■ HIGH RISK ■ Technology
AI now writes competent Solidity, which devalues the scarcity premium that made blockchain salaries silly. But smart contracts are code where bugs are unrecoverable thefts — so the paid work concentrates in security, auditing, and architecture, while boilerplate contract-writing goes to the prompt.
“AI writes smart contracts. Your Solidity skills are getting cheaper by the prompt.”
Our AI replacement risk score — how we score jobs
The job in practice: designing and writing smart contracts (Solidity or Rust, mostly), building protocol logic for DeFi, NFTs, or tokenized assets, writing exhaustive tests because deployed contracts are immutable and hold real money, integrating frontends via wallet and node infrastructure, optimizing gas costs, and — the part outsiders underrate — obsessing over security, since the industry's history is a museum of nine-figure exploits caused by one wrong line.
AI hits this niche twice. First, as a code generator: LLMs produce workable token contracts, staking logic, and standard DeFi patterns from prompts, eroding the premium once paid for scarce Solidity fluency — a large share of contract code was always boilerplate assembled from audited libraries anyway, which is exactly what LLMs are best at emitting. Second, as a market force: the sector's demand swings violently with crypto cycles, and hiring booms that once absorbed mediocre developers now don't, because a smaller senior team with AI assistance ships the same protocol. AI-powered analysis tools are also automating chunks of the audit checklist.
The countervailing physics: this is the one domain where 'the code mostly works' is a catastrophe. Immutable deployment plus adversaries who are financially motivated, sophisticated, and instant means the scarce skill isn't writing Solidity — it's reasoning about what a hostile genius can do to it. LLM-generated contracts plausibly increase demand for human security review, since confident-looking vulnerable code is now cheap to produce. Auditing, protocol architecture, formal verification, and cross-chain infrastructure remain hard, senior, and well paid. Our 55 nets out to: the commodity tier of this specialty deflates, the security tier appreciates, and the whole thing rides crypto's boom-bust cycle either way.
Automatability: our editorial assessment of current and near-term AI capability
The commodity squeeze is current: LLMs already generate standard contracts, and teams ship with fewer, more senior developers than the last boom absorbed. Through around 2030, expect entry-level contract-writing roles to keep evaporating while security, auditing, and architecture demand holds or grows — cheap AI-generated code arguably feeds the audit market. Overlay crypto's boom-bust cycle on all of it; that cycle remains the bigger employment variable.
At the senior and security end, yes — auditors and protocol architects still command strong rates because mistakes cost millions instantly. The formerly lucrative junior tier is deflating: AI writes the standard contracts that entry-level developers were paid scarcity premiums for. Add crypto's boom-bust hiring cycle, and the honest answer is: lucrative, narrow, and volatile.
It writes plausible smart contracts, which is the dangerous part. LLMs assemble standard patterns well but also produce confident code with subtle vulnerabilities — and in this domain, one subtle vulnerability is an unrecoverable theft. That's why AI generation is increasing rather than decreasing demand for skilled human security review. Never deploy generated contract code unaudited.
Security, unambiguously: exploit patterns, audit methodology, fuzzing and formal verification tools, and EVM-level internals. That's where scarcity moved when contract-writing commodified. Secondarily, keep broad backend engineering skills current — the sector's hiring swings with crypto cycles, and developers who can exit into general software or security roles ride out the winters.
Partially — AI analysis tools already catch known vulnerability classes and are absorbing the checklist layer of audits. But audits exist to catch what checklists miss: novel attack vectors, economic exploits, and cross-contract interactions that require adversarial imagination. Cheap AI-generated code flooding the ecosystem plausibly means more audit demand, concentrated on the genuinely hard findings.