Local AI assistant
Investigating small language models for useful everyday tasks, with personal data kept on the device.
Artificial intelligence
Independent technology research & developmentAdrian's
Adrian's
Sandbox Labs
A sandbox for emerging technology.
Exploring emerging consumer and computing technologies through hands-on research, software engineering, and rapid prototyping.
AI & SOFTWARE / CONNECTED DEVICES / COMPUTING PLATFORMSIDEAS → EXPERIMENTS → UNDERSTANDING
Small experiments.
Practical understanding.
Practical understanding.
A sandbox is a place to test an idea without assuming the outcome. Ours brings together artificial intelligence, connected devices, and computing platforms to explore what they can do, where they fall short, and what to try next.
01 / What we explore
All research areas Technology, put to the test.
Consumer experiences are shaped by the systems behind them. We explore both.
01 / RESEARCH
Artificial intelligence
AI that earns its place in everyday life.
02 / RESEARCHConnected technology
Devices that can work together.
03 / RESEARCHConsumer computing
New platforms, practical questions.
04 / RESEARCHCloud & distributed systems
The systems behind the experience.
05 / RESEARCHHuman-computer interaction
Keep people in control.
The interesting questions often sit between disciplines.
Explore the technology02 / In the sandbox
All experiments A question. A test. A next step.
Conceptual experiments that turn broad technology questions into focused investigations.
Smart home interoperability
Exploring the boundaries between device ecosystems, shared standards, and dependable home automation.
Connected technologyContext-aware web interface
Considering how an interface can adapt available actions without moving the goalposts for its users.
Human-computer interaction03 / Inside the sandboxBuild to learn.
Our method Build to learn.
Keep the question open.
01 /
Explore
Start with a practical question. Understand the platform, constraints, and existing approaches.
02 /
Prototype
Build the smallest implementation that can test the idea. Keep the experiment isolated and reversible.
03 /
Test
Try ordinary use, edge cases, and failure conditions. Include accessibility and privacy from the start.
04 /
Evaluate
Compare evidence with the original question. Separate measured results from interpretation.
05 /
Document
Explain what was tried, what happened, and what remains uncertain. Make the work understandable.
06 /
Iterate
Refine the question, try another approach, or stop. An idea that does not work can still produce useful knowledge.
04 / Sandbox projects
View projects From a question to a system.
PRJ / 001
Device event workbench
An inspectable workspace for understanding how connected-device events move through a system.
Concept designRead the project outline05 / From the lab
All lab notes Notes on the work ahead.
Engineering perspectives and evaluation frameworks. Clear questions before confident conclusions.
What makes a local AI benchmark useful?
Start with a task and a quality threshold, then measure the cost of getting there.
Read note: What makes a local AI benchmark useful?Designing a connected home for disconnection
Offline behavior is part of a device experience, not an exception to it.
Read note: Designing a connected home for disconnectionAdaptive should still mean predictable
Context can simplify a task, but stable navigation and user control come first.
Read note: Adaptive should still mean predictableAn open question is a good start.
Research & inquiries