Crypto Commits Down 75%, Devs Moving to AI—Did Blockchain Just Lose the Talent War or Is This Natural Selection?

General
1

I’ll be honest—when I saw the latest GitHub stats this week, it stung a bit. Weekly crypto code commits down 75% since early 2025. Active blockchain developers dropped 56%. Meanwhile, GitHub added 36 million developers overall and platform-wide commits are up 25%. The numbers don’t lie: developers are choosing AI over crypto, and it’s not even close.

The Brutal Reality

Ethereum lost 34% of its weekly active developers in just three months—down to 2,811. Solana dropped 40% to 942 developers. Even Base, which was one of 2024’s hottest Layer 2 ecosystems, shed 52% of its developer base down to 378. At the same time, AI-related repositories on GitHub exploded to 4.3 million, with repos importing LLM SDKs surging 178% in a year.

As someone trying to raise a pre-seed round right now, I’ve felt this shift firsthand. Six months ago, VCs wanted to hear about our Web3 infrastructure play. Last week, three investors asked if we’d “considered pivoting to AI agents” before I even finished my pitch deck.

But Here’s the Thing…

Before we write blockchain’s obituary, there’s another way to read this data that actually gives me hope. Yes, we lost a lot of developers—but WHO did we lose?

The exodus is almost entirely part-time contributors and developers with less than 12 months of experience. That cohort dropped 58%. But developers with 2+ years of blockchain experience? They grew 27% year-over-year and now produce roughly 70% of all commits.

Think about that. Fewer developers, more commits from experienced builders. That’s not a talent drain—that’s a shake-out of tourists who came for the bull market and left when the work got hard.

The AI Product-Market Fit Question

Look, I get why developers are moving to AI. I’ve played with Claude and GPT-5 myself—the feedback loop is instant, the use cases are obvious, and you can ship something people actually use in a weekend. Compare that to blockchain where you spend three weeks debugging a smart contract for a use case that maybe ten people understand.

AI has product-market fit right now. Crypto is still searching for it. That’s a tough pill to swallow, but it’s real talk.

The question isn’t “why are devs leaving?” It’s “what are we building that’s worth staying for?” Because the developers who remain—the ones grinding on Layer 2 security, building actual useful DeFi protocols, solving real cross-chain problems—they’re here because they believe we haven’t solved the hard problems yet.

The Startup Lens

Here’s my sports analogy (sorry, can’t help myself): This is the fourth quarter and the bandwagon fans left when we were down by 14 points. The real fans—the ones who’ve been here through three losing seasons—they’re still in the stands. And honestly? That’s the team I want to build with.

When I’m pitching investors now, I lead with this narrative: “Yes, we lost the hype developers. Good. Now let’s build something sustainable with people who actually give a damn about decentralization, not just their token allocation.”

So What Do We Do?

For those of us still building in this space—whether you’re a founder, a core contributor, or just someone who believes in what blockchain can do—I think we need to be honest about a few things:

  1. Developer experience matters. If we want talent back, our tooling needs to be as good as AI’s. Period.

  2. Real problems, real users. We need to stop building DeFi protocols that only serve other DeFi protocols. Where are the blockchain apps my parents could actually use?

  3. Sustainable business models. VCs are tired of “we’ll figure out monetization later.” If we want serious builders, we need serious business models.

What do you all think? Is this developer exodus a crisis or a healthy consolidation? Are we witnessing blockchain’s slow death or the natural selection of builders who are actually committed?

Because from where I’m sitting in Austin, trying to build a Web3 company with a small but incredibly talented team, I’m betting on the latter. But I’m curious what the rest of this community thinks.

Sources:

2

Steve, I appreciate the optimism, but I think you’re still framing this like it’s a competition we’re losing. Let me offer a different technical perspective.

Quality Over Quantity in Protocol Development

Here’s what non-technical observers miss: blockchain protocol development doesn’t scale linearly with developer count. Building a secure consensus mechanism or a zero-knowledge proof system requires a small team of deeply specialized engineers, not an army of generalists.

When I see “75% fewer commits,” I don’t panic. I ask: what KIND of commits are we losing? The answer is mostly DeFi fork #47, NFT marketplace clone #183, and half-finished tutorials that were never going to ship.

Core Ethereum development—the actual consensus layer, EVM improvements, cryptographic research—hasn’t slowed down at all. In fact, the signal-to-noise ratio has improved dramatically. We’re shipping more meaningful upgrades with fewer people bikeshedding in the GitHub issues.

AI Attracts Generalists, Blockchain Needs Specialists

The developers moving to AI are mostly application-layer engineers who can import a LangChain SDK and build something in a weekend. That’s totally fine work, but it’s not the work that blockchain needs right now.

We need cryptographers. We need distributed systems researchers. We need engineers who can spend six months optimizing a zkEVM circuit. These people don’t jump ship for the next hype cycle because they’re solving genuinely hard problems that AI hasn’t touched.

I still see incredibly high-quality contributors in every protocol I work with. The Ethereum Foundation grantees, the L2 research teams, the cross-chain bridge architects—they’re all still here, still building, still pushing the boundaries of what’s possible.

Infrastructure Maturation, Not Decline

Think about Linux kernel development. It doesn’t need 50,000 developers—it needs a few thousand deeply experienced maintainers. The same is true for blockchain infrastructure now. We’re past the “greenfield experimentation” phase and into the “hardening and optimization” phase.

Fewer developers building 100 different L1s is good. More developers concentrating on making Ethereum, Solana, and a handful of serious chains actually work at scale is healthy consolidation.

This isn’t an exodus—it’s maturation. We don’t need 10,000 developers forking Uniswap. We need 1,000 developers solving cross-chain interoperability, optimistic fraud proofs, and decentralized sequencers.

The tourists left. The builders stayed. From where I sit in Dublin, working with core protocol teams across three continents, I’m not worried at all.

3

Okay, I’m going to be really honest here because this topic has been weighing on me for months.

I’ve thought about leaving blockchain development for AI. Like, seriously considered it. Not just passing thoughts—I spent two weekends building AI projects to see if I should make the switch.

The Developer Experience Gap Is Real

Brian, I hear you on the “quality over quantity” argument, and you’re probably right about protocol development. But from the trenches of actually building DeFi applications day-to-day? The developer experience in blockchain is still frustrating as hell.

Last month I spent three entire days debugging why a transaction that worked perfectly in testing kept failing on mainnet. Turned out to be a gas estimation issue with a nested contract call. Three days. In that same time, I built and deployed an AI chat interface that’s now being used by actual users.

The feedback loop matters. With AI development:

  • Write code → Run it → See results in 20 seconds
  • Something breaks → Stack trace tells you exactly what
  • Deploy → Users can interact immediately

With blockchain:

  • Write code → Compile → Deploy to testnet → Debug → Redeploy → Debug again → Deploy to mainnet → Hope nothing breaks
  • Gas optimization hell
  • Frontend integration is still way too complicated
  • Every mistake costs real money

Why I’m Still Here (Barely)

So why haven’t I left yet? Honestly, I’m not 100% sure sometimes. Part of it is sunk cost—I spent two years learning Solidity and Web3 integration. Part of it is that I genuinely believe in the mission of decentralized finance.

But here’s the thing that keeps me up at night: Are we losing developers because our tooling isn’t good enough?

Everyone talks about “real world use cases” and “product-market fit,” but maybe the problem is simpler than that. Maybe developers are leaving because writing smart contracts in 2026 feels like writing web apps in 2006—technically possible but unnecessarily painful.

A Challenge for the Community

If we want to retain developer talent, we need to make blockchain development as enjoyable as AI development. That means:

  • Faster iteration cycles - Why does deploying to a testnet take 5 minutes?
  • Better error messages - “Transaction reverted” tells me nothing
  • Modern testing frameworks - Foundry helped, but we need more
  • Simpler frontend integration - Web3 onboarding is still terrible for new users

Brian’s right that we don’t need 10,000 developers building Uniswap forks. But we DO need to make it easier for the good developers who stay to actually build things without wanting to throw their laptops out the window.

I’m still here. Still building. Still learning. But I totally understand why my friends from bootcamp chose AI over crypto—they wanted to ship products, not fight with tooling.

Maybe this exodus is the wake-up call we needed to fix our developer experience?

4

Let me dig into the numbers a bit more here, because I think both Steve and Brian are partially right—but the full picture is more nuanced than either extreme.

Breaking Down Where Developers Are Actually Going

I pulled some additional data to understand the migration patterns. Here’s what I found:

GitHub repository growth by category (2025-2026):

  • Jupyter Notebooks (ML experimentation): +75%
  • Dockerfile repos (AI deployment): +120%
  • Repos importing LLM SDKs: +178%
  • Smart contract repos: -42%

So yes, developers are clearly moving to AI infrastructure, LLM applications, and ML tooling. But WHERE in the stack?

Most of the growth is in application-layer AI development—chatbots, document analysis, code generation tools. The kind of stuff you can build with OpenAI’s API and get users immediately.

The developers we’re losing aren’t the cryptographers working on zkEVM circuits. They’re the full-stack engineers who could build on blockchain OR build on AI—and AI offers faster feedback loops, clearer monetization, and immediate user validation.

The Veteran Developer Growth Is Actually Remarkable

Steve mentioned that developers with 2+ years experience grew 27% year-over-year. Let me contextualize that number.

If we do the math:

  • Total active developers: -56% (from ~10,400 to 4,600)
  • Veteran developers (2+ years): +27%
  • Veteran % of total commits: ~70%

That means the remaining veteran developers are producing significantly more output per capita than the aggregate workforce did before. We’re not just retaining experienced devs—they’re being MORE productive.

This pattern mirrors what I’ve seen in previous tech cycles:

  1. Mobile 2010-2012: Web developers rushed to iOS/Android, many bounced back when they realized native mobile development was harder than they thought
  2. Cloud 2014-2016: Everyone wanted to be a “cloud engineer,” most returned to their previous stacks after the hype settled
  3. Blockchain 2021-2022: DeFi summer pulled a ton of tourists, most left during the bear market

Cultural Perspective: Regional Differences

One thing that’s interesting from my conversations with Korean and Asian developer communities: blockchain development is still seen as prestigious and cutting-edge in those regions. The exodus is concentrated in Western markets where AI is getting massive VC attention and media coverage.

When I visit family in Korea, blockchain engineers are still treated like rockstars. In Seattle, my non-tech friends ask “why aren’t you working on AI?”

Cultural momentum matters for developer retention.

The Question Nobody’s Asking

Here’s what I want to know: What percentage of developers who left blockchain during previous hype cycles (ICO boom, DeFi summer) eventually came back?

If the answer is 30-40%, then this “exodus” might actually be a net positive in 12-18 months when some of these AI developers realize they’re building the 10,000th chatbot wrapper and want to work on harder problems again.

Emma’s right that developer experience matters. Brian’s right that quality matters more than quantity. Steve’s right that we need sustainable business models.

But the data suggests this is a normal tech cycle pattern, not a death spiral. The developers who stay through the hype-to-reality transition are the ones who build the next generation of infrastructure.

My money’s on the 4,600 experienced developers who are still here, not the 6,000 tourists who left for greener pastures.

5

This conversation is fascinating, but I think we’re all circling around the real issue without naming it directly.

Developer Count Follows Real-World Impact, Not Vice Versa

Everyone’s debating whether losing developers is good or bad, but that’s the wrong question. The question is: Why are developers choosing to work on AI instead of blockchain?

And the answer is brutally simple: AI is solving problems that real people actually have, right now.

I can explain to my mom what ChatGPT does in 30 seconds and she immediately understands the value. I’ve been trying to explain DeFi to my former colleagues at environmental nonprofits for three years, and they still don’t see why they should care.

The Use Case Crisis

Let me share a recent experience. Last month I tried to pitch our Web3 sustainability protocol to a climate advocacy organization I used to work with. Smart people, tech-savvy, mission-driven—exactly our target users.

Their response? “Why do we need blockchain for this? Can’t you just build a regular database with a good API?”

And you know what? They were kind of right.

The features we were selling (transparency, immutability, decentralization) didn’t solve their actual pain points (grant management, volunteer coordination, impact measurement). We were building technology in search of a problem, while AI developers are building solutions to problems people already know they have.

Where Blockchain Actually Matters

Here’s the thing—I still believe blockchain has real use cases. But they’re not where most of the developer activity was concentrated during the bull market.

Real blockchain use cases I’ve seen work:

  • Supply chain verification for carbon credits (actually prevents fraud)
  • Cross-border payments for migrant workers (actually cheaper than Western Union)
  • Digital identity for refugees (actually solves a massive humanitarian problem)

Where 80% of blockchain developers were building:

  • DeFi protocols that only serve other DeFi protocols
  • NFT marketplaces for speculative JPEGs
  • DAOs that are just glorified Discord servers with token voting

The developers leaving for AI aren’t abandoning “real” blockchain applications—they’re abandoning the 80% of crypto that was solving fake problems.

The Impact Measurement Question

Mike asked what percentage of developers come back after previous hype cycles. Here’s what I want to know: What if we measured success by user problems solved instead of developer count?

Because right now, AI development is:

  • Helping doctors diagnose diseases faster
  • Automating tedious work so people can focus on creative tasks
  • Making technology accessible to people who can’t code

Meanwhile, most blockchain development is:

  • Helping crypto traders make slightly more optimal swaps
  • Letting people speculate on digital collectibles
  • Building infrastructure for other infrastructure

A Challenge to the Community

Emma’s right that developer experience matters. Brian’s right that we need specialists. Steve’s right about sustainable business models. Mike’s right that this looks like a normal cycle.

But I think we’re all missing the point.

If we want developers back, we need to show Web3 solves problems people actually have.

Not “people who already understand blockchain.” Not “crypto natives who live on Discord.” Regular people with real problems.

Until we can do that, no amount of better tooling or VC funding will bring back the developers who left. Because at the end of the day, developers want to build things that matter.

And right now, to most people, AI matters and crypto doesn’t.

I say this as someone who’s betting my career on Web3 sustainability applications. But if we’re honest with ourselves about why developers are leaving, we might actually figure out how to build something worth staying for.