In a Brussels workshop last Christmas, software veteran Bernard Lambeau didn’t fire up his IDE. Instead, he prompted Anthropic’s Claude Code with a README sketching an idea for a new expression language. By Boxing Day, Elo was born—a complete, portable system compiling to JavaScript, Ruby, and SQL, complete with parser, type checker, standard library, CLI tools, and docs. Claude wrote every line; Lambeau steered the vision.

“In roughly 24 hours of collaboration, we built a complete expression language with a parser, type system, three compilers, a standard library, a CLI tool, and a documentation website. Not bad for a day’s work,” Lambeau and Claude declared in their joint Elo blog post. “Elo isn’t just a demonstration that AI can write code. It’s a demonstration that humans and AI can build together—each contributing what they do best.”

The feat, detailed in The Register, marks a milestone in AI-assisted development. Lambeau, PhD-holding founder of Klaro Cards and CEO of Enspirit, leveraged Claude’s Claude Max subscription—€180 monthly—to deliver what would have taken him weeks solo or months with a hire.

Elo’s Safe Core for No-Code Realms

Elo targets no-code platforms like Zapier or Klaro Cards, where non-experts craft data rules without risking exploits. Non-Turing complete by design, it enforces purity: no mutable state, no side effects, single equality operator across targets. “I’m absolutely convinced that we need better/safer/simpler programming languages inside no-code tools,” Lambeau told The Register. Mainstream tongues demand boilerplate for schema checks; Elo bakes it in.

Pipeline syntax shines: customers |> filter(active: true) |> size counts active users, chaining left-to-right for readability even sans coding chops, as noted in Korben.info. Fallback pipe handles nulls: user.nickname | user.firstname | "anonymous". Schemas validate: let Event = {name: String, date: Datetime, capacity: Int(c | c > 0)} in _ |> Event.

Temporal ops suit business logic—TODAY > signup + P30D checks trial expiry, compiling identically semantic across JS (DateTime.now().startOf('day') >= ...), Ruby (Date.today >= ...), SQL (CURRENT_DATE >= DATE_TRUNC...), per elo-lang.org.

From Prompt to Production: The Workflow

Lambeau’s method, documented in GitHub’s Elo repo CLAUDE.md and 100+ task MD files, started supervised: Claude sought permission every 20 seconds. Successes unlocked autonomy in Docker sandboxes or isolated VMs. “I review the plan, make sure we have a test strategy that’s sound, then switch Claude to autonomous mode,” he explained.

Feedback loops ruled: Claude penned tests, ran them, self-corrected. Acceptance tests live in Elo, compiling to all targets for cross-runtime proof. Modes included ‘plan’ for architecture, personas like ‘skeptic’ for bugs, per Korben.info’s 30-minute Loom demo. “Claude Code knows almost every tech stack… and does that 10x faster than I can,” Lambeau marveled.

CLI tools ease adoption: ./bin/eloc -e "2 + 3 * 4" -t ruby yields Ruby; evaluator handles JSON/CSV: ./bin/elo -e "map(_, fn(r ~> r.name))" -d @data.csv. NPM integration: import { compile } from '@enspirit/elo'; const addTen = compile('_ + 10');.

Lambeau’s Pedigree Powers the Pivot

With 30 years’ experience lecturing relational models at UCLouvain, Lambeau built precursors: Bmg for algebra, Finitio for schemas, Webspicy for API tests—all converging in Elo, now live in Klaro Cards for user-defined rules. Enspirit, his consultancy at enspirit.co, crafts AI-infused apps for Nike, Deem.

“Claude Code falls short if you don’t have a great methodology,” he cautioned. Experts thrive; novices risk “unmaintainable code.” Yet for pros, it’s “full-stack++”: tackling unfamiliar stacks seamlessly. Cost? Claude’s “PhD-level” knowledge transfer beats human hires.

Elo unifies his stack, eyeing no-code ubiquity. First integration: Klaro Cards, where non-techies tweak e-commerce, subscriptions sans peril.

Precedents and Ripples in AI Lang Labs

Lambeau joins pioneers: Steve Klabnik’s Rue, Geoffrey Huntley’s Cursed—both Claude-aided, per The Register. Hacker News at Elo thread debates portability: why multi-target for no-code? Readability wins, posters note—TODAY > signup + 30.days in Ruby vs. Elo’s ISO.

X buzz amplifies: The Register posted the story; shares from @emollick highlight Claude Code’s startup-site builds in 74 minutes. Founders like @aakashgupta warn: “If you don’t work Claude Code-native ASAP your team’s going to get left behind.” Teams ship 5 releases/engineer daily, per Anthropic insights.

Korben.info dubs it “une IA écrit un langage… sans intervention humaine,” stressing zero-trust safety: “quasiment impossible de faire une bêtise.” GitHub’s MIT-licensed repo invites TDD contributions via HACKING.md.

Broader Forces Reshaping Code Creation

Elo signals shift: AI as implementer, human as architect. Lambeau’s “lead-dev/CTO + QA role, btw; it’s just much faster” mirrors @JustJake’s Golang orchestrator in 4 hours, @alexolegimas’s 24-hour data analysis in 20 minutes. “Paradigm shift,” the latter tweeted.

Challenges linger: expertise-gated, per Lambeau. “You need very strong expertise to do it effectively.” Yet velocity compresses teams—Browser Company scales output, not headcount. No-code demands like Elo’s rise as non-techies wield AI code-gen.

Future? Relational extensions via Bmg for joins, aggregates. Playground at elo-lang.org/try beckons tinkerers. As Claude evolves, Lambeau’s blueprint—tasks, sandboxes, self-tests—may standardize human-AI forges, birthing safer data tools amid exploding low-code adoption.