FROM THE LAB.
We use prototypes as a way to answer real questions about AI: where it helps, where conventional software is stronger, what people need in order to trust it, and what happens when a promising demo meets reality. These notes document what our products are teaching us as we build.
Experiments, prototypes and product research from Darkstone Labs — including the decisions, trade-offs and lessons behind the work.
Memory as Interface: Exploring AI for Digital Legacy
Prototype / ResearchPhotographs survive. Context often does not. JulezAi explores whether AI can make a person’s documented stories, memories and voice easier to navigate without pretending technology can recreate the person behind them.
From Conversation to Action: Designing Meeting Intelligence That Preserves Context
Research / BuildingA meeting summary is useful. A searchable memory of what was actually said, decided and left unresolved is much more valuable. FireTaker is an exploration of that difference.
Scheduling Under Constraint: What Ploky Is Teaching Us About Operational Software
ResearchComplex scheduling is rarely an AI problem first. It is a constraints problem. Building Ploky has reinforced why explicit rules, priorities and auditability can matter more than adding intelligence for its own sake.