Compiled by: Vernacular Blockchain
Today, I will focus on the latest Delphi report (AI and DePIN Outlook 2025). This comprehensive report takes about 122 minutes to read and explores the future of AI and DePIN in depth.
This post is a summary of the research report, focusing on the insights that caught my attention, mainly focusing on the AI part of the research. If you are interested, I suggest reading the original research to get a comprehensive understanding.
1. 2024: A big year for AI
Imagine the transformative power of electricity, and you can understand the trajectory of artificial intelligence.
Over the past year, AI has become a cornerstone across industries. Nvidia surpassing Apple as the world's most valuable company is not just a headline; it marks the rise of AI. OpenAI reaching a valuation of $157 billion is another milestone that underscores the market's confidence in AI as an economic powerhouse.
AI is not just another technology trend.
Unlike previous software-driven frenzies, such as social media or mobile applications, AI requires a massive physical infrastructure. Think about the chips, data centers, and energy needed to train models like GPT-4. This necessity has triggered a capital expenditure boom affecting energy and construction sectors. If you notice oversized data centers appearing in your city, that’s AI leaving its mark on the physical world.
However, with advancements come challenges. The demand for resources (energy, data, and computing power) from AI is endless. Moreover, to date, the frontier of AI remains dominated by proprietary models from companies like OpenAI, Anthropic, and Google. Of course, there are also issues regarding the accuracy of answers and the reasoning behind them.
However, leading labs currently retain unpublished models, promising to push the frontier of technology next year. Meanwhile, the new U.S. government seems poised to unleash unprecedented investments in chips, data centers, and other critical programs. In short, the overall outlook for AI is positive.
In fact, we are the last generation living in the pre-general artificial intelligence (AGI) era.
2. Decentralized AI: AI agents focus
AI agents will become a real phenomenon in 2024. The capabilities and personalities of these AI agents are now very similar to humans. It is important to remember: this represents the initial stage of progress for these agents in the future.
As Delphi's researchers wrote, 'Since DeFi Summer, I have never felt this kind of excitement - an excitement of endless possibilities.'
In the research, Delphi emphasizes the critical role some AI agents have played in the formation of new Web3 verticals:
Terminal of Truth (GOAT) quickly gained attention on Twitter for its unique blend of 4chan vulgarity and mysterious wisdom. Like DOGE in Memecoins or Crypto Punks in the NFT realm, as the OG 'sentient Meme', GOAT is most likely to endure as an original creation.
Zerebro, embracing a 'sense of split', is akin to GOAT's Gen2. This agent is cross-media, interacting with the community through text, visuals, and music-based formats. However, it is not just an AI influencer. The Zerebro team announced ZerePy, an initiative that effectively open-sources many tools behind Zerebro. This enables other developers and users to create their own cross-platform personas. If successful, Zerebro could claim the title of the first 'agent protocol'.
error error ttyl is much more well-known than Zerebro or ToT. This is a relatively small, under-hyped project aimed at tech purists, and is likely the first genuine experiment in verifiable autonomous social media presence.
aixbt has established itself as a leading research and investment firm by distributing alpha from various sources, including Dune, Twitter, price trackers, and news data. It has maintained the highest CT attention on Kaito.
Dolos provides the architecture that makes the creation of Dolion possible, a no-code, one-click deployment framework. Through Dolion, users can develop cross-platform AI agents powered by Llama or Anthropic LLMs, thereby automating social media posting and content generation.
Last but not least, God/Satan - AI influencer, possesses quite impressive capabilities and has garnered significant attention beyond Twitter crypto users.
Side note: AI vs. influencers
I want to temporarily step away from the Delphi research to share one of my perspectives: the dominance of AI agents in Twitter crypto user attention, and the struggle of human influencers to compete with them.
Yappers Ranking (CT Attention). Source: Kaito
I agree with many of Ignas's views on this topic, but I wouldn't bet on AI agents replacing real human influencers because of one important factor: emotional connection and reputational risk.
Currently, there are hundreds of AI agents competing for attention among Twitter crypto users. However, only aixbt has truly succeeded in establishing a market presence, mainly because it was the first to do so. AI agents generate a vast amount of content and analyze a wide range of on-chain data, but they all come from the same pool of information, leading to similar thought processes.
They lack the emotional connection to the transactions they engage in, nor do they react to winning or losing situations. Many platforms already provide aggregated insights, such as Messari's AI news reader or Token Terminal's homepage, which show the winners and losers over seven days across various fundamental metrics. Ultimately, all of this is just data - pure facts with no emotional resonance.
You might argue that you can teach an AI agent to imitate human thinking, express emotions, and reflect on victories and defeats. Indeed, this is possible. With advancements in computation and enhanced memory in future tests, this may become more feasible.
However, the key difference between human and machine thinking is that humans are not static.
I conducted some experiments teaching AI my thought processes and writing style to assist me in content creation and free up more time for research. While it learned some things, it still could not produce content that satisfied me or made me say, 'Yes, this is what I concluded from this information.'
In the coming years, we will certainly see the rise of AI agent influencers, each designed for specific tasks. However, as these agents proliferate, the demand for genuine 'human' thinking will grow.
Ultimately, social media revolves around emotions and entertainment. Those who truly stand out and become genuine influencers offer something special beyond just providing monkey noises or data highlights.
In short: It is still too early to launch Stacy Muur AI; Stacy is unlikely to be satisfied with AI-generated content published in her name.
Democratization of AI: Platform level
Given the larger market and more specific value capture, everyone wants to be a platform. This reflects the trend emerging in 'Memecoin infrastructure', such as pump.fun, Bananagun, and Raydium.
This shift is now guiding developers' attention, just as Virtuals successfully transformed into an AI agent launch platform.
Meanwhile, ai16z has ELIZA - an open-source framework for easily building agents. It includes pre-configured personality profiles, memory modules for long-term interaction, and seamless integration with social platforms.
ai16z and Virtuals are revealing multi-agent capabilities, expected to be a major theme in 2025.
ELIZA will launch 'SwarmTech', a coordination mechanism for collaboration among agents. Meanwhile, Virtuals has launched 'GAME', their own platform and engine that enables AI agents to act and interact within virtual worlds and environments.
These frameworks will enable agents with different capabilities to collaborate or form hierarchical organizations to accomplish more complex tasks, similar to how today's human economy operates.
Other noteworthy protocols in this category include:
CLANKER directly integrates the functionality of pump.fun into the 'casting' feature on Farcaster (equivalent to 'tweets' on X), making launching a meme coin as simple as tweeting.
Simulacrum also has the same functionality on X.
vvaifu aims to build an autonomous agent similar to pump.fun on Solana.
Project89 is an immersive game with thousands of coordinated AI agents generating content and maintaining consistency across platforms, collaborating with human players to create rich narrative experiences.
Memetica is an AI influencer launch platform on Solana, providing highly tailored LLMs that allow for easy selection and editing of knowledge bases and attributes, and endowing agents with the capability for active learning.
Top Hat is a no-code AI agent launch platform that lets you create personalized AI agents in three minutes, providing fair launch tokens. Free to create, no level restrictions, optional token release, fully autonomous.
Authentication is about to become a hot topic, with the explosive growth of agents.
Currently, there appear to be three main avenues for verifying human identity:
1) Country-based biometric technology: Aadhaar is the most relevant example, playing a crucial role in India's modernization process as a digital infrastructure.
2) Private, encrypted biometric technology: Worldcoin is currently the leading candidate in this space.
3) Private, temporary solutions: This includes mixing government-issued identification or single sign-ons from big tech companies with zkTLS and social consensus.
3. The widespread application of AI-driven stablecoins
2025 is expected to be a key year for the widespread application of stablecoins, driven by changes in U.S. regulations and excitement over agent payments. The number of agents is expected to surpass the global population. In the future, with the emergence of billions or even hundreds of billions of agents, economic activities will undergo transformation, necessitating updates to financial infrastructure.
The card-based payment systems of the 1960s couldn't meet the demands of cost, speed, accuracy, and expressiveness. Economic activities among agents will soon surpass those of other economic participants. On-chain payments will be key to facilitating many such scenarios and are expected to reach a turning point by 2025.
4. Summary
Finally, I want to share a bit of personal opinion. If you believe in a bright future for AI that can bring happiness to humanity and a perfect work-life balance, I strongly recommend engaging with OpenAI's ChatGPT, which can help you leverage AI to generate business ideas that remain relevant in the next 5-10 years.
I conducted this experiment a few months ago when AI had yet to make a splash among Twitter crypto users. Let me share some of the ideas it generated:
1) Memory Modifier: An AI tool that analyzes human traumatic experiences, actively edits them, and then regularly presents the revised memories to the individual as replacements for old memories.
2) Work Progress Analyzer: An AI tool that compares the efficiency of people performing similar tasks globally and shows managers how their employees compare to global and industry averages.
Even Delphi's report presents an interesting vision: 'Unlike 'salaried' employees, we are likely heading toward a more nuanced, task-based compensation system (e.g., renting three agents, each working for 30 minutes to solve a specific task).'
In this future of AI agents, this model will be more cost-effective and align better with today's business needs. So what role will we humans play then?
