Crypto Code Commits Sink 75% Since Early 2025—Is AI Draining Builders or Just Consolidation?

Data circulated on Thursday (March 12) and attributed to Solid Intel shows crypto code commits have fallen roughly 75% since early 2025, with active contributors down about 56% over the same window. Weekly commits that used to print 800,000 to 900,000 are now closer to 200,000. Weekly contributors peaked above 10,000, and now sit below 5,000. ^[2]

What the numbers actually say (and what they do not)

The headline is simple: fewer commits, fewer contributors, sustained decline through 2025 into 2026.

But dev activity metrics are messy:

• Commits are a proxy, not a product metric. A protocol can ship meaningful upgrades with fewer commits if work consolidates into fewer repos, or if the team shifts to larger, less frequent merges.
• Repo visibility matters. More teams are building in private repositories, security-first workflows, or closed source modules, especially for infra, MEV tooling, and compliance-heavy products.
• One commit is not equal to another. "Update README" counts the same as refactoring a consensus module in many public dashboards.

The AI gravity well: real talent drain or lazy narrative?

The cleanest explanation making the rounds is also the most obvious: AI is hiring. Every company wants to be an AI company, and the comp packages for strong engineers in LLM infra, agents, and applied ML have been aggressive. ^[3] If you are an engineer who likes hard problems and fast iteration, AI currently offers:

• clearer product-market pull,
• quicker feedback loops than governance-heavy crypto,
• and fewer "why is this token down 40%?" meetings.

So yes, some builders likely pivoted. The timing lines up with the post-2024 AI tooling boom and the professionalization of the space. Crypto engineering used to be the "frontier." AI now wears that hoodie.

But "everyone pivoted to AI" is also a convenient meme explanation that can hide a more boring reality: crypto is consolidating.

Consolidation is the quiet killer of commit counts

A big chunk of crypto's 2021 to 2024 dev activity was not "core protocol progress." It was:

• copy-paste forks,
• incentive-driven new L1s and L2s,
• grant farming,
• short-lived DeFi primitives,
• and endless wallet and bridge clones chasing liquidity.

Fewer experimental chains, more standard stacks

As rollups, modular stacks, and shared sequencers mature, teams build on common frameworks rather than reinventing an L1. That concentrates work into fewer upstream repos and reduces "noise commits" across hundreds of smaller projects.

Teams ship less in public

"Maintenance mode" is a phase, not a death sentence

None of this fully explains a 75% drop, but it does explain why "crypto is dead" is not the only reading. ^[4]

The real red flag: active contributors getting cut in half

Commit counts can be gamed. Contributor counts are harder to spin. The same dataset indicates weekly active developers fell from above 10,000 at the peak to under 5,000 now. ^[2]

That suggests at least one of these is happening:

1. Builder churn is real: people left, got rekt, or found better opportunities elsewhere.
2. Projects died: fewer teams are active enough to show up week to week.
3. Work moved off the radar: private repos, internal tooling, enterprise partnerships, or closed source commercialization.

If you are trying to price long-term crypto adoption, the uncomfortable takeaway is that the public builder base is smaller than it was, even if the remaining builders might be higher quality.

Why the timing makes sense (even without an AI exodus)

The decline from early 2025 into 2026 tracks with a few broad pressures:

• Funding discipline: Easy money cycles create more repos than users. When funding tightens, "ship tokens" teams disappear first.
• Regulatory overhead: Compliance work often reduces public code output. It also pushes startups toward safer, centralized, or enterprise-friendly architectures that are not always open source first.
• User growth fatigue: If new user onboarding is not compounding, teams cut burn. That shows up quickly in developer activity.

Crypto does not need every dev to stay. It needs enough devs to ship products people actually use. The problem is that dev declines often precede ecosystem stagnation, especially for smaller chains and long-tail DeFi.

Who should worry most?

This type of drawdown does not hit everyone equally.

• Long-tail L1s and niche L2s: most at risk. If dev activity is sliding across the sector, marginal ecosystems lose the talent war first. Liquidity leaves, then apps, then devs, then the chain becomes a ghost town.
• Infra with real revenue: more resilient. If you already have fee streams, enterprise contracts, or strong distribution, you can keep core teams staffed even as the broader market shrinks.
• Ecosystems that rely on grants: vulnerable. When incentives slow, commits often slow with them.

What to watch next (no cope version)

Developer activity is a lagging indicator until it suddenly is not. The next few months should clarify whether this is a bottoming process or a deeper decay.

If weekly commits stabilize around the 200,000 level and contributor counts stop falling, watch for a "quality over quantity" narrative to become real, with fewer but stronger repos driving releases.

If contributor counts keep sliding below 5,000 and commits keep grinding down, expect more ecosystem consolidation, more abandoned chains, and a harsher market for anything that needs constant shipping to stay relevant.

Either way, the AI-versus-crypto framing is only half the story. The other half is simpler: the builder economy is maturing, and the projects that cannot justify their existence without constant hype are getting quietly sunset.