Original Title: AI Trends in the Trenches
Original author: @Defi0xJeff Original text compiled by: zhouzhou,

Editor’s Note: The AI Agent project has attracted short-term attention due to the Genesis pump, but institutions are more focused on decentralized AI infrastructure like $TAO and $GRASS. Market trends indicate that teams with strong narratives and technology have greater long-term potential. The strategy is to trade the Agent projects for short-term gains, rolling profits into DeAI infrastructure, betting on teams that can tell a good story and also deliver on technology.

The following is the original text (for the sake of readability, the original content has been reorganized):

Since the market bottom of the AI agent sector (with a market cap of about $4 billion) about two and a half weeks ago, the market is currently in a complete @virtuals_io bull market.

That's right... Currently, only Virtuals are rising. This market trend is very similar to the time when Virtuals just launched the proxy tokenization platform last year from October to November. At that time, it established itself as the pioneer in "tokenized issuance" among AI projects and was the first legitimate player to provide a top-tier issuance network (as long as you are willing to do a "fair launch").

This time it's different: they have just launched a new feature: "Genesis Launch" — a fairer distribution method that also rewards early supporters. This leads us to the first trend we want to discuss:

Fair Launch / Gamified Launch Platform

From the previous degenerate players casually investing in all sorts of random tokens on pumpfun, to now almost guaranteed 5-10 times returns in brand new projects.

The underlying mechanism relies on the introduction of a "points" system, which better aligns the interests of all parties involved. Additionally, with a fixed market value and a fixed supply launch model—each project starts with 112,000 $VIRTUAL (approximately $200,000 FDV). Each participant can claim up to 0.5% of the token share based on the points they accumulate. Points can be earned through various means: holding/staking $VIRTUAL, holding mainstream proxy tokens (there is an official list), and even being actively engaged in discussions about Virtuals on social media platforms.

Recently, Virtuals has also added a "cooling-off period" mechanism to restrict quick selling behavior, further enhancing the appeal of the Genesis Launchpad—making you think twice before selling.

Currently, the "Genesis Launch" platform has achieved remarkable success, with @BasisOS being the most successful example so far, where participants have reached returns of up to 200 times. Since then, new projects launched on Virtuals have almost all started with a return rate of 5 to 40 times.

It is also due to the explosive success of Genesis Launch that funds and attention have once again focused on the Virtuals ecosystem, raising the valuation floor of almost all proxy projects on the platform.

However, despite the market enthusiasm being reignited, "the scarcity of quality projects" remains one of the biggest challenges in the Virtuals ecosystem—this leads to the second trend:

2. Popularity and trading structure overshadow the quality of the project itself

As a speculator/trader, as long as you understand the structure of the Genesis Launch, you can make money even from projects that are average or have mediocre teams.

Because projects starting with a valuation of 200,000 USD are very likely to pump first and then dump, especially when you know that there are no "insiders" in the project and it hasn't gone through a private placement round.

As a project team, if you have a relatively novel idea, you can definitely issue the token first and discuss later.

There’s no need to create a product first, no need to figure out who the target users are, and no need to verify whether there is real market demand, let alone revenue and user growth. Just create the maximum hype and dive in (of course, having a demo is a plus, but it doesn’t matter if you don’t have one, haha).

As long as the project meets the basic requirements (a decent whitepaper, a coherent product concept, and a team that doesn't seem too underwhelming), it can basically take off smoothly at the Genesis Launch of Virtuals.

For investors, the most important lesson here is to view these launches as short-term speculative opportunities, not as medium- to long-term value investments.

Because nine times out of ten, these so-called "AI projects" are essentially just shitters packaged to look like AI. I even specifically mentioned a typical example on my Substack, feel free to check it out if you're interested.

But precisely because there are too many shitter projects now, the opportunities for truly high-quality projects have instead emerged—whether in the AI sector or in other directions outside of AI. This also leads to the third trend:

3. Potential Opportunities for DeFi Projects on Virtuals

Two months ago, I talked with the @logarithm_fi / @BasisOS team. Their product is somewhat similar to Ethena, and it focuses on something like a "Delta neutral strategy", but without the stablecoin component.

I got to know this team when they were working on Logarithm, back when they were still developing the narrative for LPDeFi (a product that uses Uniswap v3 liquidity for strategies). I really like this team, so I offered them some suggestions on token economics, launch plans, and so on.

To be honest, I didn’t think their coin could rise this much, because while their DeFi products have fundamentals, the so-called "AI" part is actually very early stage. But it turns out that none of this really matters—this project still managed to dominate the entire market.

From the success of Basis, it is clear that the Virtuals platform now provides a significant opportunity window for DeFi projects.

Even if you do not have the kind of "token incentive TVL" mechanism of traditional DeFi projects, as long as your product itself is solid and your logic is strong, combined with the traffic and sentiment of Virtuals, the TVL will naturally increase.

So if you have a mature DeFi product and want to use AI to improve user acquisition efficiency, or optimize the product's structure itself with AI technology, feel free to reach out to me. I would be very willing to brainstorm together.

In addition to the "getting market attention" layer, a new experimental gameplay is now starting to emerge on Virtuals, much like the wave of Ethereum shitter tokens in 2023-24 - which also leads to the fourth trend:

4. Use the protocol income generated by "trading volume" as a growth engine

There was a time when many small Ponzi DeFi projects emerged in the market, crazily experimenting with various token models: for example, charging a 1-3% fee on each transaction, using this portion of income to enlarge the project's treasury, supporting their Ponzi DeFi strategies, and then returning the profits/dividends to token holders.

At that time, there weren't many projects on Ethereum, and users were "wealthy and daring to take risks". With this approach, project teams could earn a six-figure or even seven-figure income in just one to two weeks.

Now, we are starting to see similar things happening in the circle of AI agents.

Each transaction on Virtuals incurs a default transaction fee of 1%, of which 70% is returned to the project team. Some other Launchpads also have a fee mechanism of 1-2%, with a return rate ranging from 70% to 100%.

For example, @Squidllora, which recently launched on the @autodotfun platform (backed by the smart support from @AlloraNetwork), uses this portion of the creator fee to expand its treasury, and then uses this money to trade mainstream coins. The strategy relies on the inference model provided by Allora (if you're not familiar with Allora, its positioning is somewhat like a "purely financial version of Bittensor" — scientists compete to see who can create the strongest crypto prediction model, covering various time dimensions).

A portion of the profits earned from trading will be used to repurchase the $SQUID token.

The advantages of this model are very obvious, especially suitable for projects with sufficient funds that do not rely on transaction fees to support their teams.

The project team can use the AI agent token as a marketing tool and user acquisition tool to gain attention, accumulate fee revenue, and thereby launch a whole new set of AI experiments.

However, looking back at the entire battlefield of AI agents, apart from the Virtuals ecosystem, other platforms basically have nothing competitive. Many teams, even with open development and continuous delivery, still haven't seen any growth. This also leads to the fifth trend:

5.Virtuals dominates, with almost no growth from other platforms.

Due to the momentum of the Genesis Launch, the valuation bottom of the Agent project on Virtuals continues to rise. However, it is important to note that the emergence of these new projects does not indicate a significant improvement in fundamentals or technology; the real catalyst is the massive optimization of trading structures— in other words, more people are willing to play the games on Virtuals now because they know they can make money.

This trend may continue until the prices of existing projects reach a temporary peak / local ceiling.

Once the price stops rising, part of the attention will naturally begin to overflow and shift to other ecosystems, such as @CreatorBid, @arcdotfun, @autodotfun, especially those projects with low market capitalization but significant fundamental improvements (new features, new products, new partnerships).

Currently, CreatorBid (CB) and Arc are the two ecosystems worth paying attention to. There are still several undervalued projects that have not started to rise: for example, the regular 3-4 targets on CB that are integrated with subnet or focus on Bittensor product development;

Or some handshake projects on Arc that are directly related to Ryzome are also worth keeping an eye on.

The best way to secure positions is to layout in advance at a low level before everyone notices the value of these tokens.

6. For institutions, AI investment opportunities remain very limited

Despite the significant price increases of major AI agent ecosystem tokens like $VIRTUAL and $AI16Z, many institutions remain on the sidelines for a simple reason: these rapidly rising tokens are only suitable for retail investors / degens, with thin liquidity and weak LP structures (especially on Virtuals).

It is precisely for this reason – the lack of mature liquidity infrastructure + the rising interest in decentralized AI – that institutional funds are beginning to shift towards investing in AI infrastructure, Agent L1, and public chain-level AI laboratories, rather than the current batch of short-term speculative Agent project tokens.

You may be curious about what these institutions are buying?

$GRASS

$TAO (including subnet)

$VANA

$FLOCK

$PROMPT (maybe)

There is still a batch of projects that have not been launched but have a clear moat, such as @NousResearch, @PluralisHQ, @PrimeIntellect.

Most of these projects are focused on the Web3 transformation of truly high-performance models and the decentralization of model ownership. They are not just the "shell" of a decentralized GPT, but rather involve genuinely complex infra + model layers. The barriers to entry are high, it’s difficult to understand, and no one knows how to invest in them.

How to better position based on these trends?

My own strategy is very simple: gradually roll the profits earned from short-term AI Agent projects (especially those with weak teams but strong Genesis rallies) into the DeAI Infra project.

It is important to understand that these infrastructure projects will not experience short-term explosions, as most of them do not have consumer products and are being built slowly at the Infra layer.

But just as OpenAI, Grok, and Anthropic have suddenly evolved in areas such as everyday tasks, real-time retrieval, and code generation, the Web3 model will also experience a qualitative change at some point—truly capable of handling native tasks of Web2 / Web3.

Should we then heavily invest in DeAI Infra?

No, it's not. Crypto is essentially a narrative and distribution-driven market.

Distribution ≫ Technology, this is a game of "90% storytelling + 10% modeling."

As long as a project knows how to tell a story, has good UI/UX, understands how to launch, how to design a token mechanism, and how to attract and retain users, it can thrive.

Therefore, my core investment logic remains: invest in teams that understand the dual driving forces of "narrative + technology."

It's like how Web2 VCs invest in some vertical SaaS; although the underlying models may be open-source, the product barrier is still maximized through their own data and operations.

I believe this logic still holds in the short term in Web3, especially now that hype and community are the main driving forces, and things that are easy to understand tend to sell well.

The market is also moving in this direction now:

·Genesis Launch is becoming an "industry default"

·More and more teams are starting to collaborate with Infra projects, not just for hype, but also to actually use each other's technologies.

While engaging in distribution, we are also developing real AI technology, where narrative and practical work begin to merge.

Quick Summary

The Genesis Launch mode of Virtuals is dominating market attention and returns;

Most projects hype > technology, essentially still short-term trading;

DeFi projects with fundamental value have found incremental opportunities in the Virtuals ecosystem.

Trading volume drives creator income, sparking a new wave of experimental fervor;

Virtuals is currently leading, but other ecosystems like CreatorBid and Arc may soon catch up;

Institutions are still off the market, with more betting on decentralized AI Infra;

The optimal strategy is to roll the short-term profits of the Agent project into DeAI Infra and bet on teams that understand both narrative and technology.

"Original link"

：