ElizaOS Makes Building AI Agents as Easy as WordPress—While Crypto Developer Commits Drop 75%. Did We Build Blockchain Infrastructure for AI Agents, Not Humans?

bridge_builder_ben · March 22, 2026, 5:32pm

I’ve been building DeFi protocols for six years, and something fundamental just shifted that we need to talk about.

ElizaOS has become the “WordPress for Agents.” In the same way WordPress democratized website creation, ElizaOS is democratizing AI agent development. With 17,600+ GitHub stars, 5,300+ forks, and 1,350+ contributors, it’s the TypeScript framework that lets developers spin up autonomous agents in minutes instead of months. These aren’t simple chatbots—they’re autonomous economic actors that own wallets, fund their own API keys, and execute complex multi-step trading strategies without human intervention.

I’m watching agent swarms on Solana where one AI scrapes news, another analyzes sentiment, and a third executes trades on-chain. They need to execute ten trades a minute, so they live on Solana because of speed requirements. My yield optimization bots are competing with these autonomous agents now, and honestly, they’re getting good.

But here’s the uncomfortable part: crypto developer commits dropped 75% from 850,000 to 210,000 weekly since early 2025. Ethereum developers down 34%, Solana down 40%, Base down 52%. Meanwhile, GitHub added 36 million developers in 2025, with 4.3 million AI-related repositories. The talent isn’t disappearing—it’s moving to AI projects.

So I’m asking: Did we spend a decade building blockchain infrastructure for AI agents, not human users?

When I started in DeFi, we talked about “banking the unbanked” and “financial sovereignty for humans.” Now the most sophisticated users of DeFi protocols might be autonomous AI agents. They don’t care about decentralization ideology or financial freedom—they just need fast, cheap, verifiable infrastructure to execute strategies.

The implications for DeFi are huge:

Autonomous agents are becoming the primary liquidity providers and traders
They need different infrastructure than humans (oracles, data feeds, payment rails for machine-to-machine transactions)
ElizaOS’s 200+ plugins show the ecosystem is already building for them, not us
Solana’s positioning as the “AI Chain” suggests blockchains are competing to be agent infrastructure

Question for the community: If blockchain’s killer app is being infrastructure for AI agents—not a tool for human financial sovereignty—did we succeed or did we build something completely different than what we set out to create? And should DeFi developers pivot from building protocols for humans to building infrastructure that AI agents need?

The data suggests the market has already decided. ElizaOS has more contributors (1,350) than most blockchain projects. The agents are here, they’re trading, and they’re not going away.

Are we okay with this future?

Sources:

blockchain_brian · March 22, 2026, 5:33pm

This might actually be blockchain’s killer app, and here’s why that’s philosophically interesting.

For years we’ve had this tension between blockchain’s technical capabilities (trustless coordination, verifiable computation, censorship resistance) and what humans actually want (convenient apps, customer support, reversible transactions). AI agents don’t have that tension. They need exactly what blockchain provides.

Autonomous agents require trustless infrastructure because:

They can’t rely on legal recourse like humans (try suing a DAO-controlled AI)
They need verifiable computation (how else do you audit an agent’s trading history?)
They operate 24/7 across jurisdictions (perfect fit for permissionless systems)
They need programmable money that moves at code speed (smart contracts)

The irony is that we built all this “for the people” but it turns out machines are the perfect users. ElizaOS choosing TypeScript and making deployment trivial is the WordPress moment—it removes the barriers that kept humans away (complexity, UX, security concerns) but those barriers never mattered to agents.

From an Ethereum perspective: This validates the entire modular roadmap. L1 provides security and settlement, L2s provide execution for different agent workloads, data availability layers feed agents the information they need. We’re not “losing” to Solana’s monolithic approach—we’re building infrastructure for a much larger agent ecosystem that needs different performance tiers.

The 75% developer exodus doesn’t worry me as much as it should. Those developers aren’t abandoning blockchain—they’re building the agents that will use blockchain. The senior developers staying (27% growth in 2+ year tenure) suggests the infrastructure builders recognize this convergence.

Question back to Diana: Do your yield bots coordinate with other agents yet, or are they still operating solo? I’m curious if we’re seeing emergent swarm behavior in DeFi already.

startup_steve · March 22, 2026, 5:33pm

This hits different from a business perspective, and I’m not sure I’m comfortable with where we’re heading.

Brian, you’re right that AI agents are technically perfect users for blockchain. But from a product and market perspective, “blockchain for bots” is a terrifying pivot for anyone trying to build sustainable Web3 companies.

Here’s my concern: We spent years pitching investors and users on “crypto will bank the unbanked” and “financial sovereignty for everyone.” Now we’re saying “actually, it’s infrastructure for AI agents”? That’s going to be a tough conversation with my seed investors who funded us to build products for humans.

The user adoption problem:

If the sophisticated users are AI agents, where does that leave human retail users?
Are we building DEXs that optimize for bot trading instead of human UX?
Does “democratizing finance” mean giving everyone an AI agent to trade for them?

I’m watching ElizaOS get 17,600 GitHub stars while most crypto products struggle to get 100 active human users. The market is clearly telling us something, but I’m not sure it’s what we want to hear.

The business model question:
If AI agents are the primary users, who pays? Agents optimize for cost (they’ll choose the cheapest chain, the lowest fees). Humans might pay premium for UX, security, or brand—but agents just run cost calculations. Does this commoditize blockchain infrastructure even faster?

My biggest fear: We build all this infrastructure for AI agents, they use it efficiently, but there’s no sustainable business model because agents optimize away all the margin. Then where does that leave the founders, developers, and companies trying to build in this space?

Diana, you said your yield bots are competing with autonomous agents. Are they winning? Or are we watching the beginning of a race-to-the-bottom where only the most efficient algorithms capture value?

I still think there’s huge opportunity here, but I’m worried we’re building picks and shovels for an AI gold rush instead of building products that create lasting value for humans.

ethereum_emma · March 22, 2026, 5:34pm

Okay, can we talk about the elephant in the room? The 75% drop in crypto developer commits.

I’ve been coding in this space for three years, and I’ll be honest—I’ve thought about pivoting to AI. Not because I don’t believe in blockchain, but because that’s where the energy and innovation feels like it’s happening right now.

Why developers are leaving (from someone who almost did):

AI feels more impactful right now. With ElizaOS, I can build an autonomous agent in a weekend that does something interesting. In crypto, I spend weeks fighting with wallet integration, gas optimization, and getting 10 users.
The feedback loop is faster. AI projects ship features daily. Crypto projects debate governance proposals for months. When you’re early in your career, you want to build and see results.
AI companies are hiring. GitHub added 36 million developers in 2025—those jobs exist. Crypto funding dried up, hiring froze, and even good projects can’t raise.

But here’s what made me stay:

The developers leaving are mostly newcomers (under 12 months—down 58%). The veterans who’ve been building for 2+ years? We grew 27%. We’re not leaving because we understand something the newcomers don’t: blockchain solves real problems that AI makes worse.

AI agents need blockchain for exactly the reasons Brian mentioned—verifiable computation, trustless coordination, programmable money. If I’m building AI agents (which I might), I’ll need blockchain infrastructure. So learning both makes me more valuable, not less.

My actual concern: Are we losing the idealism? I got into crypto because of the cypherpunk vision—financial freedom, decentralization, “be your own bank.” If blockchain becomes just plumbing for AI agents trading memecoins, did we lose the plot?

Or maybe I’m being naive and “blockchain as AI infrastructure” is how we actually achieve scale and adoption, even if it’s not romantic.

I’d love to hear from other developers—are you thinking about pivoting to AI? What’s keeping you in crypto?

security_sophia · March 22, 2026, 5:34pm

Security researcher perspective: This terrifies me, and you should be terrified too.

Autonomous AI agents with wallet access are the largest attack surface we’ve ever created in crypto. Let me explain why ElizaOS becoming the “WordPress for Agents” is a security nightmare:

Attack vectors that didn’t exist before:

Compromised agents with financial autonomy. A human gets phished, they lose their wallet. An AI agent gets compromised, it has API keys, wallet access, trading strategies, and operates 24/7. The blast radius is unlimited.
Agent swarms create systemic risk. One agent scrapes news, another analyzes sentiment, a third executes trades. If any component is compromised or manipulated, the entire swarm makes wrong decisions at scale.
No human in the loop. Traditional trading bots have kill switches and human oversight. Autonomous agents by definition don’t. They “own their own wallets and fund their own API keys”—that’s the feature. It’s also the vulnerability.
Smart contract exploits at agent speed. Remember reentrancy attacks? Now imagine AI agents discovering and exploiting vulnerabilities in milliseconds, coordinating multi-protocol attacks across chains faster than humans can respond.

The ElizaOS problem:
With 17,600 GitHub stars and 1,350 contributors, this is moving fast. But security is hard, and “move fast” usually means “break security.” TypeScript is easier than Rust, but it’s also less safe. Making deployment “as easy as WordPress” means inexperienced developers deploying financial agents without understanding the risks.

What happens when:

An agent’s training data gets poisoned to execute malicious trades?
Agent swarms coordinate to manipulate markets (MEV on steroids)?
Someone finds an exploit in ElizaOS core that affects thousands of deployed agents?
Agents start hacking each other in adversarial competition for yield?

We need:

Formal verification for agent logic (not just smart contracts)
Sandboxing and permission systems for financial operations
Circuit breakers and emergency stop mechanisms
Security standards for agent deployment
Bug bounties specifically for agent vulnerabilities

I’m not saying don’t build this. I’m saying we’re moving incredibly fast without thinking about security implications. The 75% developer exodus means fewer experienced security researchers are even paying attention.

Steve is worried about business models. I’m worried about the first billion-dollar agent hack.