Intelligence
is an architecture.
We study the relationships between computation, memory, reasoning, action, and evaluation — where small design choices can change the whole system.
RESEARCH / SYSTEMS / SYMPLOKAEvery capability has a cost.
Compute, memory, latency, and verification change what a system can do—and what it takes to use it.
Four questions.
One connected field.
Open a program for its research questions, method, evaluation criteria, and selected original sources. Together, these tracks connect mathematical structure to working software.
Efficient intelligence
Use computation where it changes the result.
Agents and agency
Give agents goals, tools, and reviewable boundaries.
Memory and reasoning
Bring relevant, sourced context into later decisions.
Evaluation and safety
Test behavior, reveal limits, and keep people in control.
Better structure changes what intelligence can do.
Selective computation, persistent knowledge, and verifiable action give us concrete ways to rethink how an intelligent system is built.
Read our vision ↗We look at the relationships between computation, context, reasoning, action, and evaluation.
Explore the systems view ↗From a question
to a method.
Choose a program to explore its central question, method, and the lines of inquiry it brings together.
The shape of intelligence matters as much as its scale.
When can selective computation outperform always-on scale?
A fair comparison uses clear baselines, matched tasks, and stated resource budgets. It considers task quality alongside latency, memory use, and total compute, so an efficiency claim remains meaningful in practice.
Ambition, grounded
in measurement.
Clear questions, careful comparisons, and transparent analysis help turn a promising idea into useful knowledge.
Define the question
State what a system should do and what evidence would count as progress.
Compare fairly
Use clear baselines and consistent data, compute, and evaluation conditions.
Study failure
Look beyond ideal cases to edge conditions, errors, and recovery.
Share what matters
Explain methods, trade-offs, and findings so others can learn from them.