Winners LabAll experiments Dependencies
Which steps rely on other decisions being correct.
WINNERS LAB / 0701 / PURPOSE
Make the reasoning visible.
Lattice explores visible planning for complex AI work. It investigates how goals, dependencies, decisions, and revisions can be communicated so that people can meaningfully steer a system.
WHY IT MATTERS
Clearer plans could help people identify mistakes earlier, understand tradeoffs, and intervene before a small misunderstanding grows.
02 / RESEARCH QUESTIONS
What we look closer at.
Can visible plans make long-running agents easier to steer?
Legibility
Whether a plan is useful to a human reviewer, not only internally consistent.
Intervention
Where a person can redirect the work without having to start over.
03 / OBSERVATION
Observe. Question. Refine.
The research lens for Lattice: the questions that guide careful observation and review.
01Map the objective+
Identify the choices and dependencies behind a goal.
02Expose decision points+
Make consequential branches and unresolved questions visible.
03Revise the path+
Study how the plan changes when constraints or priorities shift.
People remain part of the picture.
Winners Lab’s experiments are carefully watched to deepen understanding of AI and support the betterment of humanity. This page describes the research purpose; it is not a public application or a live activity feed.
04 / PERSPECTIVE
A useful distinction.
A visible plan is not a transcript of private model reasoning. The focus is useful decision summaries and observable task structure.
Read the lab’s research framework ↗