RESEARCH / OPEN QUESTIONS

What can a computer become?

I’m interested in the limits we treat as fixed: what a device can compute, where its work must run, and whether better algorithms can change the question before more hardware enters it.

I build systems to test those limits. Sometimes the answer involves another machine. Sometimes it is a better local calculation, or work that never needed to happen. These notes follow the decisions, experiments and reversals that changed my understanding.

01 / Local CPU experiments

OBP v2

Before I decide where a calculation should run, can I work out which parts actually need to happen?

Removing unnecessary work produced large gains; moving it was a different question.

Algorithms / Performance / Distributed systems

Read the research

02 / Prediction experiments and simulations

ASPR

If moving data is expensive, could I prepare it before the application asks for the next calculation?

Predicting some useful inputs is easier than predicting everything a calculation needs.

Prediction / Data residency / Performance

Read the research

03 / Lossless encoding experiments

ReSynth

Can I send a smaller description of an input and reconstruct exactly the bytes the calculation needs?

Fewer bytes can help, but only if encoding and reconstruction cost less than the transfer they remove.

Compression / Data residency / Performance

Read the research

04 / Native-call interception experiments

Splice

Could I change where a calculation runs without asking the application to become a distributed program?

Redirecting a call is possible; preserving everything its caller expects is the harder problem.

Interception / Runtime engineering / Systems security

Read the research

05 / Live application experiments

AutoCAD

Could I send a calculation out of a live drawing session and return an answer AutoCAD would actually use?

Live result consumption worked on selected geometry. Local execution delivered the useful performance gain.

Interception / Performance / CAD

Read the research

06 / Execution and interoperability experiments

Runtime

What happens to a fast calculation once I include everything needed to run it and return a usable answer?

Correct remote execution often lost; removing setup from a timer did not remove its cost.

Runtime engineering / Distributed systems / Performance

Read the research

07 / Historical execution experiments

Fusion Alpha

Could I divide a workload between local and external compute without losing the benefit to startup and data movement?

A ready GPU could win on a larger resident input; cold execution and planner decisions told a less encouraging story.

Distributed systems / Performance / AI infrastructure

Read the research