AI Agent Investment Review: There are not a hundredfold opportunities, but a lack of decisive action

In the rapidly changing AI market, investors need keen insight and decisive action to seize hundredfold k-fold opportunities. This article reviews the investment dynamics and successes of the past few months. This article is from an article written by Tomb and compiled by TechFlow. (Summary: The Economist shouts: 2025 is the era of "AI agents", but we must pay attention to three difficulties) (Background supplement: Decentralized Finance integrates AI: how DeFAI's three core scenarios promote the evolution of Decentralization Finance) Key points of this article: Review the development of agents and AI in the past few months; Review some of the obvious investment opportunities in the market; Take a break during the New Year period, observe many things from the perspective of a bystander, and deeply reflect on past successes and failures; Next, I will try to share all my thoughts as clearly as possible; Looking back at the dynamics of agents and AI developments over the past few months, AI Agent Review was very satisfied with the performance at the end of last year. From September to early October, we entered a lot of storytelling areas about AI, agents, and infrastructure (Infra) very early. Without overcomplicating the analysis of these trends, it is clear that these are the next stages of the AI supercycle (with more important catalysts to come). This mature judgment comes from a long history of capturing narratives in the marketplace. When you're able to identify these trends, you can invest money quickly without too much hesitation. Even now, I see some people arguing about irrelevant issues such as the encapsulation of ChatGPT, the specific function of an agent, and the failure of a protocol to respond in a timely manner. But do these really matter? We are here to make money, not to prove ourselves right. The key is to be clear about where you are in the market cycle. Each cycle has its own unique rules and strategies. This is why many people hold on to old projects for too long, thus missing out on some hot trends of "simple mode". By "old" I mean not really old, but "relatively old" relative to the pace of rapid development of the industry. Keep in mind that new hotspots emerge every day, and money flows faster and faster — if teams can't articulate their ideas or show a clear roadmap, money goes to protocols that are better communicative and able to get projects off the ground quickly. When I first came into contact with Goat, I realized its potential and followed it very early — and it surprised me how long it took others to migrate to Base or start exploring agents on Sol via the Cross-Chain Interaction bridge. This phenomenon is largely due to the "Mid Curve Effect" (an overly complicated way of thinking) and market panic. If you didn't realize these problems until now, you need to seriously reflect on why. I am very grateful for the past few months because I haven't experienced too many serious losses. I jumped at AI16z at 30 million Market Caps, got on board at aixbt at 30 million Market Caps, and some games and dialogue projects laid out at under 10 million Market Caps. Of course, there are some missed opportunities, such as Zerebro, Fartcoin, Snai, and Swarm Node. But it doesn't matter, it's impossible to take all the opportunities, it's important to reflect on why they were missed and why they succeeded. Quality As always, quality is always the key to victory: high-level developers; Excellent branding; Innovative ideas or unique technologies with a competitive advantage; a cohesive, high-quality community; Clear and valuable information and communication; These are the core elements that we have been following. If two or more of the above are missing, there is a risk. Why? Because there will be other teams that can do better. For example, in the current environment of Virtuals, as long as there is a high-quality team to launch an AI agent or related infrastructure, I will invest quickly at an early stage (i.e., "all in"). Doxxed developers, they've been in the field for over 5 years and have worked on multiple protocols — as soon as I spotted such a team early on, I would have made a decision to get on board. In fact, you don't even need to invest much money to get a decent return. For example, I found out about GEKKO when it first started, when Market Cap was only $4 million. Because it was developed by the Axal team and received investment from a16z – which was enough for me, I decided to invest. As a result, it went from 4 million Market Cap to 40 million, which is a very good profit. I'm sure there will be further rises. Turn $4000 into $40000, which is a year's salary for some. Sometimes, you don't need to invest a huge amount of money, and $2,000 or $4,000 can bring a considerable return. Don't underestimate the potential of these funds. The other one I've been laying out is Acolyt, a very strong team with a similar narrative to GEKKO. As I mentioned earlier, if there are 100 different AI agents that can provide high-quality market intelligence (alpha), the only thing that distinguishes them is identity, user interface (UI), branding, developer level, etc. – no different from other areas of life. That's what I'm always looking for: teams that have been building for a long time — top developers who are passionate, keep tweeting, and keep launching new products. They're the only way my money goes, and I wouldn't think about it anywhere else. Things to avoid Avoid undisclosed one-person development teams, especially when building speculative agent projects, where the risk is high. Another thing to be wary of is the agent project initiated by individual "influencers". Such projects are often abandoned shortly after announcement because developers realize that the responsibility exceeds expectations. I have to admit that there are some narratives that I personally am very optimistic about, but they do encounter some difficulties in their development. This is not to say that these projects are bad, but they are moving forward relatively slowly. It's really frustrating to put money into a project that looks promising, only to find that it doesn't perform as expected, while other projects around it achieve a 2-3x rise in a short period of time. These projects are often of the "slow heat type" or end up being "adjunctive experiments" for other projects. To give a few examples, such as $GRIFT+$REALIS+$OMEGA. I still hold some of them, but practically I should have stopped losing money earlier and shifted funds to stronger AI infrastructure projects like REI and ARC. The potential of these projects is more pronounced and their performance is more stable. At the time, I was investing about $100,000 to $120,000 in these "experimental" projects, expecting them to capture a larger market share. But in retrospect, projects like Griffain with Solana support are significantly stronger, while projects like Realis take more time to mature, perhaps because their ideas are too far ahead of their time, or because the team...