Weekly code commits to crypto and blockchain projects have fallen hard since early 2025, just as AI is pulling developers, funding and attention into a different fight. That should worry you more than another round of token price noise.
Ask a computer science graduate in 2026 where the money and the momentum are, and the answer isn't Solidity anymore. It's AI. Weekly code commits to crypto and blockchain repositories have fallen from roughly 850,000 a week to about 210,000 since early 2025, a 75 percent drop, according to Artemis data reported by CoinDesk. Active developers working on blockchain projects fell 56 percent over the same stretch, to around 4,600 people. That's not a quiet patch. That's an exodus.
The destination is obvious. CoinDesk also reported that GitHub added 36 million new developers in 2025, pushing its global base past 180 million, while platform-wide commits rose roughly 25 percent year over year. Repositories built around Jupyter Notebooks, the everyday workspace for machine learning, grew about 75 percent. Dockerfile repositories tied to AI applications jumped roughly 120 percent. The same engineers who once shipped smart contracts are still writing code. They're just pointing their nights, weekends and seed-stage ambition somewhere else.
Individual chains show the damage up close. Ethereum's weekly active developer count dropped 34 percent over three months to 2,811. Solana fell 40 percent, to 942 developers as of March, according to the Artemis figures cited in the report. Base, Coinbase's layer 2 network, lost 52 percent of its developers, down to 378. These aren't obscure projects tucked away on the edge of the market. They're the networks crypto has been using to sell its 2026 institutional story.
You should not shrug at that.
Open-source developer activity is not the same thing as market value, and Bitcoin has never needed a GitHub popularity contest to move. But blockchains are software systems before they're financial narratives. Someone has to review cryptography, patch clients, audit bridges, maintain wallets and write the boring infrastructure that stops a clever exploit from becoming a front-page disaster. If the bench gets thinner while the attack surface gets larger, the risk does not politely wait for the next bull market.
The age profile makes the shift sharper. Crypto.news reported on research showing that software developers aged 22 to 25 have lost nearly 20 percent of their employment since late 2022, the period when generative AI tools went mainstream, while older developers at the same companies saw employment grow. Blockchain leaned heavily on young, self-taught engineers who arrived during the 2021 run with GitHub handles, Discord reputations and little patience for old finance. That was part of the charm. It was also part of the labor model.
Now AI is taking the easy pitch. It has more venture funding, more corporate buyers and a clearer route into normal software budgets. A founder can sell an AI coding assistant, customer service agent or data workflow to a company that already has the problem and already has the budget. Try selling a new chain to that same buyer in 2026 and you spend half the meeting explaining why it isn't another token scheme. Frankly, that matters.
Here's the part that should worry crypto more than any commit chart. In May, Solana co-founder Anatoly Yakovenko warned publicly that AI could break the post-quantum cryptography blockchain networks are racing to adopt, and that it could get there before quantum computers do. His point was not that quantum risk has disappeared. It was that newer signature schemes may contain mathematical or implementation weaknesses the industry does not yet understand well enough. AI tools are getting good at searching ugly problem spaces, and cryptography has plenty of ugly problem spaces.
Yakovenko's proposed answer was practical: use a 2-of-3 multisig setup spread across different signature systems, rather than betting a wallet on one cryptographic scheme. He also warned that Ethereum's layer 2 networks are not quantum safe. You don't have to accept every part of that argument to see the larger issue. Crypto is losing the people who harden systems at the same moment the tools for probing those systems are getting stronger.
There is one wrinkle worth keeping in the piece because it stops the story from becoming too neat. Not every measure says Solana is shrinking. SolanaFloor and KuCoin both reported that Solana's broader developer count rebounded through the first half of 2026, climbing past 10,700 active contributors by mid-year by some counts, after the network pulled in more new developers in 2025 than Ethereum. So this is not a simple obituary for crypto development.
It is a warning about where the deepest talent is choosing to spend its time.
The industry can live with lower token prices. It has done that before. It can live with regulatory fights, slower retail adoption and another round of loud conferences promising quieter infrastructure. What it cannot live with is a thinning group of builders being asked to secure more money, more bridges, more layer 2s and more cryptography against attackers using better tools every month.
If crypto wants developers back, it has to offer more than ideology and upside. It needs real security work, real users and projects that look worth spending a career on. AI is giving engineers the simplest pitch in technology right now: build here, and people will use it. Crypto has to answer that with more than a token chart.
Also read: Shanghai Bets on a New Quantum Computing Zone to Beat Rival Chinese Cities, Singapore Will Fine Cloud Giants a Million Dollars for Going Dark, Sakana AI's Ren Ito Joins the UN's New AI for Good Commission
