For all the talk about the future of Web3, most of the industry has been stuck in a strange paradox: the most secure smart contract languages are also the hardest to write. Move is a perfect example. It gives developers an asset model grounded in formal reasoning—the kind of mathematical guarantees blockchain has always needed—but the complexity of that model has kept it firmly in the hands of experts.
For years, we’ve known what was missing: a way to make intelligent smart contracts possible without dumbing down the safety rules that keep decentralized systems trustworthy.
That gap is finally closing.
The University of Surrey and Endless Protocol have introduced Endless Move-LM, an open-source AI model built specifically for the Move programming language. To anyone outside the technical community, this may sound incremental. Inside the field, it represents a structural shift. We are moving from “AI as a helper” to AI as a participant in secure blockchain engineering.
This isn’t hype, and it isn’t a marketing phrase. It’s a reality we’ve been heading toward for over a decade, and now the tools have caught up.
Why Move Needed a Purpose-Built AI Model
Move is not another Solidity clone. It was designed to solve core problems: asset safety, predictable behavior, formal constraints around resource movement. Its architecture is elegant, but it doesn’t tolerate developer sloppiness—nor should it. If a programmer mishandles a resource rule, the entire project can break down.
The challenge is simple: The stricter the language, the steeper the learning curve.
Traditional AI coding assistants don’t understand Move. Not really. They pattern-match text and produce synthetic guesses. That works for JavaScript. It does not work for a language where asset semantics matter as much as syntax.
Endless Move-LM is the first model trained specifically on:
- Move’s resource logic
- Its module structure
- Common error patterns
- Real-world developer workflows inside the MoveVM
- Best practices from system-level codebases across multiple chains built on Move
It doesn’t “autocomplete” Move—it reasons about it.
The Bridge Between Intent and Security
One of the first things people misunderstand about blockchain development is how quickly intent becomes irrelevant. Smart contracts execute blindly. They don’t care what a developer meant. Only what was written.
This is where I see the deepest value in Endless Move-LM.
A developer can describe, in plain language, the behavior they want: a multi-asset escrow, a staking flow, a token with custom movement rules. The model translates that intent into well-structured Move modules, then immediately surfaces whether anything violates resource constraints or introduces unwanted side effects.
It compresses the feedback loop in a way previously impossible.
We’re not automating creativity; developers still decide what to build. We’re automating the “is this safe and correct?” layer that, until now, required hours of testing, trial-and-error, and one or two Move experts reviewing everything manually.
Why This Matters for Web3’s Next Phase
Web3 has been stuck in a pattern for years:
- Developers build applications.
- Auditors try to catch mistakes.
- Eventually, the weakest project in the ecosystem collapses and everyone suffers.
The stakes are too high for this cycle to continue.
When we talk about “AI-native Web3,” this is what we mean: systems where the intelligence is baked into the development environment itself, not bolted on after the fact.
Endless Move-LM integrates directly into Endless Chain’s infrastructure, giving developers:
- Automated error detection, especially around resource management
- Real-time security guidance
- Optimization suggestions that understand Move’s constraints
- Natural-language-to-code generation that doesn’t sacrifice correctness
This combination shortens development timelines, but, more importantly, it elevates the baseline of safety across an entire ecosystem.
From Smart Contracts to Autonomous On-Chain Agents
Where this ultimately leads is far bigger than “AI helping developers write faster.” As these models improve, three breakthroughs become possible:
1. AI-driven formal verification
Not as a separate phase—but as a continuous part of development.
2. Multi-chain intelligent behavior
AI that understands Move on different chains and adapts modules to their local configurations.
3. Autonomous agents that operate with enforceable constraints
Agents that can act, reason, transact, and adapt within the MoveVM’s security model—without exposing users to unpredictable behavior.
Every major blockchain innovation has followed the same pattern: first we build tools, then the tools build the ecosystem.
Endless Move-LM is one of those tools.
Lowering the Barrier Without Lowering the Standard
There’s a misconception in the blockchain world that lowering barriers always leads to security degradation. That’s only true when shortcuts replace discipline. The Move language was built to prevent shortcuts, and Endless Move-LM respects those constraints.
If anything, it enforces them more rigorously than most developers do on their own.
The result is a strange and welcome shift: Move becomes easier to use while becoming even safer in practice.
That kind of shift rarely happens in software engineering.
Where We Go From Here
The launch of the model doesn’t mark an endpoint; it marks the beginning of an entirely new research frontier.
Our next focus areas include:
- Multi-chain compatibility
- Automated, AI-assisted auditing
- High-trust autonomous agent
- Cross-language reasoning between Move and other formal, resource-secure environments
It took a long time for the AI world and the blockchain world to converge. They had fundamentally different cultures and priorities. But as smart contract ecosystems mature, intelligence isn’t just a bonus — it’s becoming a requirement.
If blockchain is going to power global-scale systems, the infrastructure needs to think, reason, and adapt alongside human developers.
Endless Move-LM is an early proof that this is not only possible—it’s already happening.
By Professor Yu Xiong, Founding Advisor, Luffa
Professor Yu Xiong is Chair of Business Analytics and Associate Vice-President at the University of Surrey, as well as co-founder and Chief Scientist of Endless Protocol, a Web3 startup that reached unicorn status in 2025. A recognized bridge-builder between academia, industry, and government, Xiong has helped launch over 40 startups, advised UK policymakers on emerging tech, and led major sustainability and innovation initiatives. His work offers a new model for translating academic research into global enterprise.
Luffa is a next‑generation social operating system for the fan economy, giving creators ownership over their communities while allowing fans to turn attention into tangible value. The platform unifies wallet, messaging, loyalty, and engagement in a decentralized environment: fans earn rewards for actions like chatting, tipping, minting tokens, joining “SuperGroups,” and completing quests—forming a living fan graph with real‑world worth. Luffa emphasizes privacy and security: it is built with end‑to‑end encryption and zero centralized backups, and supports mnemonic‑based registration without requiring phone or email.
Luffa runs on
This article is published under HackerNoon’s Business Blogging program.