After the burst of the AI Agent bubble, where are the next trends and projects?

Original Title: "Post-AI Agent Bubble: Where's the Real Value in Web3 AI?"

Author: ++0xJeff++

Compiled by: Deep Tide TechFlow

Overview

AI agents have rapidly grown to over $20 billion in just a few months, but have also collapsed just as quickly. However, this field is maturing. Infrastructure, decentralized AI, and practical applications are taking over. How the next wave is forming and why it’s worth paying attention to.

In the fourth quarter of last year, we witnessed the rapid growth of the "AI agent" industry, scaling from zero to over $20 billion in just a few months—ranging from the interesting, charismatic, entertaining, and almost absurd "agents" to financial agents promising to change the world through trading and investment, making you very wealthy. And not just agents that make you rich, but many investment DAOs… those investing in other agents (or agent) DAOs (3,3).

From Hype to Infrastructure

We all understand that in a new industry (accompanied by new catalysts like Web2 AI, Trump's election supporting crypto and AI), people do not care about the fundamentals. Anything that generates a lot of discussion, looks highly hyped, and has a cool demo can easily surpass a market cap of $100 million.

++@virtuals_io++ became a standout in this ecosystem, controlling market promotion, attracting builders' attention, telling the best stories, and weaving the best narratives. This attracted builders to launch projects on Virtuals while also drawing retail traders' attention for hype.

Then, ++@elizaOS++ emerged, taking a different approach—open-source AI, providing tools for any developer wishing to create "gold-digging" agents. A massive movement formed around this idea, with adoption rates skyrocketing, and the number of stars and forks on GitHub increasing rapidly (and still growing).

The valuation of major Virtuals grew to over $5 billion, with Eliza accounting for about half at its historical peak, while many other interesting agents reached 8-9 figure peaks, such as AIXBT reaching a billion dollars. Of course, the situation is very different now, with newly launched agents averaging trading prices between $3 million and $10 million. Older well-performing agents average trading prices between $10 million and $50 million. Valuation caps have compressed, with total market cap shrinking from $20 billion to a range of $4-6 billion.

Infrastructure Momentum and Web2 Acceleration

The market is now focused on "pure fundamentals," with people more concerned about infrastructure and decentralized AI, especially as AI models in Web2 continue to accelerate at a rapid pace—Meta's Llama, OpenAI's GPT, Grok, DeepSeek, and Alibaba's Qwen are releasing new improvements and optimized models every month. You can see that ChatGPT's image generation model immediately created a viral "Miyazaki-fication" trend upon release.

In addition, due to the enhanced capabilities of AI models, the consumer layer in Web2 is advancing at a faster pace, making previously impossible things possible—Lovable, Bolt, Cursor, and Windsurf enable developers to release more products faster. Agent workflows and AI agents are ubiquitous. The entry barrier is lowered, and the switching cost for users is almost zero. If you don’t like one app, you can easily find a competitor's service or product with a better price or UI/UX.

Awakening of Data Ownership

Amid all this, many people began to ponder: "If there are so many agent applications utilizing centralized technology, who owns my data? Where does my data go? If I discuss something private with AI, will it keep it confidential? Or will it flow elsewhere?" This is particularly important because ++@OpenAI++'s recent update allows ChatGPT's memory feature to reference all your past chat records to provide more personalized responses.

Bro… this sounds cool and could spark a wave of personalized AI agents, co-drivers, personal assistants, therapists, companions, etc. You can imagine the implications when your data is owned or controlled by others.

The Rise of Decentralized AI (DeAI)

I made some predictions last year, one of which is that decentralized AI will emerge in the second quarter of 2025, with infrastructure enhancing confidentiality, transparency, verifiability, and ownership, thus gaining more adoption and attention as the demand for these features becomes stronger.

There are three independent trend sectors, with many trends interweaving or intersecting between them:

• Web2 AI venture capital trends (YC companies launching vertical agents, a16z positioning future consumer trends through its arguments, Perplexity launching its AI fund)

  

• Web3 AI venture capital trends (DeAI infrastructure investment, distributed training, inference networks, etc.)

  

• Web3 AI retail trends (AI agent ecosystem, consumer agents, AI consumer applications)

  

Web2 vs Web3 AI: A World of Difference

For Web2, the total addressable market (TAM) is significantly larger than Web3, as many businesses seek to transform or optimize their operations through AI, improving workflows to generate more leads, more conversions, more sales, retain more customers, reduce management costs, and operate at a higher level. Many businesses are looking for solutions that can address highly specific pain points within their domains.

This optimization demand has attracted many young startup founders to seek better ways to introduce AI agents to improve workflows. Compared to traditional SaaS, the solutions provided by AI agents can significantly save capital or generate more leads. This allows agent startups to charge higher subscription fees for their usage (which is why we see many startups reaching 7-8 figure annual recurring revenue in just a few months).

For Web3 venture capital, the trends here are very different, as blockchain provides the perfect layer for decentralized AI (DeAI), such as verifiable/immutable transaction trails, trustless environments, decentralized computing, and trust-minimized AI reasoning and training. In short, the future direction is to let people know how their data is being processed, understand the thought processes of AI, own their data, own the models, own the use cases, and be incentivized to share (without censorship), etc. Web3 venture capital has been investing in these futures.

Why Retail Loves Agents (Even if They Don’t Fully Understand Decentralized AI Yet)

For the Web3 retail sector, decentralized AI (DeAI) is very difficult to grasp, as it requires you to learn a lot of terminology and understand important content (which sometimes feels like an alien language). This is why the retail sector tends to choose the easiest-to-understand things—starting with "Web3 AI agents" that can chat, be funny, and do some entertaining activities.

As the retail sector continues to stay in this industry, they gradually realize that this is not enough to create sustainable value for users (yes, many AI agents are useless and lack creativity). This realization (combined with market instability) has prompted market consolidation, with useless agents gradually disappearing while useful agents continue to survive (albeit with significantly reduced valuations).

People are beginning to recognize the need for a core AI product with practical use cases. This realization has prompted teams to either develop real AI products or collaborate with AI companies that possess real technology, such as @AlloraNetwork and ++@opentensor++ (Bittensor).

This shift has two benefits:

1. It gives people a better understanding of the infrastructure they are less familiar with.

2. It provides AI agents with practical use cases to showcase to the community.

Before this shift: Agents had basic skills/use cases (chatting, post analysis)

After this shift: Agents have advanced useful skills (AI-driven betting, trading, liquidity provision, mining, etc.)

Agents like ++@AskBillyBets++ and ++@thedkingdao++ have become ideal agents to showcase the Bittensor subnet, bringing cool technology into the mainstream.

The Bittensor Ecosystem

What I find interesting about the Bittensor ecosystem is that it is a decentralized AI-filled ecosystem where you can invest. Nowadays, most decentralized AI (DeAI) can only be invested in by venture capital or behind-the-scenes strategic investors, as it is still in the early stages, and many projects have yet to issue tokens.

But Bittensor allows anyone to hold their ++$TAO++ and stake it to the subnet they wish to support, thus converting it into the subnet's alpha token (immersing in decentralized AI projects).

I have publicly shared my fatigue with cross-chain and trading experiences, but the technology, products, and atmosphere there, especially ++@rayon_labs++, are truly impressive.

The reason I like Rayon Labs is that they are building products optimized for consumer-facing UI/UX. Given the nature of dTAO, the market determines the emission and pricing for each subnet, making it increasingly important for each subnet to build products that are easy to understand and master.

Rayon has many cool subnets (the coolest might be Gradients, which is an automated machine learning platform where you can easily train models), and even cooler is their latest flagship product, the Squad AI agent platform, where you can create agents through drag-and-drop block building (node builder style, similar to ++@figma++ for AI agent creation).

Final Notes

I am still in the early stages of my journey into Bittensor, and I will later publish a dedicated article sharing the interesting things I discover and how to leverage opportunities within it. If you wish to learn about other trends and changes in the market, please check out this article.

Personal note: Thank you for reading! If you are an owner or researcher of a Bittensor subnet, building or researching cool things in the Bittensor ecosystem, I would love to connect and learn more.