Web3 is sick, but the cure is not AI

Author: Zhou, ChainCatcher

In the past year, the pace of breakthroughs in AI technology has far exceeded expectations. From GPT-4o to the emergence of various AI Agent tools, new capabilities have been breaking boundaries every few months.

Especially with the explosive popularity of next-generation AI Agent products like OpenClaw, many in the crypto industry have felt for the first time that AI is beginning to touch their work boundaries.

As a result, a wave of "AI anxiety" has spread throughout the crypto circle.

From macro strategies to micro work methods, from leading institutions to frontline practitioners, many crypto projects and professionals have begun to reassess their value and work models—will what they do be replaced? Does the industry still have a future? These questions have been repeatedly raised over the past year.

1. The projects and institutions panicked first

Responses in products

When AI Agents can directly take over keyboards and mice to execute tasks on computers, users can completely perform on-chain operations through AI without needing to open exchanges or wallets. So, where will the entry value of exchanges be reflected?

Earlier this month, leading exchanges like OKX, Binance, and Coinbase announced their AI product layouts in quick succession.

Among them, OKX launched the Onchain OS AI upgrade, supporting AI Agents to autonomously operate across more than 60 blockchains and over 500 decentralized exchanges; Binance announced it would provide each AI Agent with a Binance-level brain, embedding exchange-level trading intelligence directly into AI Agents; Coinbase introduced the Agentic Wallet for AI autonomous operations.

The logic behind these moves is simple: users can operate with AI, but as long as the infrastructure remains with us, the entry point remains with us.

However, while the excitement is palpable, the actual data is much cooler.

A report from a16z mentioned that previously reported data claimed AI agents completed $24 million in payments within 30 days, but after independent verification and excluding inflated figures, the actual amount was about $1.6 million, shrinking nearly 15 times.

This indicates that the on-chain payment scale of AI Agents is still far from what the outside world has portrayed; it is more about emotions and narratives running ahead of actual data.

David Gan, founder of Inception Capital, stated in an interview with ChainCatcher that not all AIs need Crypto; today, most AI applications are essentially still focused on information processing, content generation, and workflow efficiency. However, once AI Agents transition from being able to speak to being able to act—calling services, managing budgets, initiating transactions, and completing settlements—Crypto will gradually shift from being an option to becoming infrastructure.

Personnel reductions

In February of this year, fintech company Block laid off about 4,000 employees. CEO Jack Dorsey previously admitted that the layoff decision might have been a mistake and stated that the rapid development of AI technology prompted the company to restructure its team of 6,000. However, the backlash came quickly; a month later, Block quietly recalled some employees due to clerical errors and a shortage of personnel in infrastructure roles.

The crypto industry has not been spared either. Recently, Crypto.com announced it would cut about 12% of its positions to advance enterprise-level AI integration; Gemini has reduced its workforce by about 30% since layoffs at the beginning of the year and is introducing AI tools to improve productivity; the Algorand Foundation announced a reduction of about 25% of its staff, and Messari also announced personnel optimization.

The trend of AI replacement has spread beyond the tech circle and has genuinely impacted the crypto industry. The forced implementation of AI tools, reduction in hiring, and reassessment of job necessity have all been put on the table.

However, there are also more objective views. Crypto practitioner Forest Bai told ChainCatcher that layoffs are influenced by both the bear market and the use of AI. For business owners, when the cost-benefit ratio brought by AI is lower than that of humans, replacement will occur; this logic existed even before the emergence of LLMs.

2. The anxiety of individual practitioners runs deeper

Exchanges can mask their anxiety with product actions, but for individual practitioners, it seems that not using AI tools is equivalent to being out of touch with the times.

CZ tweeted that after installing the lobster (OpenClaw), he didn't have to do anything, but all his time was spent adjusting the lobster that couldn't do anything.

This self-deprecation resonated widely, with one crypto builder bluntly stating that non-technical individuals have little chance of succeeding with OpenClaw due to its high debugging difficulty.

For ordinary practitioners, just configuring the required data calls and skill modules has already deterred most people. The excitement is real, but the truly actionable scenarios are still a game for a select few.

Deeper anxiety comes from those who have already left.

In the past two years, some of the smartest and most restless individuals in the crypto circle have appeared in the AI world.

• OpenSea co-founder Alex Atallah founded OpenRouter, directly transferring his integration thinking accumulated in the crypto industry to AI infrastructure.
• Leopold Aschenbrenner, a core member of the FTX Future Fund, now manages an AI investment fund worth billions of dollars, specifically betting on power infrastructure, semiconductors, and computing centers.
• Avital Balwit, also from the FTX Future Fund, is now the chief of staff for Anthropic CEO Dario Amodei, participating in the company's top-level strategic decisions.

Their departure is not because crypto has failed. Or rather, crypto was their training camp, giving them a sense of risk, sensitivity to power structures, and the ability to make judgments in highly uncertain environments. They took these abilities to a larger battlefield.

For those who remain, it is less about the anxiety of being replaced by AI and more about a sense of helplessness. Well-known crypto KOL 0xSun admitted that for most ordinary people, whether the AI industry thrives has little to do with whether they can enjoy the industry's dividends. Most people, due to mathematical ability, education, and upbringing, cannot truly enter the AI industry.

Forest Bai pointed out that the current anxiety among crypto practitioners stems more from the continuous decline in market prices rather than just the disruption of new technologies. Another crypto KOL also suggested that in the AI era, the core capability of individuals will shift from productivity to decision-making ability; AI does not replace people but replaces those parts of people that can already be standardized, replicated, and automated. Those who remain may have the opportunity to redefine their position in this industry.

3. The narrative dilemma is real, but AI is not the remedy

The entire crypto circle's reaction to AI has a common presumption: Crypto needs AI to save its narrative. This logic is understandable but does not hold up under scrutiny.

Using new narratives to cover old problems

The narrative dilemma in the crypto industry over the past two years is real. The vision of Web3 is grand, but the user base that remains has always been limited; DeFi's transformation of the traditional financial industry is still ongoing, the creator economy bubble of NFTs has burst, and the metaverse has long since cooled down.

After each narrative falls silent, the industry needs to find a new story to maintain appearances, and this time it is AI's turn.

David Gan pointed out that many projects make a typical mistake by first assuming that AI and Crypto should definitely be combined, then looking for scenarios to justify this combination. Truly good projects should be such that removing either side would significantly decrease the product's value, and this threshold is very high.

David and Daniil Liberman, co-founders of the decentralized AI computing network Gonka, also stated in an interview with ChainCatcher that most projects attempt to integrate AI at the narrative or token economic level rather than at the infrastructure level. The real challenge is not to add AI to crypto but to build systems that can provide access to open and scalable computing resources.

Chris Feng, CEO of AI infrastructure Axis Robotics, added in an interview with ChainCatcher that labeling products as AI is essentially an attempt to catch up with emotional dividends rather than solving real problems. The key dividing line is simple: are you solving real-world problems?

The true dilemma in the crypto industry has never been about narratives being good enough but about the lack of real application scenarios, inability to retain users, and failure to generate sustainable commercial value.

AI cannot change this. Launching AI features on exchanges will not make more people need crypto; labeling projects with AI will not truly activate on-chain activities.

Where is the real intersection of AI and Crypto?

Of course, some are seriously contemplating the real intersection of AI and Crypto.

As AI Agents move from dialogue to execution, they begin to require autonomous payments and on-chain settlements, and the account systems and KYC processes of traditional financial systems are not compatible; stablecoins and programmable contracts have real utility in this scenario.

Chris Feng believes that Agents need autonomous payment capabilities, verifiable identities, and traceable operation logs, which do not have natural solutions in traditional Web2 systems, and Crypto happens to fill this gap.

This trend is reflected in the data; as the demand for AI agent payments grows, the supply of USDC has rebounded to near historical peaks, with adjusted transaction volumes increasing by over 90% year-on-year, and it is increasingly being used in scenarios outside of crypto trading. Meanwhile, Stripe, Cloudflare, and Google have also embedded on-chain payment standards into their respective AI agent protocols.

Bill Sun, a researcher deeply engaged in the AI field, bluntly stated in an interview with ChainCatcher that Circle's true competitor should be Stripe, not USDT. In the quasi-financial layer, companies that take strategic value seriously, like Stripe, are needed to create truly valuable infrastructure. He also mentioned that the crypto industry has focused too much on short-term retail trading and speculation over the past few years, wasting a lot of meaningful development energy that should have been used on truly valuable things.

The direction of on-chain payments is real, but it may solve AI's problems, not crypto's. As Forest Bai said, although he recognizes stablecoins as a direction for AI payment tracks, their current proportion is very small; stablecoins and blockchain payment networks are existing infrastructures that may not require something entirely new.

David Gan stated that the value of AI to Crypto lies not in providing the industry with another trendy label but in whether it has the opportunity to make the chain truly serve software, machines, and automated execution.

If the industry can seize this opportunity to genuinely enter the machine economy and open network collaboration at the foundational level, it is a structural opportunity; if it merely treats AI as a new narrative to stick onto an old model, it will quickly cool down like past narratives.

Conclusion

Ultimately, the AI anxiety in the crypto circle is a stress response from an industry that has overdrawn its narrative after encountering external shocks. The anxiety is real, but AI is not the remedy.

In the past few years, after each major narrative collapsed, the crypto industry has self-repaired in the same way: finding new stories, creating new concepts, and attracting a new batch of people to enter the market.

This time is no exception; exchanges are intensively launching AI features, practitioners are flocking to test the lobster, and projects are rushing to label themselves, but behind the excitement, the question remains unanswered: what problems can the on-chain world truly solve for real users?

This question, perhaps, AI cannot answer.