Original author: 0x Jeff
Original translation: Luffy, Foresight News
The crypto market has experienced pullbacks again and again, and liquidity is becoming increasingly thin. The new AI agent project that was successfully launched recently has a peak market capitalization (MC) of about $10 million. Success here means that the project has achieved product-market fit (PMF), can provide value to real users, and has generated (or is about to generate) revenue.
This is in stark contrast to 3-4 months ago, when AI agent projects that achieved product-market fit were valued at up to $100M+, especially those that positioned themselves as “agent + framework/launch platform tokens.” For example, $AVA, a 3D agent, also captures value from its own launch platform and the projects it supports through its audio-visual layer.
Old Model: Agent as Framework
The old model was to launch a proxy project to demonstrate its functionality, attract demand from developers who wanted to build their own proxies, and require these developers to hold/burn/use their own tokens to gain access to the framework. Whats the problem? The crypto Twittersphere places an excessive premium on framework tokens, and these framework proxies are all the same and lack differentiation. In most cases, they dont even have an actual product, but just talk big on Twitter to drive up the token price.
The first generation of AI agent products were primarily conversational agents. This is unique to crypto because we value community building more than other fields, similar to the founder-led marketing model (founders gain attention through publicity). It seemed like a good idea to have agents promote on Twitter to increase attention for the project. When this model first appeared in November 2024, it really worked. But now, there are 420,690 agents constantly promoting all over the world, most of which are simple, repetitive, and frankly, annoying.
New Model: Agency as a Business
The old model has been ruthlessly eliminated by the market. If you still want to build an AI agent project now, here are your ideas:
Launching an agency project means you will be running a startup and managing up to three products at the same time: the core product, the token, and the agency.
1. Core products (actual business)
Your core product should solve real problems and shouldn’t just be a conversational agent, but to be a real product you need to be valuable to your users.
Example:
Predictive models that improve betting odds and help users win more in sports betting (e.g., @AskBillyBets).
Cryptoasset prediction models that optimize trading, minimize impermanent loss, and maximize returns for liquidity providers (LPs) (e.g., @Co d3 xOrg, @gizatechxyz, @Almanak__).
An AI agent research search engine that aggregates insights from top quality information sources such as Cookie, Kaito, Nansen, Messari, Aixbt, Coingecko (CG), Dexscreener and Bubblemaps to assist in investment decisions (no team has yet successfully achieved this, we need a product similar to Perplexity for AI agents).
Before launching a token, building a core product should be every team’s top priority. You need to make sure there is a real market demand for the product and that users are willing to pay for it. Otherwise, you’ll end up in a crypto “death spiral” that’s probably worse than what traditional startups face:
High operating costs.
The cost of acquiring customers using token incentives.
Token price plummets → reputation is damaged → no one cares about your project.
If your token price drops significantly, it becomes a curse. Most people in this space won’t care about your project, no matter how well it’s progressing or how strong your core product is.
Instead of relying on token incentives, focus on attracting customers through your product and devising a profit model that balances growth and revenue generation.
@KaitoAI’s pattern is a great case study:
They have developed an enterprise-grade product: an encrypted search engine focused on social/emotion/narrative, and provide real value by charging users, projects, and the ecosystem.
They launched the Mindshare Dashboard, which has become the standard tool for tracking narratives and trends.
They further launched the Yapper rankings, allowing key opinion leaders (KOLs) to spontaneously share everywhere and use it as a symbol of identity.
They also incentivize users to interact on Twitter with real rewards through non-fungible token (NFT) whitelisting and $KAITO token airdrops.
Although Kaito’s model is difficult to replicate, the lessons learned from it are: first find the fit between product and market, generate revenue, and get people excited about the project before launching the token. Once you have gained attention (heat) and revenue, launch the token to take the project to new heights.
Communication is also crucial. Many projects have great products but poor communication. If no one knows what you are doing, no one will care about your project.
2. Tokens (coordination tools)
We have moved from a “VC token” model to a “fair launch” model that promotes high circulation and low fully diluted valuation (FDV) tokens. But fair launches are not really fair, and each token launch strategy has pros and cons.
If you launch your project’s token with a high circulation and low fully diluted valuation structure, you will not be able to raise money from venture capital (VC) and angel investors (because the valuation is low). However, you can use the token as a marketing tool to increase awareness of your project.
Many teams will launch two types of tokens:
Proxy tokens: used to increase the attention of the project.
Ecosystem Tokens: Raise funds from venture capitalists and angel investors at higher valuations.
But this leads to inconsistent expectations. The community expects to receive an airdrop, and when the ecosystem token is launched, funds flow from the proxy token to the ecosystem token, causing the proxy token price to plummet.
Managing the core product, proxy tokens, and ecosystem tokens while ensuring that each part accrues value is extremely complex and difficult.
Ideally, there should be a token that accrues all value from the core product. Historically, projects that can generate revenue and funnel it back into the token (either through buybacks or revenue distributions) have survived in the long term.
Tokens should be complementary to the core product, not a necessity.
3. Agent (supplementary products)
The agent here refers to the conversational agent built using frameworks such as ElizaOS, GAME, ARC, and Pippin.
While these proxies integrate on-chain and off-chain functionality, they should be complementary to the core product.
Agencies should enhance the core product by changing the user funnel:
Instead of asking users to find and use your product, let agents push the product to users.
This could mean: Using proxies to showcase products directly on Twitter, via text or video.
Use agents as AI assistants to change the way users interact (similar to ChatGPT’s abstract mode).
The agent itself acts as an interface and performs tasks in the background.
There are exceptions. One example is Aixbt, which provides real-time social and sentiment insights from Twitter, allowing users to get real-time quality signals earlier than others. Aixbt has become the number one key opinion leader in the crypto Twitter circle by continuously providing quality information, demonstrating the power of its terminal. In this case, the agent itself is the product.
However, this is extremely difficult to replicate. Most teams should focus on strengthening their core product first.
Cookie DAO is an excellent example of product-first thinking:
Get started with a free AI agent dashboard to acquire users.
Shifting to a freemium model, locking up COOKIE tokens to unlock premium features.
Profits are made by providing an Application Programming Interface (API) to projects and agents.
Introducing Agent Cookie, pushing insights directly to Twitter.
Summarize
In 2020-2021, you needed Solidity programming knowledge to launch a token. Now, platforms like Pump.fun make it easy to tokenize anything.
This changes the way people think, and instead of focusing on building real products, people are just launching tokens. It’s like “garbage in, garbage out”, with money flowing from one garbage project to another.
We need to change this.
To build a sustainable project, we need to run our agency project like a startup. Don’t just think about attracting attention from the crypto Twitter circle or getting funding from venture capitalists and angel investors, but build a product with lasting value, not for the next 6 months, but for the next 6 years.
