After the number of developers was halved: Crypto is not dead, it has just handed over talent to AI

Author: Xinyang & Ethan, IOSG

In 2026, the GitHub activity curve of the Crypto open-source community completed an astonishing "bottoming out." The monthly active developers dropped from 45K at the peak in 2022 to about 23K. This halving of surface data sparked discussions on social media about "narrative exhaustion." However, when we dissect this curve's cross-section, what we see is not a contraction of the industry, but a profound "de-leveraging of talent."

▲ Data Source: Electric Capital Developer Report, based on Crypto Ecosystems Github

1. Who left? Who is still here?

The main ones who left are newcomers. In February 2024, the number of new developers reached 5,462 in a single month, followed by a sharp decline, with a 52% attrition rate for those who had been in the industry for less than a year. Most of these individuals entered during the bull market, working on NFT minting contracts, forking DeFi protocols, and creating front ends for new L2s. These positions are highly dependent on market enthusiasm; once the excitement fades, projects cease operations, and the roles disappear. Data shows that newcomers' code contributions never exceeded 25% of the total, indicating that they were never part of the core circle of the industry from the start.

▲ Newcomers surged in during the bull market and left during the bear market; Established devs (with over 2 years of experience) reached a historical high during the same period.

Data Source: Electric Capital Developer Report On the other hand, developers with over two years of experience increased during the same period, reaching a historical high and contributing about 70% of the code volume. Maria Shen, GP at Electric Capital, stated directly: "When we look at the established developers group, it is growing and looks very healthy."

They stayed not because they had no other options.

Technically, the core work in crypto now involves infrastructure development that typically requires years of accumulation to understand: protocol layer development, security audits, cross-chain architecture. These tasks require years of experience to truly engage with and cannot be eliminated by market fluctuations.

Economically, many veterans hold unvested tokens, governance powers, and equity relationships within protocols. Their accumulation in this industry has created real barriers and returns. From an ecological distribution perspective, they are voting with their feet: Bitcoin developers grew by 64.3% over two years, Solana by 11.1%, while Cosmos declined by 51.1% and Polkadot by 46.9%. Veterans are concentrating on ecosystems with real users and revenue, leaving projects that still rely on narratives to sustain themselves.

▲ Source: Coincub Web3 Jobs Report 2025

Data Source: Web3.Career Changes in job structure also confirm the same thing. In 2025, the highest proportion of new Web3 positions was not developers, but Project & Programme Management, accounting for over 27%. This is counterintuitive for an industry known for being technology-driven, but the underlying logic is not complex: the industry has transitioned from a construction phase to an execution phase, with over 100 chains needing integration. Institutional clients have different compliance and security requirements, and DAO governance needs to balance the interests of diverse stakeholders. This is not traditional project management but involves coordination and judgment in an environment where rules are still being formed.

On the surface, the industry is shrinking, but the core density is actually increasing. The bear market of 2018-2019 also saw a significant loss of developers, but it was followed by the emergence of phenomenal projects like Uniswap, Aave, and OpenSea, which defined the bull market of 2020-2021. The builders who remain this time have more mature infrastructure, and the AI era has provided them with a larger stage than the previous cycle.

2. What abilities do those who remain possess?

What special abilities has the crypto industry honed in builders? To answer this question, we need to return to the underlying principles of blockchain. Between the cycles of bull and bear markets, this industry has always operated under the same foundational rule: code is law, and execution is final.

In 2016, the DAO incident saw an attacker exploit a recursive call vulnerability to steal $36 million. The code had no bugs; the logic executed as expected, but the boundaries were not anticipated by the designers. In 2021, the Poly Network cross-chain bridge was attacked, and $610 million was transferred within hours. No platform could stop it, no institution could revoke it, and no legal clauses could provide recourse. This is a structural characteristic that distinguishes crypto from almost all other industries: the margin for error is zero, and post-event intervention is virtually non-existent.

This environment has forced the development of a set of abilities that are rarely needed in other industries: building operational systems from scratch that strangers are willing to participate in under conditions of missing rules and trust.

This ability encompasses two levels. One is to establish trust from scratch, relying solely on code and mechanisms to make strangers willing to put real assets in. The second is to make judgments under dual uncertainties of technology and economy, without regulatory frameworks, historical data, or industry standards to refer to, yet still design operational systems.

Both levels have concrete validations in crypto. Uniswap has no company guarantees, no KYC, no customer service; anyone can put funds into the liquidity pool, relying only on trust in a few hundred lines of code and an economic mechanism, achieving daily trading volumes in the hundreds of billions of dollars. MakerDAO has no central bank endorsement, no deposit insurance, and purely relies on on-chain governance and collateral mechanisms to maintain the stability of DAI. During DeFi Summer, it was even more extreme, with no regulatory framework, no auditing standards, and no historical data to refer to; builders designed AMMs, lending protocols, and liquidity mining, achieving billions of dollars in TVL from concept to execution in just a few months. This ability manifests differently among builders at the protocol layer, application layer, and governance layer, but the underlying principles are the same.

The AI era is creating a structurally similar problem. The decision-making process of models is opaque, and the output results cannot be independently verified. AI agents are beginning to autonomously execute trades and allocate funds, but the accompanying rule systems and constraints do not yet exist. Large model companies control both the models and the evaluation standards, leaving users without effective verification means. Computing power is highly concentrated in a few top firms, leading to monopolistic pricing when demand surges. These issues point to a core problem: the trust issue of autonomous systems, which is replaying in the larger scale process of AI.

Crypto builders have been dealing with such issues in an environment without external authoritative rules for years; the previous scenarios were on-chain protocols, and now they have shifted to AI. A group of individuals has directly brought the abilities accumulated in crypto into AI and achieved results.

3. How are these abilities being repriced in the AI era?

Cases of transitioning from crypto to AI have become increasingly common in recent years, but upon closer examination, the things they take away are not the same.

The most intuitive path is the direct transfer of hardware and experience. The three founders of CoreWeave, Michael Intrator, Brian Venturo, and Brannin McBee, started mining Ethereum with GPUs in 2017, expanding from one machine to thousands. They shut down their mining business in 2022, and two months later, ChatGPT was released, turning their GPUs directly into AI computing power supply. In March 2025, they went public on Nasdaq with an IPO valuation of about $23 billion, and their market cap peaked at nearly $70 billion.

OpenSea co-founder Alex Atallah dealt with the aggregation and routing of extremely heterogeneous assets in the NFT market, applying the same experience to AI model routing by founding OpenRouter, which served over 5 million developers within two years and reached a valuation of $500 million.

Another type of migration is more noteworthy. NEAR founder Illia Polosukhin is one of the co-authors of the Transformer paper. After leaving Google, he initially aimed to build AI applications using natural language but encountered a practical issue during development: the need to make cross-border payments to data annotators worldwide, most of whom do not have bank accounts, and blockchain technology became the best solution for this payment challenge.

Now, NEAR is transforming into an AI infrastructure platform, focusing on user-owned AI and decentralized confidential machine learning (DCML), allowing users to use AI services without exposing their data. The decentralized architecture experience accumulated in NEAR has become the most difficult starting point to replicate in this direction.

Circle co-founder Sean Neville founded Catena Labs after leaving, positioning it as an AI-native bank, directly transferring his understanding of stablecoin infrastructure to AI agent financial scenarios, with a16z crypto leading a $18 million seed round. Nader Dabit, a senior developer at Aave and Lens Protocol, shifted to Cognition, bringing his experience in building developer ecosystems across multiple crypto protocols into the AI agent tools space.

These individuals are taking away not just GPU hardware or user networks, but also intuitive mechanism design, experience in building developer ecosystems, and the judgment to build trustworthy systems from scratch in the absence of rules. These abilities correspond precisely to the three structural gaps encountered in scaling AI. Aggregation and Optimization of Computing Power Computing power is the most direct bottleneck for scaling AI. Training and inference require a large number of GPUs, with fluctuating demand; cloud vendors are expensive and have long wait times, and companies do not want to stockpile hardware themselves. This problem has two levels: how to aggregate and allocate computing power, and how to use the aggregated computing power more efficiently. Crypto builders have direct transferable experience in both areas.

Hyperbolic addresses the distribution and trust issues. Founder Jasper Zhang has brought decentralized mechanism design into the AI computing power arena: tokens incentivize dispersed GPU holders to contribute idle computing power, but the more core issue is trust.

Why should one trust the computation results provided by a stranger node? The core innovation, PoSP, uses random sampling and game theory to make honesty the dominant strategy for nodes, requiring no full verification, with low overhead, scalability, and reliable results. This mechanism directly migrates from the logic of verifying unfamiliar node behavior in crypto.

MoonMath addresses the efficiency issue. Its predecessor, Ingonyama, focused on ZK hardware acceleration, improving the speed of ZK proof generation several times under extreme computational constraints. Now, it has shifted to the Physical AI performance layer, working on sparse attention acceleration for video diffusion models (LiteAttention), low-rank decomposition for FFN layers (LiteLinear), and acceleration of training backpropagation (BackLite). From ZK acceleration to AI inference acceleration, the underlying capability remains the same: making mathematics run faster under extreme computational constraints. The track has changed, but the accumulation has not been wasted. AI Governance and Incentive Mechanism Design When multiple AI agents begin to collaborate on tasks, how can we ensure that they do not disrupt the overall system in pursuit of their individual goals? Each participant is pursuing its own objective function, and there is no guarantee that the system will operate normally when they are combined, while the execution speed of agents far exceeds the window for human intervention.

This is a type of problem that crypto builders have repeatedly dealt with in DAO governance and tokenomics design: allowing participants with completely different interests to operate in the direction preset by the system without a central authority. The answer provided by crypto is economic mechanisms, where violations incur real economic costs, and the rules are written in code and executed automatically.

EigenLayer has directly migrated this mechanism to the AI scenario. Through a restaking mechanism, nodes must stake assets before participating in collaboration; non-compliance or violations trigger automatic penalties, and the rules are not suggestions but rigid boundaries with real economic costs. EigenCloud extends this logic to verifiable computation and collaborative governance for AI agents, requiring agents to operate within preset limits while pursuing their own goals. Using economic mechanisms to constrain agents is much more reliable than using ethical guidelines.

Autonomous Payment for AI Agents There is also a more fundamental question: how do agents pay? Traditional payment systems are designed for humans; credit cards require account opening, bank transfers require authorization, and every step assumes the operator is human, has identity, and will wait. Agents do not wait; they may initiate numerous requests per second, and each request may involve micro-payments, rendering traditional payment pipelines ineffective in this scenario.

Stablecoins and on-chain rules are the infrastructure that crypto builders have already established, natively supporting programmable, authorization-free, and round-the-clock operations. These three features are precisely the hard requirements for agent payment scenarios; what is missing is a layer of protocol that connects stablecoins to the agent workflow.

x402 was launched by Coinbase in May 2025, activating the HTTP 402 status code and embedding stablecoin payments directly into HTTP requests, allowing agents to complete payments simultaneously while initiating requests, without needing accounts, with settlement in about two seconds. As of April 2026, the x402 protocol has processed over 165 million transactions, with a cumulative transaction volume of about $50 million and 69,000 active agents (data source: x402 Foundation). Cloudflare, AWS, Stripe, and Anthropic MCP have all integrated. Agent payments have already become a track with real traffic.

The three directions correspond to the three structural gaps encountered in scaling AI: aggregation and efficiency of computing power, incentive alignment for multi-agent collaboration, and infrastructure for autonomous payments. These three problems do not have ready-made answers in traditional software architecture, but there are corresponding handling experiences in the crypto industry. The capabilities have not disappeared; they have simply found new carrying scenarios.

4. The New Positioning of Builders: From Contract Writers to Rule Makers for AI

The scaling of AI is creating a functional gap that did not exist before. It is not a gap in technical talent, but a gap for those who can design trust mechanisms in autonomous systems. As the target of services shifts from humans to AI, the role of crypto builders is also being redefined.

The table below compares the dimensional changes of specific functional paradigms:

The core difference between the two paradigms lies not in the technology stack, but in the way trust is established and the logic of rule execution. In the pre-AI era, crypto builders faced human participants, where rules were written into contracts, and the margin for error was zero, but the boundaries of the system were relatively clear.

In the AI-native era, when the interaction objects become autonomously operating AI agents, the problems that need to be solved are: agent behavior is unpredictable, execution speeds far exceed human intervention windows, and the boundaries of the system itself need to be redefined under greater uncertainty. The functional positioning of crypto builders is shifting from "writing secure contracts" to "designing trustworthy mechanisms for AI autonomous systems."

Top institutions' hiring is already reflecting this change:

▲ Q1 2026 hiring trends for core AI/data positions at leading exchanges

Source: Gate Research Institute In 2026, the hiring trends of leading exchanges and institutions clearly reflect this trend: they are no longer simply hiring AI engineers or crypto developers but are looking for individuals who can connect the two sides, understanding both on-chain incentive distortions and governance games, while deeply embedding AI tools into crypto workflows and designing mechanisms that align agents with regulators and users in the long term.

The direction of capital allocation has also reflected this judgment. Paradigm is raising a new fund of up to $1.5 billion, expanding its investment scope from crypto to AI and robotics. Haun Ventures has completed a $1 billion Fund II, focusing on financial infrastructure that merges crypto and AI, particularly supporting autonomous trading and coordination for AI agents, as well as stablecoins and agent-to-agent economic systems.

a16z crypto has completed a $2.2 billion fifth fund (Crypto Fund V), explicitly stating that the fund will be 100% directed toward the crypto field. In the face of the complexity and opacity of the AI era, they will focus on the transparency, verifiability, and decentralization characteristics of crypto applications. According to PitchBook data, in 2025, about 40% of VC investments in the US crypto sector flowed to companies involved in AI business, a significant increase from 2024.

Similarly, crypto builders transitioning to AI show clear differences in the paths chosen under different market environments.

In the US, as the regulatory environment becomes relatively clearer, protocol layer innovations have gained real survival space. The density of capital networks is high, with short paths from ideas to financing and a larger margin for error. Projects like Hyperbolic, EigenCloud, Gensyn, and Ritual share the common characteristic of designing new mechanisms from scratch rather than simply integrating applications into existing systems. Top VCs have clear investment theses on directions like "verifiable computation, agent coordination, decentralized ML," and are willing to provide ample margin for early technological exploration.

The situation in Asia is different. Singapore and Hong Kong take on more roles in compliance implementation and institutional fund transfers, with a relatively conservative regulatory framework and lower tolerance for pure protocol layer innovations. Builders with crypto backgrounds transitioning to AI often choose application layer and industry integration paths—leveraging the user base, payment capabilities, or data assets accumulated in crypto to quickly access AI products and services.

This is not a gap in capabilities but a difference in path selection caused by varying market signals and regulatory environments: the US encourages foundational mechanism innovation and early technological exploration, while Asia emphasizes compliance friendliness, rapid monetization, and deep integration with traditional industries.

Returning to the GitHub curve mentioned at the beginning. The monthly active developers dropped from 45K to 23K, which superficially appears to be a contraction of the industry. However, among those who remain, the proportion of established developers has reached a historical high, and they are flocking to ecosystems with real users, while being repriced in the AI industry in unprecedented ways.

As AI scaling encounters structural bottlenecks such as computing power aggregation, autonomous payments for agents, verifiability of data and decisions, and privacy coordination, these builders at the intersection of Crypto and AI, with their long-accumulated sensitivity to rules, incentives, and authenticity, are gradually transforming into the scarce system-level capabilities needed in the AI era.

As an investment institution that has been deeply engaged in crypto infrastructure since 2017, IOSG's judgment on this line goes beyond mere observation. We participated in the investment in EigenLayer's restaking mechanism before it was widely recognized in the market, led the seed round investment in Ingonyama (now MoonMath) betting on the migration of ZK hardware acceleration to the AI performance layer, and invested in Hyperbolic in 2024, optimistic about its path to solving decentralized computing power trust issues with crypto-native verification mechanisms.

The common logic behind these layouts is that the trust, coordination, and verification issues encountered in AI scaling will ultimately require the mechanism design capabilities accumulated by the crypto industry to solve. We believe that the intersection of crypto and AI is not just a narrative, but a structural opportunity that is unfolding.