Private beta opens soon. Apply for early access
TopOS Labs preprint · v4 · operating company Propose a collaboration

TopOS Labs · applied research · 2026

Research with a shipping deadline: two-to-four-week cycles that end in evidence

The TopOS Labs research bench1 · with academic and industry partners2

Propose a collaborationRead the method
Abstract

We run short, bounded research cycles across autonomous multi-agent systems, post-quantum cryptography, consensus mechanisms, and spatial computing. Each cycle begins with a sharp question and a success threshold, and ends in one of three outcomes: a validated prototype, a documented result, or a negative finding the next team can build on. Partner data and IP stay scoped to the program that produced them.

In plain terms: TopOS Labs is an operating company. You bring a sharp research question; the bench ships a working prototype — or a documented negative result — in two to four weeks.

Frame question & threshold
Baseline & threat model
Smallest prototype
Evaluation vs. threshold
Transfer or retire
week 1week 2week 3week 4

Figure 1. Anatomy of one cycle. Both outcomes count — a negative finding ships as a report, not a silence.

1Research areas

[cs.MA]

Autonomous multi-agent systems

Coordination, verification, and safe delegation for agents that act on real systems.

[cs.CR]

Post-quantum cryptography

Hybrid key exchange and signatures sized for constrained devices and networks.

[cs.DC]

Consensus mechanisms

Energy-efficient, high-throughput agreement for applications with demanding governance.

[cs.HC]

Spatial computing

Interfaces for spatial data, sensing, and on-device perception.

2Method

Questions are chosen because a product can use the answer. Every cycle is scoped to the smallest prototype that can confirm or refute a stated threshold, evaluated honestly, and written up with methods, results, and limits.

Prototypes that pass ship with the documentation and IP terms a partner needs to adopt them. Those that fail ship as negative findings — often the more valuable artifact for the next team.

outcome = threshold(prototype, evidence) ∈ { transfer, iterate, retire } (1)

Algorithm 1 — one research cycle
require question q, threshold τ, budget ≤ 4 weeks
1:  baseline ← measure(current_state) ▹ week 1
2:  p ← smallest_prototype(q) ▹ weeks 2–3
3:  e ← evaluate(p, τ) ▹ reported as measured
4:  return transfer(p) if e ≥ τ else report(e)

3Programs and data boundaries

Table 1. TopOS Labs is an operating company, not a tenant platform. Each program's data and IP stay with its sponsor.
StageScopeDescription
Programsponsor + questionA sponsor, a question, and agreed IP terms.
Experimentprogram-scopedPrototypes and data scoped to the program; never pooled across sponsors.
Resulttransfer-readyAn adoptable artifact, or a documented negative finding.

4Contributors — the research bench

Inventors and first-bringers: the Labs' internal bench for prototyping, evaluation, and cadence, provisioned through Idio. These are operating personas, not customer seats.

Daedalus

Greek

prototype_forge Built the labyrinth and the wings — invention under deadline; the prototype seat

Väinämöinen

Finnish

song_smith The eternal sage who sings things into being and forged the Sampo — invention by iteration

Imhotep

Egyptian

deified_inventor The human who became a god through invention — the architect seat

Oannes

Mesopotamian

knowledge_bringer The fish-sage who taught humanity civilization — first-principles invention

Kui Xing

Chinese

exam_star The ugly-footed star official who tests the scholars — the benchmark/eval seat

Nanna

Mesopotamian

moon_calendar The moon — the four-week cadence made god

Mamaragan

Australian Aboriginal

lightning_rider Rides thunderbolts and speaks in the storm — the high-voltage experiment seat

Amirani

Georgian

chain_breaker The Caucasus' Prometheus — stole fire, kept defying the gods

Menrva

Etruscan

north_minerva The Etruscan craft-wisdom — older register for applied knowledge

Bragi

Norse

mead_poet God of the improvised verse — demos and narrative; invention you can show

Khonsu

Egyptian

time_traveler The moon's wanderer — marked the temple doors and bends time

Tāne

Māori

sky_splitter Pushed sky and earth apart to make room — making space for new things to exist

AEthics, provenance, and data availability

Governed like the products we ship — the same identity, policy, provenance, and consent controls run under every Colossal Capital product.

  • AI provenanceEvery AI-assisted result in TopOS Labs is disclosed and traceable — which persona acted, which model ran, on whose authority, from which sources.
  • Built for AI regulationTransparency notices, human oversight, risk documentation, and audit logs designed to support EU AI Act, Colorado AI Act, and NIST AI RMF / ISO/IEC 42001 programs.
  • Private by designEntity-scoped storage, data minimization, encryption in transit and at rest, and GDPR / CCPA-CPRA request handling. Customer data never trains a shared model without opt-in.
  • Fine-grained provisioningIdio SSO over SAML and OIDC, entity context, role and attribute policy, and a named, scoped, revocable identity for every agent.
  • Humans at the gateConsequential actions wait for an eligible approver, and every decision lands in an exportable audit trail.

Private models, at the scope you choose — nothing rolls up on its own.

  • UserOne researcher's notebook preferences
  • ContentOne experiment's authorized data
  • EntityOne program's private models
  • OrganizationApproved results across a sponsor's programs

Partner data is used only within its program and is never pooled across sponsors. Shared ecosystem learning is a separate, sanitized, opt-in path. AI transparency notice.

Correspondence

Bring us the hard question.

We take on a small number of research collaborations each quarter. Leave an address and a sentence about the problem.