Mentals.ai (𝕄³)

This is a great thread on the AI agents space by @hosseeb. Here’s my summary with some additional comments (not trading advice): 1. The “AI agent” craze will continue probably throughout 2025. But it will die off eventually. Hype cycles come and go, and many crypto narratives fade over time. While the current AI agent crypto projects are seeing a surge in interest, many are likely to fail in the long run. Memes might have a slightly better survival rate because they rely on strong communities. For products, however, a strong community isn’t enough—products need adoption and genuine business value. Without these, they face an uphill battle to stay relevant. Most AI crypto projects lack both, as explained below. 2. These things are not really agents. These are chatbots with memecoins attached; they are barely agentic at all besides posting on Twitter. One of the major trends in the AI agent crypto space is the branding of chatbots as full-fledged AI agents. While these chatbots can be entertaining and might even replace influencers in some cases, their revenue models are shaky, and large-scale adoption is uncertain. By mass adoption, we don’t mean the number of followers or interactions—it’s about users paying to use the product at scale. Sustainable revenue models are rare. At best, they might replicate influencer-style partnerships, but even that wouldn’t be enough to support strong tokenomics. 3. And no, it’s not going to be in trading either. AIs will not give everyone their own “trading agent” or miniature hedge fund. While there may be research bots that help users analyze markets or execute trades, these use cases don’t offer lasting viability. Platforms like Mentals allow anyone to build entire teams of AI agents using just natural language, no coding required. This accessibility raises the question: why pay for a trading bot when you can build your own or use one created by others for free? While UX could differentiate some projects, the ongoing commoditization of intelligence means the cost of such solutions is likely to approach zero. 4. So if the big story is not chatbots and not trading bots, what else is there? Here’s my core thesis, which for some reason almost nobody is talking about: the truly impactful AI agents will be software engineering agents. AI agents have the potential to impact every facet of life. They can code, automate tasks, and assist in decision-making. This goes far beyond chatbots or trading bots. As @hosseeb explains, “What is the costly input preventing there from being more applications, more wallets, better infrastructure, better everything? The answer is software.” AI agents won’t just write code—they’ll enable the creation of entire products or businesses. Cost of intelligence? Approaching zero. Cost of marketing? Approaching zero. Cost of coding? Approaching zero. As @sama speculated, we could soon see unicorn companies built entirely by a single founder, powered by AI agents. Overall, the role of crypto in the AI ecosystem remains limited today. Does crypto bring something truly essential to most AI projects? In many cases, the answer is no. Current examples include: • Bots with memecoins • Trading bots • Research bots • Gaming NPC AI agents Can these products exist without a token? Absolutely. Does the token add value? Beyond community financing, its role is often negligible. Next, @hosseeb discusses what actual Crypto x AI could mean. 5. Truly autonomous agents will use crypto to pay each other. For autonomous agents that cooperate and require monetary exchange, crypto and smart contracts are a natural fit. They offer decentralized, trustless transactions, making them particularly useful for this type of ecosystem. This is the direction @mentals_ai is exploring, though the concept of fully autonomous agents is still in its early stages. It may take years for these systems to mature and achieve significant adoption. Another promising area is the monetization of unique, proprietary data that powers AI agents. Data that provides agents with specialized capabilities could create new marketplaces. For example, owners of such data could rent out AI agents to others, enabling agents to perform tasks autonomously—imagine an Upwork, but driven entirely by AI agents. While crypto could simplify secure and efficient transactions in this model, it isn’t strictly necessary. However, it could enhance transparency, decentralization, and scalability. 6. We will also see more and bigger scale experimentation around decentralized training and inference. This could become a major area of growth. Decentralized training and inference offer transformative potential, though the focus here leans more toward edge infrastructure than blockchain itself. Blockchain remains a useful tool for transactions in such systems, but its necessity remains unproven. In the meantime, sure, trade AI-flavored memecoins. But real agents are going to have a lot more impact than tweeting and pumping their own tokens. 👏 What’s your take on the AI agents space? Let’s discuss in the comments! 👇