Crypto is not dead; it has just handed over talent to AI.

Author | Xinyang & Ethan @ IOSG

In 2026, the activity curve of the open-source crypto community on GitHub completed a remarkable “bottoming out.” From a peak of 45k monthly active developers in 2022, it fell back to about 23k. This halving on paper sparked discussions on social media about “narrative exhaustion.” However, when we dissect this curve’s cross-section, what we see is not industry contraction, but a profound “talent de-leverage.”

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

Who left? Who is still here?

The main departure was newcomers. In February 2024, monthly new developers reached 5,462, then sharply declined, with a churn rate of 52% within less than a year of entering the industry. Most of these people flooded in during the bull market, working on NFT minting contracts, forking DeFi protocols, and building frontends for new L2s. These roles are highly dependent on market hype; once the hype fades, projects halt operations, and roles disappear. Data shows that newcomers’ code contributions have never exceeded 25% of the total, indicating these people were never part of the core industry circle from the start.

▲ Newcomers flood in during bull markets and leave during bear markets; established developers (with over 2 years of experience) hit record highs during the same period

Data source: Electric Capital Developer Report

On the other side, developers with over two years of experience actually increased during the same period, reaching new highs and contributing about 70% of the code. Electric Capital’s GP Maria Shen’s assessment is straightforward: “When we look at the group of established developers, it’s growing and appears very healthy.”

They stay not because they have no other options.

Technically, the core work in crypto now involves infrastructure development—protocol layer development, security audits, cross-chain architecture—which requires years of experience to master. These tasks demand deep expertise; they cannot be replaced simply because hype wanes. They are not easily eliminated from the market.

Economically, many veterans hold unvested tokens, governance rights within protocols, and equity relationships. Their accumulated assets have become real barriers and sources of returns in this industry. From an ecosystem distribution perspective, they are voting with their feet: Bitcoin developers grew 64.3% over two years, Solana +11.1%, while Cosmos declined 51.1%, Polkadot down 46.9%. Veterans are consolidating into ecosystems with real users and revenue, leaving behind projects still sustained by narratives.

▲ Source: Coincub Web3 Jobs Report 2025

Data source: Web3.Career

Changes in job structure also confirm the same trend. In 2025, among new Web3 positions, the largest share was Project & Program Management, accounting for over 27%. For an industry known for its technical focus, this is counterintuitive, but the logic is simple: as the industry shifts from construction to execution, with over a hundred chains to integrate, institutional clients demanding compliance and security, and DAO governance balancing conflicting stakeholder interests—this is not traditional project management but coordination and judgment in an environment where rules are still forming.

On the surface, the industry appears to shrink, but its core density is actually increasing. The 2018-2019 bear market also saw massive developer attrition, yet it gave rise to iconic projects like Uniswap, Aave, OpenSea, which defined the 2020-2021 bull run. The builders who remain now have more mature infrastructure, and the AI era provides them with an even larger stage than before.

What do the remaining builders bring to the table?

What special skills has the crypto industry cultivated in builders? To answer this, we need to revisit the fundamental principles of blockchain: under the cycles of bull and bear markets, this industry always operates on the same underlying rules—code is law, execution is final.

In 2016, during The DAO incident, attackers exploited a recursive call bug to steal $36 million. The code had no bugs; it executed exactly as intended, but the boundary conditions were not anticipated by the designers. In 2021, the Poly Network cross-chain bridge was attacked, transferring $610 million within hours. No platform could halt it; no institution could revoke the transactions; no legal recourse was available. This is a defining structural feature of crypto: zero tolerance for fault, almost no post-incident intervention.

This environment has forced the development of a rarely needed skill elsewhere: building trustworthy, operational systems from scratch in conditions of rule and trust absence, where strangers are willing to participate.

This skill involves two levels. First, establishing trust from zero, without relying on external authorities—only code and mechanisms make strangers willing to put real assets in. Second, making judgments under dual uncertainty—no regulatory framework, no historical data, no industry standards—yet still designing systems that can operate reliably.

Both levels are validated in crypto. Uniswap has no company guarantee, no KYC, no customer service; anyone can deposit funds into liquidity pools, relying solely on trust in a few hundred lines of code and an economic mechanism, enabling hundreds of billions of dollars in daily trading volume. MakerDAO has no central bank backing, no deposit insurance; it relies purely on on-chain governance and collateralization to maintain DAI’s stability. During DeFi Summer, even more extreme—no regulatory framework, no audit standards, no historical data—builders designed AMMs, lending protocols, liquidity mining, creating billions in TVL within months. These capabilities manifest differently across protocol, application, and governance layers, but the underlying principles are the same.

The AI era is creating a highly similar structural problem. Model decision processes are opaque, outputs cannot be independently verified. AI agents are beginning to autonomously execute trades and allocate funds, but the rules and constraints for these are not yet established. Major model companies control both the models and evaluation standards, leaving users with limited verification tools. Computing power is concentrated among a few top firms, leading to monopolistic pricing during demand surges. These issues point to the same core: trust in autonomous systems, recurring at larger scales in AI.

Crypto builders, having long dealt with such issues in environments without external authority rules—initially on-chain protocols, now in AI—have already begun to transfer their skills into AI, producing tangible results.

How can these skills be revalued in the AI era?

The shift from crypto to AI has become increasingly common in recent years, but a closer look reveals that what they carry away differs.

The most straightforward path is direct transfer of hardware and experience. CoreWeave’s founders—Michael Intrator, Brian Venturo, and Brannin McBee—began mining Ethereum with GPUs in 2017, scaling from a single machine to thousands. They shut down mining in 2022, and two months later, ChatGPT launched. Their GPUs became AI compute resources, and by March 2025, they went public on Nasdaq with an estimated valuation of about $23 billion, peaking near $70 billion. OpenSea co-founder Alex Atallah, experienced in aggregating and routing highly heterogeneous assets in NFT markets, applied similar expertise to AI model routing, founding OpenRouter, serving over 5 million developers within two years, with a valuation of $500 million.

A more noteworthy category involves strategic shifts. Illia Polosukhin, co-author of the Transformer paper, left Google to build AI applications with natural language. During development, he faced a real-world problem: cross-border payments for data annotation workers worldwide, most of whom lack bank accounts. Blockchain technology became the best solution. Now, NEAR is transforming into an AI infrastructure platform, focusing on user-owned AI and decentralized confidential machine learning (DCML), enabling users to access AI services without exposing data. The decentralized architecture experience accumulated at NEAR has become a hard-to-copy starting point. Sean Neville, co-founder of Circle, left to create Catena Labs, positioning as an AI-native bank, applying stablecoin infrastructure knowledge to AI agent finance, raising $18 million in seed funding from a16z crypto. Nader Dabit, senior developer at Aave and Lens Protocol, shifted to Cognition, bringing developer ecosystem experience from multiple crypto protocols into AI agent tools.

These individuals are not just taking hardware or user networks—they are carrying intuition in mechanism design, developer ecosystem building, and trustless system construction in rule-scarce environments. These skills directly address three structural gaps faced at AI scale.

Aggregation and optimization of compute power

Compute is the most direct bottleneck for AI scaling. Training and inference require massive GPUs; demand fluctuates wildly, cloud providers are expensive and have queues, companies prefer not to stock hardware. The issues are twofold: how to aggregate and allocate compute, and how to use it more efficiently. Crypto builders have direct transferable experience in both.

Hyperbolic tackles trust and allocation. Founder Jasper Zhang introduced decentralized mechanisms: tokens incentivize distributed GPU owners to contribute idle compute, but trust remains the core issue. Why trust a stranger’s calculation? PoSP (Proof of Staked Participation) uses random sampling and game theory to make honesty the optimal strategy, avoiding full verification, low overhead, scalable, and reliable—directly migrated from crypto’s logic of verifying stranger nodes.

MoonMath addresses efficiency. Originating from Ingonyama, which optimized ZK hardware acceleration, it now focuses on physical AI performance—sparse attention acceleration for diffusion models (LiteAttention), low-rank decomposition of FFN layers (LiteLinear), and backpropagation speedups (BackLite). From ZK proof acceleration to AI inference acceleration, the underlying capability remains: making math run faster under extreme computational constraints. The track has shifted, but the skills are not wasted.

AI governance and incentive mechanism design

When multiple AI agents begin collaborating, how to ensure they don’t sabotage the system pursuing their own goals? Each participant optimizes their own objective, with no guarantee that the combined system remains functional, especially as agents act faster than human intervention can respond.

This is a recurring problem crypto builders have addressed in DAO governance and tokenomics: enabling stakeholders with conflicting interests to operate systemically without central authority. Crypto’s answer: economic mechanisms—violations incur real economic costs, rules coded into smart contracts, and enforced automatically.

EigenLayer directly transferred this mechanism to AI. Through restaking, nodes must pledge assets before participating; violations trigger automatic penalties—rules are not suggestions but rigid boundaries with real economic consequences. EigenCloud extends this logic to verifiable AI computation and collaborative governance, ensuring agents stay within predefined bounds while pursuing their goals. Using economic mechanisms to constrain agents is far more reliable than relying solely on ethical guidelines.

Autonomous payment for AI agents

Another fundamental issue: how do agents pay? Traditional payment systems are designed for humans—credit cards require accounts, bank transfers need authorization, each step assumes a human operator with identity and patience. Agents don’t wait; they may send thousands of requests per second, each involving micro-payments. Traditional payment channels fail here.

Stablecoins and on-chain rules form the infrastructure crypto builders have established—programmable, permissionless, 24/7 operation. These features are exactly what agent payments require; what’s missing is a protocol connecting stablecoins to agent workflows.

x402, launched by Coinbase in May 2025, activates HTTP 402 status code, embedding stablecoin payments directly into HTTP requests. Agents initiate requests and pay simultaneously, no accounts needed, settlement in about two seconds. By April 2026, x402 handled over 165 million transactions, totaling around $50 million, with 69,000 active agents (source: x402 Foundation). Companies like Cloudflare, AWS, Stripe, and Anthropic MCP are integrated. Agent payments are now a real traffic channel.

These three directions correspond to the three structural gaps in AI scaling: compute aggregation and efficiency, multi-agent collaboration incentives, and autonomous payment infrastructure. While these problems have no ready solutions in traditional software architecture, crypto industry experience provides relevant answers. Capabilities have not disappeared—they’ve just found new application scenarios.

The new role of builders: from contract coders to rule designers for AI

AI’s scaling is creating a previously unseen functional gap—not a talent shortage, but a trust mechanism design gap. When the object of service shifts from humans to AI, the role of crypto builders is being redefined.

The table below compares the paradigm shifts in specific functions:

The core difference between the two paradigms is not the tech stack but the way trust is established and rules are enforced. Pre-AI, crypto builders dealt with human participants, rules embedded in contracts, zero tolerance for fault, with relatively clear system boundaries. In the AI-native era, when the interaction target becomes autonomous AI agents, the challenge is: their behavior is unpredictable, execution speeds surpass human intervention windows, and system boundaries must be redefined under greater uncertainty. The role of crypto builders is shifting from “writing secure contracts” to “designing trustworthy mechanisms for autonomous AI systems.”

Leading institutions are already reflecting this change in their hiring:

▲ Q1 2026 top exchanges actively opening AI/data core roles

Source: Gate Research Institute

In Q1 2026, top exchanges and institutions’ hiring clearly shows this trend: they are no longer just recruiting AI engineers or crypto developers, but people who can connect both worlds—understand on-chain incentive distortions and governance games, embed AI tools deeply into crypto workflows, and design mechanisms to align agents with regulators and users over the long term.

Capital allocation also reflects this judgment. Paradigm is raising a new fund up to $1.5 billion, expanding from crypto into AI and robotics. Haun Ventures completed a $1 billion Fund II, focusing on crypto-AI integrated financial infrastructure—supporting AI agent autonomous trading, payments, stablecoins, and agent-to-agent economies. a16z crypto closed its fifth fund at $2.2 billion (Crypto Fund V), explicitly investing 100% in crypto. Facing the complexity and opacity of the AI era, they emphasize applying crypto’s transparency, verifiability, and decentralization features. According to PitchBook, in 2025, about 40% of US crypto VC investments involved companies with AI components—significantly higher than in 2024.

The path for crypto builders shifting into AI varies across markets:

In the US, with clearer regulation, protocol-layer innovation has room to grow. Capital networks are dense, ideas to funding are short, and tolerance for failure is higher. Projects like Hyperbolic, EigenCloud, Gensyn, Ritual focus on designing new mechanisms from scratch rather than simple integrations. Top VCs have clear investment theses around verifiable computation, agent coordination, decentralized ML, and are willing to fund early-stage exploration.

In Asia, the situation differs. Singapore and Hong Kong mainly handle compliance and institutional capital transfer, with more conservative regulation and lower tolerance for pure protocol innovation. Crypto-leaning builders turning to AI often choose application-layer and industry integration paths—leveraging existing user bases, payment capabilities, or data assets to quickly deploy AI products and services.

This is not a capability gap but a market signal and regulatory environment-driven path choice: the US encourages foundational mechanism innovation and early tech exploration, while Asia emphasizes compliance, rapid monetization, and deep integration with traditional industries.

Returning to the initial GitHub activity curve: the drop from 45K to 23K monthly active developers may seem like industry contraction. But among those remaining, the proportion of established devs hit record highs, flowing into ecosystems with real users, and being re-priced by AI industry in unprecedented ways. When AI scaling encounters structural bottlenecks like compute aggregation, autonomous agent payments, verifiable data and decision-making, these long-accumulated sensitivities to rules, incentives, and authenticity are gradually transforming into scarce systemic capabilities for the AI era.

As an investment firm deeply involved in crypto infrastructure since 2017, IOSG’s view is not just observational. We invested early in EigenLayer’s restaking mechanism before it gained broad recognition, led seed rounds in MoonMath (formerly Ingonyama) for ZK hardware acceleration towards AI performance, and invested in Hyperbolic in 2024, optimistic about its crypto-native verification approach to decentralized compute trust. The common logic behind these investments: the trust, coordination, and verification challenges faced at AI scale will ultimately require the mechanism design expertise accumulated in crypto. We believe that the intersection of crypto and AI is not just a narrative but a structural opportunity in the making.