IOSG | After Developer Count Halved: Crypto Isn’t Dead—It’s Just Letting AI Take the Talent
TechFlow Selected TechFlow Selected
IOSG | After Developer Count Halved: Crypto Isn’t Dead—It’s Just Letting AI Take the Talent
The number of crypto developers has halved, but the proportion of core developers has increased—indicating the industry is not dead, but pivoting toward AI.
Navigating Web3 tides with focused insightsAuthors: Xinyang & Ethan, IOSG
In 2026, the GitHub activity curve of the open-source crypto community completed a remarkable “bottoming out.” Monthly active developers fell from a peak of 45K in 2022 to roughly 23K—a paper-based halving that sparked social media discussions about “narrative exhaustion.” Yet when we dissect this curve’s cross-section, what emerges is not industry contraction, but a profound “talent deleveraging.”
▲ Data source: Electric Capital Developer Report, based on Crypto Ecosystems GitHub
I. Who Left? Who Stayed?
Most who left were newcomers. In February 2024 alone, 5,462 new developers joined—followed by a sharp decline. Within one year of entering the field, 52% dropped out. These individuals largely entered during bull markets, building NFT minting contracts, forking DeFi protocols, or developing frontends for new L2s. Their roles were highly dependent on market momentum; once热度 faded, projects shut down and positions vanished. Data shows newcomers never contributed more than 25% of total code—meaning they were never part of the industry’s core from the outset.
▲ Newcomers flooded in during bull markets and exited during bear markets; Established devs (2+ years’ experience) hit an all-time high concurrently
Data source: Electric Capital Developer Report
Meanwhile, developers with over two years of experience rose—not declined—reaching an all-time high and contributing ~70% of total code volume. Maria Shen, GP at Electric Capital, put it plainly: “When we look at the established developer cohort, it’s growing—and looks very healthy.”
They stayed not because they lacked alternatives.
Technically, core crypto work today is infrastructure development demanding years of accumulated expertise: protocol-layer engineering, security auditing, cross-chain architecture—skills requiring deep, long-term mastery, not easily discarded when market sentiment shifts.
Economically, many veterans hold unvested tokens, governance rights, and equity stakes across protocols—accumulated assets forming real barriers to entry and tangible returns. Looking at ecosystem distribution, they’re voting with their feet: Bitcoin developers grew 64.3% over two years, Solana +11.1%, while Cosmos declined 51.1% and Polkadot 46.9%. Veterans are concentrating in ecosystems with real users and revenue—and exiting those sustained solely by narrative.
▲ Source: Coincub Web3 Jobs Report 2025
Data source: Web3.Career
Shifts in job structure confirm the same trend. In 2025, the largest share of new Web3 roles wasn’t developers—but Project & Programme Management, accounting for over 27%. Counterintuitive for a tech-driven industry, yet logical: the sector has moved from construction into execution. With 100+ chains needing integration, institutional clients demanding new compliance and security standards, and DAO governance requiring balance among stakeholders with divergent interests—this isn’t traditional project management. It’s coordination and judgment within environments where rules themselves are still being formed.
The industry may appear to be shrinking on the surface—but its core density is rising. The 2018–2019 bear market similarly saw massive developer attrition, yet paved the way for breakout projects like Uniswap, Aave, and OpenSea—defining the 2020–2021 bull run. This cycle’s remaining builders operate atop far more mature infrastructure—and the AI era offers them a vastly larger stage than the last.
II. What Capabilities Do Those Who Stayed Cultivate?
What unique capabilities does the crypto industry actually forge in its builders? To answer, we return to blockchain’s foundational principle—unchanged across bull and bear cycles: code is law; execution is final.
In 2016, the DAO hack exploited a recursive call vulnerability to drain $36 million. The code had no bugs—it executed exactly as designed—but designers hadn’t anticipated that edge case. In 2021, Poly Network’s cross-chain bridge was compromised, transferring $610 million in hours. No platform could halt it; no institution reverse it; no legal clause recover it. This zero-tolerance, near-zero-post-hoc-intervention environment is crypto’s structural hallmark—distinguishing it from nearly every other industry.
This environment forces the cultivation of a rare capability: building functional, trustless systems from scratch—systems strangers willingly participate in, despite absent external authority or preexisting trust.
This capability operates on two levels. First: establishing trust from zero—relying solely on code and mechanism, not third-party validation, to convince strangers to deposit real assets. Second: making sound judgments amid dual technological and economic uncertainty—no regulatory frameworks, no historical data, no industry standards—to design systems that actually function.
Both levels have concrete validation in crypto. Uniswap operates without corporate backing, KYC, or customer support. Anyone deposits funds into liquidity pools trusting only hundreds of lines of code and an economic mechanism—achieving billions in daily trading volume. MakerDAO maintains DAI’s stability without central bank backing or deposit insurance—relying purely on on-chain governance and collateral mechanisms. During DeFi Summer, builders launched AMMs, lending protocols, and liquidity mining—all without regulatory guardrails, audit standards, or historical precedent—scaling from concept to multi-billion-dollar TVL in months. This capability manifests differently across protocol, application, and governance layers—but shares the same foundational logic.
The AI era is now generating structurally similar challenges. Model decision-making remains opaque; outputs lack independent verifiability. AI agents autonomously execute trades and allocate capital—yet supporting rule systems and constraint mechanisms remain undeveloped. Major model firms control both models and evaluation standards, leaving users without effective verification tools. Compute power concentrates in a few top-tier firms, enabling monopolistic pricing during demand surges. All these problems converge on one core issue: trust in autonomous systems—replaying at scale in AI.
Crypto builders have spent years solving precisely such problems—without external authority or binding rules. The context merely shifted: from on-chain protocols to AI. And a cohort has already transplanted crypto-honed capabilities directly into AI—with measurable results.
III. How Are These Capabilities Being Repriced in the AI Era?
Crypto-to-AI transitions have become increasingly common—but what builders carry across differs significantly.
The most direct path is hardware and experiential transfer. CoreWeave’s founders—Michael Intrator, Brian Venturo, and Brannin McBee—mined Ethereum with GPUs starting in 2017, scaling from one rig to thousands. They shut down mining in 2022; two months later, ChatGPT launched. Their GPU fleet instantly became AI compute infrastructure. In March 2025, CoreWeave went public on Nasdaq with a $2.3B IPO valuation—later peaking near $7B.
OpenSea co-founder Alex Atallah tackled hyper-heterogeneous asset aggregation and routing in NFT markets—then applied identical expertise to AI model routing, founding OpenRouter. Within two years, it served >5M developers and achieved a $500M valuation.
A second migration path is even more revealing. NEAR co-founder Illia Polosukhin co-authored the Transformer paper. After leaving Google, he initially aimed to build AI applications using natural language—but encountered a practical hurdle: paying global data labelers cross-border, many lacking bank accounts. Blockchain emerged as the optimal payment solution.
NEAR is now pivoting to an AI infrastructure platform, focusing on user-owned AI and decentralized confidential machine learning (DCML)—enabling users to leverage AI services without exposing raw data. NEAR’s accumulated decentralized architecture expertise forms an exceptionally hard-to-replicate foundation for this direction.
Circle co-founder Sean Neville founded Catena Labs—an AI-native bank—transferring his stablecoin infrastructure expertise directly into AI agent financial use cases. a16z crypto led its $18M seed round. Nader Dabit, a senior developer from Aave and Lens Protocol, joined Cognition—bringing his crypto protocol developer-ecosystem-building experience into the AI agent tooling space.
These builders aren’t just carrying GPUs or user networks—they’re bringing mechanism-design intuition, developer-ecosystem-building experience, and judgment honed in rule-less environments to construct trustworthy systems from scratch. These capabilities map precisely onto three structural gaps AI faces at scale.
Compute Aggregation & Optimization
Compute is AI’s most immediate bottleneck. Training and inference demand vast GPU resources, subject to volatile demand. Cloud providers are expensive and backlogged; enterprises resist hoarding hardware. This problem has two dimensions: how to aggregate and allocate compute—and how to use aggregated compute efficiently. Crypto builders possess directly transferable experience on both fronts.
Hyperbolic tackles allocation and trust. Founder Jasper Zhang imported decentralized mechanism design into AI compute: tokens incentivize distributed GPU owners to contribute idle capacity—but the deeper challenge is trust.
Why trust a stranger node’s computation result? Its core innovation, PoSP, uses randomized sampling plus game theory to make honesty the dominant strategy for nodes—eliminating full verification, reducing overhead, enabling scalability, and ensuring reliability. This mechanism migrates directly from crypto’s logic for verifying behavior of unknown nodes.
MoonMath solves efficiency. Formerly Ingonyama, it specialized in ZK hardware acceleration—boosting ZK proof generation speed several-fold under extreme computational constraints. Now focused on Physical AI performance layers, it develops sparse attention acceleration for video diffusion models (LiteAttention), low-rank decomposition for FFN layers (LiteLinear), and accelerated backpropagation for training (BackLite). From ZK acceleration to AI inference acceleration—the underlying skill remains identical: making math run faster under extreme compute constraints. The domain changed—but the expertise didn’t go to waste.
AI Governance & Incentive Mechanism Design
When multiple AI agents collaborate on tasks, how do we ensure pursuit of individual goals doesn’t destabilize the overall system? Each participant optimizes its own objective function; no guarantee their combined actions preserve system integrity—and agent execution speeds far exceed human intervention windows.
This is precisely the problem crypto builders repeatedly solved in DAO governance and tokenomics: aligning participants with divergent incentives to operate cooperatively—without central authority. Crypto’s answer is economic mechanisms—where violations trigger real financial penalties, codified and enforced automatically.
EigenLayer transplants this logic directly into AI. Its restaking mechanism requires nodes to stake assets before joining collaboration; non-compliance or violations trigger automatic penalties. These aren’t suggestions—they’re rigid boundaries backed by real economic cost. EigenCloud extends this logic to verifiable computing and collaborative governance for AI agents—ensuring agents pursuing self-interest remain bounded within predefined parameters. Constraining agents via economic mechanisms is far more reliable than relying on ethical guidelines.
AI Agent Autonomous Payments
A more fundamental question remains: how do agents pay? Traditional payment systems are built for humans: credit cards require accounts; bank transfers need authorization—each step assumes human identity, deliberation, and patience. Agents don’t wait. They may initiate thousands of requests per second—each involving micro-payments—rendering traditional payment rails entirely inoperative.
Stablecoins and on-chain rules constitute infrastructure crypto builders have already built—natively supporting programmability, permissionlessness, and 24/7 operation. These three properties are precisely the hard requirements for agent payments. All that was missing was a protocol bridging stablecoins into agent workflows.
x402, launched by Coinbase in May 2025, activates HTTP status code 402—embedding stablecoin payments directly into HTTP requests. An agent initiates a request and completes payment simultaneously—no account required, settlement in ~2 seconds. As of April 2026, x402 has processed >165M transactions, totaling ~$50M in volume, with 69,000 active agents (source: x402 Foundation). Cloudflare, AWS, Stripe, and Anthropic MCP have all integrated it. Agent payments are now a live, high-traffic vertical.
These three directions map to AI’s three structural bottlenecks at scale: compute aggregation & efficiency, multi-agent incentive alignment, and autonomous payment infrastructure. None have ready-made answers in traditional software architecture—but crypto has corresponding battle-tested experience. Capabilities haven’t vanished—they’ve found new contexts.
IV. Builders’ New Role: From Contract Writers to AI Rule-Designers
AI scaling is creating a previously nonexistent functional gap—not for technical talent, but for professionals capable of designing trust mechanisms within autonomous systems. As the service recipient shifts from humans to AI, the crypto builder’s role is being fundamentally redefined.
The table below compares shifts across functional paradigm dimensions:
The core difference between paradigms lies not in tech stacks—but in how trust is established and rules enforced. In the pre-AI era, crypto builders faced human participants: rules were encoded in smart contracts, fault tolerance was zero—but system boundaries remained relatively well-defined.
In the AI-native era, interaction partners are autonomous AI agents—whose behaviors are unpredictable, whose execution speed dwarfs human intervention windows, and whose very system boundaries must be redefined amid heightened uncertainty. The crypto builder’s role is shifting—from “writing secure contracts” to “designing trustworthy mechanisms for AI autonomous systems.”
Top-tier institutions’ hiring already reflects this shift:
▲ Top exchanges’ actively open AI/data core roles, Q1 2026
Source: Gate Research Institute
In 2026, leading exchanges and institutions clearly signal this trend: they’re no longer hiring pure AI engineers or crypto developers—but individuals who bridge both domains—understanding on-chain incentive distortions and governance games, deeply embedding AI tools into crypto workflows, and designing mechanisms that align agents, regulators, and users over time.
Capital allocation likewise confirms this view. Paradigm is raising a new fund up to $1.5B, expanding investment scope from crypto into AI and robotics. Haun Ventures closed its $1B Fund II, prioritizing financial infrastructure at the crypto-AI intersection—especially payments, stablecoins, and agent-to-agent economic systems enabling autonomous AI trading and coordination.
a16z crypto closed its $2.2B fifth fund (Crypto Fund V), explicitly committing 100% to crypto. Facing AI’s complexity and opacity, it will prioritize applications of crypto’s transparency, verifiability, and decentralization. Per PitchBook, ~40% of U.S. crypto VC funding in 2025 flowed to companies operating in both AI and crypto—up sharply from 2024.
Crypto builders transitioning to AI chart distinct paths depending on market conditions.
In the U.S., clearer regulatory clarity has granted protocol-layer innovation genuine breathing room. Dense capital networks shorten the path from idea to funding, allowing greater tolerance for experimentation. Projects like Hyperbolic, EigenCloud, Gensyn, and Ritual share a common trait: designing novel mechanisms from scratch—not merely integrating AI tools atop existing systems. Top VCs publish explicit investment theses around “verifiable computing,” “agent coordination,” and “decentralized ML”—and provide ample runway for early-stage technical exploration.
Asia tells a different story. Singapore and Hong Kong serve primarily as compliance gateways and institutional capital conduits—operating under comparatively conservative regulatory frameworks with lower tolerance for pure protocol-layer innovation. Crypto-background builders moving into AI thus favor application-layer and industry-integration paths—leveraging crypto-acquired user bases, payment capabilities, or data assets to rapidly embed AI products and services.
This isn’t a capability gap—it’s a divergence in market signals and regulatory environments driving distinct strategic choices: the U.S. favors foundational mechanism innovation and early technical exploration; Asia emphasizes compliance readiness, rapid monetization, and deep integration with legacy industries.
Returning to that GitHub curve: monthly active developers fell from 45K to 23K—superficially signaling industry contraction. Yet the cohort that remains features a record-high share of established developers—converging on ecosystems with real users—and being repriced by the AI industry in unprecedented ways.
As AI scaling confronts structural bottlenecks—compute aggregation, agent autonomous payments, verifiable data and decisions, privacy-preserving coordination—these builders stand at the crypto-AI intersection. Their long-honed sensitivity to rules, incentives, and authenticity is gradually crystallizing into AI-era scarcity: systemic-level capability.
As an investment firm deeply rooted in crypto infrastructure since 2017, IOSG’s perspective goes beyond observation. We invested in EigenLayer’s restaking mechanism before it gained broad market recognition; led Ingonyama’s (now MoonMath) seed round—betting on ZK hardware acceleration’s migration to AI performance layers; and backed Hyperbolic in 2024, endorsing its path to solve decentralized compute trust using crypto-native verification mechanisms.
The unifying logic behind these bets is clear: AI’s scaling challenges around trust, coordination, and verification will ultimately require the mechanism-design expertise cultivated in crypto. We believe the convergence of crypto and AI is not narrative—it is a structural opportunity unfolding in real time.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
IOSG Ventures
