and the ones that rule it out.
- 06Design considerationsImperative rules, each with the reasoning behind it.
- 07In the fieldExamples, each labelled observed or speculative.
- 08Related patternsWhat this requires, precedes, complements or fights with.
- 09ElsewhereThe same idea in other libraries, linked rather than copied.
- 10SourcesWhat the pattern is built on, grouped by kind.
- 11Still figuring outOpen questions. The learning shown in public.
Sections appear only when there is something to put in them, so a young pattern is short rather than padded with empty headings.
Status
- seed
- A candidate: name and intent only. Kept in the repo but off the index until it is written up.
- draft
- Mechanism and problem written, with few sources so far.
- reviewed
- Every section filled, at least three sources and at least one observed example.
- stable
- Reviewed, then revisited after real-world use or feedback.
Seeds let candidate patterns live in the repo without making the public library look thicker than it is. They are hidden on the index until you ask for them.
Trust jobs
What the human actually needs from the machine at that moment. A pattern does one or two of these; more than two usually means it is really two patterns.
- Know
- See the machine's state and reasoning.
- Predict
- Anticipate what it will do next.
- Intervene
- Act, override or stop it.
- Recover
- Get back to a safe state.
- Calibrate
- Trust it the right amount over time.
The autonomy scale
Each pattern names the range of automation levels it applies to. The scale below takes SAE's driving-automation levels as its reference point, generalized because this library also covers robots, drones and warehouse systems.
- 0 None
- The human performs the task. The machine does nothing on its own.
- 1 Assistance
- The machine handles one narrow part of the task. The human does the rest and stays responsible throughout.
- 2 Partial
- The machine controls several parts of the task at once, but the human supervises continuously and is expected to intervene.
- 3 Conditional
- Within defined conditions the machine runs the task itself. The human can disengage, but must take over when asked.
- 4 High
- Inside its operating envelope the machine handles everything, including reaching a safe state on its own if nobody responds.
- 5 Full
- No operating envelope limits, and no expectation of a human fallback.
How patterns relate
A relationship is written once, on one pattern, and shows up on both — inverted where that is what makes it read correctly. If The Handoff requires Shared Mental Model, then Shared Mental Model lists The Handoff under “Required by”.
- Requires
- The other pattern has to be in place for this one to work at all.
- Complements
- They reinforce each other, but neither depends on the other.
- Alternative to
- Two ways of solving the same problem. Pick one.
- In tension with
- Applying one makes the other harder. The trade-off is the point.
- Precedes
- One runs before the other in the control loop.
Risk and reversibility
Together these decide how much friction a pattern can justify. They are not used to navigate the library — they are there to be argued with, and to show where the coverage is thin. The risk × reversibility view lays every pattern out on both axes.
- Low
- A mistake is an inconvenience. Nothing is damaged and nobody is hurt.
- Elevated
- A mistake costs time, money or equipment, or puts someone in a situation they have to recover from.
- Critical
- A mistake can injure or kill, or do damage at a scale that reaches past the people operating it.
- Reversible
- Undo it and the world is as it was.
- Costly
- It can be put right, but it takes time, money or an apology.
- Irreversible
- Once it has happened, it cannot be taken back.
Observed and speculative examples
Every example in “In the field” carries one of two labels. The distinction is not decoration: while examples are still being collected, it is the thing that keeps the library honest about how much of it is evidence and how much is argument.
- Observed
- A real system or a documented incident, with a source you can follow.
- Speculative
- The author's own concept — plausible, but not yet seen in the field.
How to cite
Cite an individual pattern, not the library as a whole:
Johnson, L. (2026). {Pattern name}. Trust Layer Patterns. trust-layer-patterns.vercel.app/patterns/{slug}
For example:
Johnson, L. (2026). The Handoff. Trust Layer Patterns. trust-layer-patterns.vercel.app/patterns/handoff
The patterns are shareable under CC BY 4.0, so you are free to quote and adapt them with attribution. The changelog records what changed and when, if you need to cite a particular state of a pattern.