A Brief Discussion on FLock: Initial Exploration of the AI Large Model Launchpad Model

Written by: Haotian

I listened to the FLock 2025 annual performance report, and during the meeting, they mentioned something about creating a Launchpad for AI large models, which really caught my interest.

What? Another Launchpad? How does the large model issue assets? It's actually quite easy to understand; just draw an analogy and it will be clear:

The launchpad of AI agents like Virtuals Protocol, driven by the application layer, provides asset incentives through a token mechanism to help agents evolve from “chatting” to x402 “payment”, and ultimately to the goal of “autonomous trading” and providing complex services.

The AI Model Launchpad that the FLock project aims to create is driven by an infrastructure layer, providing assets to the trained large models, specifically a large number of vertical scenario models, such as medical diagnosis, legal documents, financial risk control, and supply chain optimization, among others.

Although the training costs of these vertical models are relatively controllable, the commercialization path is extremely narrow; either sell to big companies or open-source for love, and there are very few sustainable monetization methods.

FLock intends to reconstruct this value chain using Tokenomics, providing assets to the fine-tuned large model, thereby allowing data providers, computing nodes, validators, and others who contribute to model training to have the possibility of obtaining long-term profit rights. When the model is called and generates income, it can be continuously distributed according to the contribution ratio.

Creating a Launchpad for large models sounds fresh at first, but essentially it is about driving product development through financialized methods.

Once the model is assetized, the trainers will have the motivation for continuous optimization, and once the profits can be continuously distributed, the ecosystem will have the ability to self-generate.

The benefits of doing this are undeniable. For example, the recent popular nof1 large model trading competition currently only has general large models participating, without specialized fine-tuned models in the competition. The reason is the lack of an incentive mechanism. Excellent specialized models usually tend to earn quietly and cannot be exposed. However, if there are assets involved, it becomes significant. This type of large model Arena competition becomes a public stage to showcase strength, and the competitive performance will directly affect the asset performance of large models. Can you imagine the possibilities?

Of course, at present, FLock has only proposed a direction and has not yet been implemented. It is still unknown what the similarities and differences are between the asset issuance model and the agent asset issuance model.

But one thing is certain: how to ensure that the asset issuance model is based on real demand rather than just increasing numbers, and how to effectively ensure PMF (Product-Market Fit) within vertical scenarios, all of these are issues. It should be said that the problems faced by Agent applications in the wave of token issuance are also unavoidable.

I'm just looking forward to it. What different kinds of gameplay will there be for the Launchpad direction for the Model?