By: Haotian
I listened to FLock's 2025 annual performance report, and one thing mentioned in the meeting about creating a Launchpad for large AI models really caught my attention.
What? Another Launchpad? How do you issue assets for large models? Actually, it's quite easy to understand with an analogy:
Launchpads for AI Agents like Virtuals Protocol are application-layer driven, issuing assets for Agents. They use token incentive mechanisms to help Agents evolve from "can chat" to x402 "can pay", and ultimately to the final goals of "can autonomously transact" and provide complex services.
However, the AI Model Launchpad that FLock plans to create is infrastructure-layer driven, issuing assets for trained large models, specifically a large number of vertical scenario models, such as those for medical diagnosis, legal documents, financial risk control, and supply chain optimization, etc.
Although the training costs for these vertical models are relatively manageable, their commercialization paths are extremely narrow. They either get acquired by big companies or go open-source, fueled purely by passion, with few sustainable monetization methods.
FLock intends to restructure this value chain using Tokenomics, issuing assets for fine-tuned large models, thereby giving data providers, computing nodes, validators, and other participants involved in model training a possibility for long-term profit rights. When the model is called and generates revenue, it can be distributed continuously according to the contribution ratio.
The idea of a Launchpad for large models sounds novel at first, but essentially, it's about using financialization methods to drive product development.
Once a model is assetized, trainers gain continuous motivation for optimization. Furthermore, if profits can be distributed continuously, the ecosystem develops the ability to be self-sustaining.
The benefits of this approach are undeniable. For instance, the recently popular nof1 large model trading competition currently only has general large models participating. There are no fine-tuned, specialized large models competing, primarily due to a lack of incentive mechanisms. Excellent specialized models usually prefer to operate discreetly for profit and wouldn't be exposed publicly. But if they carry assets, the significance changes entirely. Such large model Arenas become public stages for showcasing strength, and competitive performance would directly impact the asset performance of the large models. Doesn't that just open up a world of possibilities?
Of course, currently, FLock has only proposed a direction; it hasn't been fully implemented yet. The specific similarities and differences between issuing assets for models and issuing assets for Agents remain unknown.
But one thing is certain: How to ensure that model calls for asset-issuing models are based on real demand rather than artificial inflation? How to effectively ensure Product-Market Fit within vertical scenarios? These are all challenges. It should be said that the issues encountered during the wave of Agent application token launches will likely be present here too.
I'm just very curious, what different approaches might emerge in this direction of creating Launchpads for Models?
