Authors｜Xinyang & Ethan @ IOSG

In 2026, the Crypto open-source community's GitHub activity curve underwent a dramatic "bottoming out." From a peak of 45,000 monthly active developers in 2022, it plummeted to approximately 23,000. This halving on paper sparked discussions on social media about "narrative exhaustion." However, when we analyze the cross-section of this curve, what we see is not an industry contraction, but a profound "talent deleveraging."

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

Who left? Who's still here?

The departures were primarily driven by newcomers. In February 2024, the number of new developers reached 5,462, before declining sharply, with a 52% attrition rate for those less than a year in the industry. Most of these individuals entered the market during the bull run, working on NFT minting contracts, forking DeFi protocols, and providing front-end development for new L2 platforms. These roles are highly dependent on market hype; once the hype dies down and projects cease operation, the positions disappear. Data shows that newcomers' code contributions never exceeded 25% of the total, indicating they were never part of the industry's core circle from the outset.

▲ Newcomers flocked in during the bull market and left during the bear market; established devs (2+ years of experience) hit a record high during the same period.

Data source: Electric Capital Developer Report

On the other hand, the number of developers with more than two years of experience actually increased during the same period, reaching a record high and contributing about 70% of the code volume. Maria Shen, GP of Electric Capital, made a straightforward assessment: "When we look at the group of established developers, it is growing and looks very healthy."

They stayed not because they had no other choice.

Technically, the core work of crypto now involves infrastructure development that generally requires years of experience to understand: protocol layer development, security auditing, and cross-chain architecture. These tasks require years of experience to truly master and cannot be eliminated by the market simply when the hype dies down.

Economically, many veterans hold unvested tokens, governing power within protocols, and equity relationships. Their accumulated experience in the industry has created real barriers to entry and tangible rewards. Looking at the ecosystem distribution, they are voting with their feet: Bitcoin developers grew by 64.3% in 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 are still relying on narratives to survive.

▲ Source: Coincub Web3 Jobs Report 2025

Data source: Web3.Career

The changing job structure also confirms this. In 2025, the largest proportion of newly added Web3 positions was not developers, but Project & Programme Management, exceeding 27%. This is counterintuitive for an industry known for its technology-driven approach, but the underlying logic is not complex: the industry has moved from the construction phase to the execution phase, over 100 chains need to be integrated, and institutional clients have completely different compliance and security requirements. DAO governance needs to find a balance among stakeholders with differing interests. This is not traditional project management, but rather coordination and judgment in an environment where rules are still being formed.

While the industry appears to be shrinking, its core density is actually increasing. The bear market of 2018-2019 was also accompanied by a significant loss of developers, but it was followed by phenomenal projects like Uniswap, Aave, and OpenSea, defining the bull market of 2020-2021. The builders who remain in this round have more mature infrastructure, and the AI ​​era has given them a much larger stage than the previous one.

What abilities did those who stayed bring?

What special abilities has the cryptocurrency industry honed in its builders? To answer this question, we need to go back to the underlying principles of blockchain. Through bull and bear market cycles, this industry always operates on the same fundamental rule: code is law, and execution is the final outcome.

In the 2016 DAO incident, attackers exploited a recursive call vulnerability to steal $36 million. The code had no bugs, and the logic executed exactly as expected; the only problem was that the design hadn't anticipated the potential consequences. In 2021, the Poly Network cross-chain bridge was attacked, and $610 million was transferred within hours. No platform could halt the transfer, no institution could revoke it, and no legal provisions could offer recourse. This is the structural characteristic that distinguishes crypto from almost all other industries: zero fault tolerance and virtually no post-incident intervention.

This environment has forced out a set of capabilities that are rarely needed in other industries: building a working system from scratch that strangers are willing to participate in, in the absence of rules and trust.

This capability encompasses two levels. First, it involves building trust from scratch, relying solely on code and mechanisms to encourage strangers to deposit their real assets, without depending on any external authority. Second, it enables the design of functional systems despite technological and economic uncertainties, lacking regulatory frameworks, historical data, and industry standards for reference.

Both layers have concrete verification in crypto. Uniswap has no corporate guarantee, no KYC, no customer service; anyone can put funds into the liquidity pool, relying solely on trust in a few hundred lines of code and an economic mechanism, achieving a daily trading volume of tens of billions of dollars. MakerDAO has no central bank backing, no deposit insurance, and relies purely on on-chain governance and collateral mechanisms to maintain the stability of DAI. The DeFi Summer was even more extreme; there was no regulatory framework, no audit standards, and no historical data to refer to. Builders designed AMMs, lending protocols, and liquidity mining, going from concept to billions of dollars in TVL in just a few months. This capability manifests differently in builders at the protocol, application, and governance layers, but the underlying principle is the same.

The AI ​​era is creating a structurally similar problem. Model decision-making processes are opaque, and outputs cannot be independently verified. AI agents are beginning to autonomously execute transactions and allocate funds, but supporting rules and constraints are still lacking. Large model companies control both the models and evaluation standards, leaving users without effective verification methods. Computing power is highly concentrated in a few top-tier companies, leading to monopolistic pricing during periods of explosive demand. These problems all point to the same core issue: the trust problem in autonomous systems, which is being replayed on a larger scale in the development of AI.

Crypto builder has been handling these kinds of problems for years in environments without external authoritative rules, except that the previous scenario involved on-chain protocols, and now it's AI. A group of people have already directly applied the capabilities accumulated in crypto to AI and achieved results.

How will these capabilities be repriced in the AI ​​era?

The shift from crypto to AI has become increasingly common in recent years, but upon closer examination, they have taken away different things.

The most straightforward path is the direct transfer of hardware and experience. CoreWeave's three founders, Michael Intrator, Brian Venturo, and Brannin McBee, started mining Ethereum with GPUs in 2017, scaling from a single GPU to thousands. They shut down mining in 2022, and two months later launched ChatGPT, transforming their GPUs into AI computing power. In March 2025, they went public on Nasdaq with an IPO valuation of approximately $23 billion, and their market capitalization subsequently peaked at nearly $70 billion. OpenSea co-founder Alex Atallah, who had experience dealing with the aggregation and routing of highly heterogeneous assets in the NFT market, applied the same experience to AI model routing, founding OpenRouter, which served over 5 million developers within two years and reached a valuation of $500 million.

Another type of migration is even more noteworthy. NEAR founder Illia Polosukhin, a co-author of the Transformer paper, initially wanted to build AI applications using natural language after leaving Google. However, she encountered a real problem during development: she needed to make cross-border payments to data labelers around the world, most of whom did not have bank accounts. Blockchain technology became the best solution to this payment problem. Now, NEAR is transforming into an AI infrastructure platform, with its core focus 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 at NEAR has become the most difficult starting point to replicate in this direction. After leaving Circle, co-founder Sean Neville founded Catena Labs, positioning itself as an AI-native bank, directly migrating his understanding of stablecoin infrastructure to AI agent financial scenarios. a16z crypto led an $18 million seed round. Nader Dabit, a senior developer at Aave and Lens Protocol, moved to Cognition, bringing his experience in building the developer ecosystem across multiple crypto protocols into the field of AI agent tools.

This group of people took away not just GPU hardware or user networks, but also intuition for mechanism design, experience in building a developer ecosystem, and the judgment to build trustworthy systems from scratch when rules are lacking. These capabilities precisely address the three structural gaps encountered in the scaling up of 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, demand fluctuates greatly, cloud providers are expensive and have long waiting lists, and enterprises are unwilling to stockpile their own hardware. This problem has two aspects: how to aggregate and allocate computing power, and how to use the aggregated computing power more efficiently. Crypto Builder has directly transferable experience in both of these aspects.

Hyperbolic addresses the issues of allocation and trust. Founder Jasper Zhang brought decentralized mechanisms to the AI ​​computing power arena: tokens encourage decentralized GPU holders to contribute their idle computing power, but the core issue is trust. Why should one believe that a computational result from an unknown node is correct? The core innovation, PoSP, uses random sampling and game theory to make honesty the dominant strategy for nodes. It eliminates the need for full verification, has low overhead, is scalable, and provides reliable results. This mechanism is directly transferred from the logic of crypto's verification of unknown node behavior.

MoonMath addresses the issue of efficiency. Its predecessor, Ingonyama, focused on ZK hardware acceleration, increasing the speed of ZK proof generation several times over under extreme computational constraints. Now, its focus has shifted to the Physical AI performance layer, working on sparse attention acceleration (LiteAttention) for video diffusion models, low-rank decomposition of FFN layers (LiteLinear), and training backpropagation acceleration (BackLite). From ZK acceleration to AI inference acceleration, the underlying capability remains the same: making mathematics run faster under extreme computational constraints. The focus has changed, but the accumulated experience 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 don't disrupt the overall system in the pursuit of their individual goals? Each participant is pursuing its own objective function, and there's no guarantee that the system will function properly when they are combined, especially since the agents' execution speed far exceeds the window for human intervention.

This is a type of problem that Crypto Builder has repeatedly addressed in its DAO governance and tokenomics design: allowing stakeholders with completely different interests to operate according to a pre-set system direction without a central authority. Crypto's answer is an economic mechanism; violations incur real economic costs, and the rules are written in the code and executed automatically.

EigenLayer directly migrated this mechanism to AI scenarios. Through restaking, nodes need to stake assets before participating in collaboration; failure to fulfill obligations or violations trigger automatic penalties. These rules are not suggestions, but rigid boundaries with real economic costs. EigenCloud extends this logic to the verifiable computation and collaborative governance of AI agents, ensuring that agents pursue their goals within predefined boundaries. Constraining agents with economic mechanisms is far more reliable than constraining them with ethical guidelines.

AI Agent Autonomous Payment

There's an even more fundamental question: how does the agent make payments? Traditional payment systems are designed for humans. Credit cards require account opening, bank transfers require authorization, and each step assumes the operator is human, has an identity, and will wait. Agents, however, don't wait. They might initiate a large number of requests per second, each potentially involving micro-payments. Traditional payment channels simply fail in this scenario.

Stablecoin and on-chain rules are the infrastructure already built by crypto builder, natively supporting programmability, authorization-free operation, and 24/7 operation. These three characteristics are precisely the hard requirements of agent payment scenarios; all that's missing is a protocol layer to integrate stablecoin into the agent workflow.

Launched by Coinbase in May 2025, x402 activates the HTTP 402 status code, embedding stablecoin payments directly into HTTP requests. The payment is completed simultaneously with the agent's request, requiring no account and taking approximately two seconds to settle. As of April 2026, the x402 protocol had processed over 165 million transactions, with a cumulative transaction volume of approximately $50 million and 69,000 active agents (source: x402 Foundation). Cloudflare, AWS, Stripe, and Anthropic MCP have all integrated with it. Agent payments have become a sector with real user traffic.

These three directions correspond to three structural gaps encountered in the scaling of AI: the aggregation and efficiency of computing power, the incentive alignment for multi-agent collaboration, and the infrastructure for autonomous payments. These three issues do not have ready-made answers in traditional software architecture, but the crypto industry has corresponding experience in handling them. The capabilities haven't disappeared; they've simply found new application scenarios.

Builder's new positioning: from someone who writes contracts to someone who sets the rules for AI.

The scaling up of AI is creating a previously non-existent skill gap. It's not a shortage of technical talent, but a shortage of people capable of designing trust mechanisms within autonomous systems. As the target audience shifts from humans to AI, the role of the crypto builder is 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 how trust is established and the logic of rule execution. In the pre-AI era, crypto builder interacted with human participants; rules were written into contracts, with zero tolerance for errors, but the system boundaries were relatively clear. In the AI-Native era, when the interacting object becomes an autonomous AI agent, the problem to be solved is that the agent's behavior is unpredictable, its execution speed far exceeds the human intervention window, and the system boundaries themselves need to be redefined under greater uncertainty. The role of crypto builder is shifting from "writing secure contracts" to "designing trusted mechanisms for autonomous AI systems."

The hiring practices of leading institutions are already reflecting this change:

▲ AI/Data Core Positions Actively Opened by Leading Exchanges in Q1 2026

Source: Gate Research Institute

The recruitment practices of leading exchanges and institutions in 2026 clearly reflect this trend: they are no longer simply recruiting AI engineers or crypto developers, but looking for people who can connect the two sides, who understand on-chain incentive distortion and governance game theory, can deeply embed AI tools into crypto workflows, and design mechanisms that allow agents to align with regulators and users in the long term.

Capital allocation trends have already reflected this assessment. Paradigm is raising a new fund of up to $1.5 billion, expanding its investment scope from crypto to AI and robotics. Haun Ventures completed its $1 billion Fund II, focusing on financial infrastructure integrating crypto and AI, particularly payments, stablecoins, and agent-to-agent economic systems that support autonomous transactions and coordination by AI agents. a16z crypto completed its $2.2 billion Crypto Fund V, explicitly stating that the fund will invest 100% in the crypto space. Facing the complexity and opacity of the AI ​​era, they will focus on applications of crypto's transparency, verifiability, and decentralization. Furthermore, according to PitchBook data, in 2025, approximately 40% of VC investment in the US crypto sector went to companies also involved in AI businesses, a significant increase compared to 2024.

While both crypto builder programs are shifting towards AI, the paths they choose vary significantly depending on the market environment.

In the US, with a relatively clearer regulatory environment, protocol-level innovation has gained real room to thrive. The high density of capital networks, short path from idea to funding, and significant tolerance for failure are key advantages. 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-tier VCs have published research papers specifically on areas such as verifiable computing, agent coordination, and decentralized machine learning, and are willing to provide ample room for error in early-stage technological exploration.

The situation in Asia is different. Singapore and Hong Kong play more of a role in compliance implementation and institutional funding transfer, with relatively conservative regulatory frameworks and lower tolerance for pure protocol-level innovation. When builder companies with crypto backgrounds transition to AI, they tend to choose application-level and industry integration paths—leveraging their accumulated user base, payment capabilities, or data assets from crypto to quickly integrate AI products and services.

This is not a difference in capabilities, but rather a difference in path choices caused by different market signals and regulatory environments: the United States encourages innovation in underlying mechanisms and early-stage technological exploration, while Asia emphasizes compliance-friendly practices, rapid monetization, and deep integration with traditional industries.

Returning to the GitHub curve mentioned at the beginning, the number of monthly active developers dropped from 45K to 23K, seemingly indicating an industry contraction. However, among those who remain, the proportion of established developers has reached an all-time high, indicating a surge towards ecosystems with real users and a revaluation of their value by the AI ​​industry in unprecedented ways. As AI scaling encounters structural bottlenecks such as computing power aggregation, agent-driven autonomous payments, data and decision verifiability, and privacy coordination, these builders, at the intersection of Crypto and AI, are gradually transforming their long-accumulated sensitivity to rules, incentives, and authenticity into a system-level capability that is scarce in the AI ​​era.

As an investment firm deeply involved in crypto infrastructure since 2017, IOSG's assessment of this field goes beyond mere observation. We invested in EigenLayer before its restaking mechanism was widely recognized, led the seed round of Ingonyama (now MoonMath) betting on ZK hardware acceleration's migration to the AI ​​performance layer, and invested in Hyperbolic in 2024, optimistic about its approach of using crypto's native verification mechanisms to solve the trust problem in decentralized computing power. The common logic behind these investments is that the trust, coordination, and verification problems encountered in scaling AI 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.