日本語

Begin a
conversation.

Request a briefing or propose research. If you work on biological compute, persistent agents, or applied AI in clinical settings — we want to hear from you.

01 — WHO WE TALK TO

Four routes. One lab.

Research labs

Persistent-agent architecture, neural-substrate computation, memory continuity, and multi-agent mesh design. Joint benchmarks, co-authored papers.

research@kodasoken.com →

Infrastructure partners

Bio-compute platforms, neuromorphic vendors, HPC operators — run KoLo against real substrate constraints: latency, noise, state recovery, energy, observability.

partners@kodasoken.com →

Medical & healthcare investors

Healthcare AI infrastructure, Japanese medical technology, clinical documentation, and regulated-environment deployments for an aging society.

partners@kodasoken.com →

Press & analysts

Editorial enquiries, briefings, and on-the-record conversations. We respond to thoughtful questions.

press@kodasoken.com →

02 — THE LAB

KodaSōken Intelligence Labs.

Tokyo and Abu Dhabi. We build the KoLo Platform — a continuity-first runtime for persistent agents — and operate the research mesh that tests it.

hello@kodasoken.com →

03 — WRITING TO US

What makes a first message useful.

None of this is required. It simply shortens the distance between the first message and a conversation worth both people’s time.

The problem, not the productWhat is going wrong operationally. We would rather hear the constraint than a specification of what you think we sell.
What already existsThe systems in place, who runs them, and what has already been tried. Replacement is rarely the answer we reach for.
Who has to agreeClinical, legal, IT, procurement. Knowing the approval path early prevents a conversation that cannot land.
Data and jurisdictionWhere the data lives, what may leave the building, and which regulator has an interest.
Timing that is realA budget cycle, a contract end, an audit. Vague urgency helps neither of us.

04 — WHAT HAPPENS NEXT

Four shapes an engagement takes.

A conversationNo slide deck. We ask what breaks today and say plainly whether we are useful.
A scoped assessmentA narrow look at one workflow, with a written finding you keep whether or not we continue.
A controlled pilotOne department, one measurable outcome, a defined end date and an agreed definition of failure.
Research collaborationJoint benchmarks or co-authored work, with method and limitations published as agreed.

A pilot without an agreed definition of failure is not a pilot. It is a purchase with extra steps.

05 — WHAT WE WILL NOT DO

Worth saying before you write.

Promise a result we have not measuredWhere we have no evidence, we say so rather than borrowing a number from someone else’s deployment.
Take a clinical decisionOur systems support operations and documentation. Clinical judgement stays with clinicians.
Ask you to rip anything outThe platforms are built to sit above the systems you already run.
Take a pilot we cannot staffA pilot we cannot support properly damages both sides. We will say no.

A collaboration, a pilot, a partnership.

We read every message and reply within reasonable engineering turnaround.

See the KoLo Platform →The research systems ↘HPC & BioCompute Lab ↘
Begin a conversation →