Architecture
Framework
The public architecture: Right Model, Right Version, Right Role, Right Culture, and Right Compute, with customer work at the center.
Research / Thinking
Thinking about how organizations keep intelligence understandable
Model Culture explores how organizations can understand, compare, and retain control of the intelligence they use in real work. This space is for the ideas surrounding the framework: organizing intelligence, independence, evidence, provenance, and the economics of AI work.
Research / Thinking tracks questions Model Culture is actively exploring. Where public writing or findings are released, they can be linked here. Otherwise, the topic remains an area of inquiry.
Positioning
The architecture, the inquiry, the software, and the open questions
Model Culture Studio is working private software. Research / Thinking tracks questions that remain open around the public architecture. One does not replace the other.
Inquiry
Research / Thinking
Active questions about organizing intelligence, workload fit, optionality, evidence, and change. Public writing can be linked here as it is released.
Private application
Implemented software
Model Culture Studio is a working private R&D and evaluation application used to operationalize and develop the methodology.
Unresolved
Open research questions
How organizations should notice meaningful change, retain evidence across providers, and keep optionality without renting their understanding remains an open field of work.
Themes
Research areas
Each theme is an area of inquiry. Links will appear here as public writing is released.
Organizing Intelligence
The starting problem is organizational: which intelligence should do which job, under which conditions, and what should remain knowable after the choice is made.
Configuration vs. Model Identity
A model family name does not identify what was evaluated. Version, role, surrounding context, and compute can make the same family a different configuration.
Workload Shape
The shape of the work (task, context, concurrency, tools, and expected response) connects the role being asked of the AI to runtime and compute choices.
Product Culture
Culture concerns observable behavior while performing the role: evidence discipline, uncertainty handling, verification behavior, communication, and consistency. It is not only prompting.
Evidence and Provenance
A decision should remain inspectable after models, vendors, versions, or people change. Provenance is part of that inspectability.
Runtime and Compute
Compute includes feasibility, economics, privacy, latency, throughput, and local, cloud, or hybrid operating context. It is not merely hardware.
Independence and Optionality
Organizations may use commercial, open-weight, local, sovereign, cloud, or hybrid intelligence. Understanding of that intelligence should remain independent of any one provider.
Revalidation and Change
When the work, the version, or the operating conditions change, prior evidence may no longer apply. How an organization notices that, and what it re-evaluates, remains an active question.
Economics of AI Work
Nominal token price is not the same as the cost of completing the work. Cost per successful task is a supporting economic concept.
Continue
The public architecture remains the framework
Research / Thinking sits around that architecture. The next useful step is the five-factor framework, or a conversation about a specific workload.