Symbolic in. Real result out.

Watch a program collapse into meaning — then come alive and run. EML is a semantic overlay for humans and AI agents.

scroll ↓

The compression

A loop collapses into three symbols

The same computation, written twice. EML keeps the meaning and drops the ceremony — and it transpiles back, deterministically.

Python

N = 100
r = sum(i**2 for i in range(1, N+1))
print(r)

↓

EML

N^+100
Σ(i^2, i in [1:N]) => r
r^0

Not magic — a toolchain

One deterministic pass

Symbols become runnable, observable Python through a rule-based pipeline. No LLM in the core.

EML→

normalize→

lex→

parse → AST→

semantic→

emit→

Python→

run / trace

Execution is the interface

It actually runs. Right here.

No video, no mockup. The program below executes in your browser via EML’s execution-truth interpreter, emitting a phosphor-jsonl-v1 trace.

eml://run · phosphor-jsonl-v1running…

N^+100
Σ(i^2, i in [1:N]) => r
r^0

// scroll in to execute

stdout

Not a video. Not a mockup. Executed in your browser. · No backend, no local Python for this demo.

Open full Playground

Human × Agent

One artifact, two readers

Humans read the projection. Agents read the structure — AST, semantics, trace. EML is a shared language between them.

Human view

Σ(i², i∈[1:N])

A dense, readable projection.

Agent view

{ Sum: { expr: Power(i,2),
  range: [1, N] } }
→ eml:sum · eml:assign · eml:output

Structured AST + execution trace.

AI-native interface

Not just a website — a semantic node

The page you read is for people. The /ai/ surface you can call is for agents — same EML, two readers. A public, non-visual, machine-readable layer plus bounded tools any agent can invoke.

Human UI Layer

Landing, workbench, and the in-browser playground.

Machine Corpus Layer

/ai/ — the v1.0 spec, EBNF, AST / trace / error schemas, verified examples.

Agent Tool Layer

/ai/tools/* — parse, transpile both ways, interpret, trace, round-trip.

Call it from a terminal — live, deterministic

curl -s https://efficientnewlanguage.org/ai/tools/transpile-python \
  -H 'content-type: application/json' \
  -d '{"source":"N^+100\nΣ(i^2, i in [1:N]) => r\nr^0"}'

llms.txt /ai/ manifest.json OpenAPI

Try EML in the browser.

Open Playground Read the spec GitHub