# What Must Be True ## Ten conditions for the technology to support the platform-company thesis Each condition links to a result elsewhere in this map. "Against" means that the indications at this time show the opposite. "Insufficient" means that there are no indications for or against the thesis. | # | Assumption | Indication today | Concrete test | | --- | --- | --- | --- | | 1 | The platform layer is really shared. A deployment for a new customer in an existing vertical starts to operate in about two months. The work is mostly configuration. | **Against** | An inspection of the repository for the last three deployments. The inspection covers the elapsed time, the engineer-days, and the share of shared code and custom code. | | 2 | The company owns the data and the model weights behind each vertical. Or the company has permanent rights to use them again. | **Against** | A matrix of data rights for each vertical. An inspection of the terms of the institute and of the customer. The schedules of background IP. | | 3 | The proprietary part of the low-code engine is large. Each third-party component has a license that agrees with the use of the component by the company. | Insufficient | The bill of materials. A walk-through of the code. An inspection of the licenses. | | 4 | The fine-tuned model has better performance than an off-the-shelf model that has the same retrieval. | Insufficient | A blind test on a held-out task set. | | 5 | Two or more deployments in production give outcomes that a person can measure. The customer confirms the outcomes. | Insufficient | Calls with the customers. The calls give a baseline, a period, and a result. | | 6 | The stated mid-market buyer can pay for the deployment architecture. | Mixed | The topology for each customer. The total cost of ownership against the value of the ticket. | | 7 | The cost of the deployment for each customer decreases when the number of customers increases. | Insufficient | The engineer-hours for each deployment and the change of the engineer-hours with time. The integration margin for each quarter. | | 8 | The engineering team can do the work for the current customers and for the funded roadmap. | Insufficient | A named roster. The roles, the tenure, the vesting, and the hiring plan. | | 9 | An enterprise at the semiconductor tier and an overseas enterprise examine the protection of the systems of the company. The company can get approval from the two. | Insufficient | Completed questionnaires on protection from customers. A penetration test. A certification plan. | | 10 | No university, research council, or initiative has a claim on the core IP. | Insufficient | The IP assignment of the founder. Confirmation from a third party. The terms of the initiative. | > [!tip] How to read the table > Conditions 1 and 2 are the thesis. If the two conditions do not hold, the company is a good vertical-AI solutions business with an institute-led channel. If the two conditions hold, the platform narrative holds. All other conditions are execution risk. ## Why it matters This document is the underwriting document. It changes all of this map into ten statements that a test can show are incorrect. It works with [[Dispositive Questions]], which lists the questions to put to the company. It also works with [[Technical DD Framework]], which is the general method of the vault. ## Related - [[Foreman MOC]] - [[Platform or Solutions - The Central Question]] - [[Borrowed Data Is Not a Moat]]