01 Phase 1 – IR Hardening

Production-grade deterministic execution core.

  • Determinism guarantees audit
  • Scheduler correctness & time-slicing
  • Snapshot + replay stability
  • Structured diff reliability
  • Performance profiling & memory tuning
  • CSS coverage expansion (core parity subset)
  • JS DOM mutation completeness (~30 core APIs)
Rune IR runtime stable for structured apps and internal products.
02 Phase 2 – Web Compatibility Maturation

Expand real-world app coverage.

  • HTML → IR translator stabilization
  • DOM shim → IR mutation bridge completion
  • Canonical read API completeness
  • Observer APIs (Resize / Intersection)
  • fetch + storage shims
  • Complexity threshold tuning
  • Real-site validation (HN, lite CNN, etc.)
  • CEF fallback transparency controls
70–80% of workflow-heavy web apps run in IR path.
03 Phase 3 – Agent-First Runtime

Make IR natively agent-operable.

  • IR snapshot serialization format
  • Checkpoint & replay engine
  • Action dispatch abstraction
  • Structured semantic labeling
  • Human-in-the-loop UI
  • Deterministic workflow editor integration
  • Diff visualization tools
Reliable, replayable agentic workflows on real apps.
04 Phase 4 – Arth Expansion

Deterministic execution without browser semantics.

  • Arth language v1 stabilization
  • HIR validation & auditing tools
  • Multi-backend compilation (VM / Cranelift / LLVM)
  • Sandboxed plugin model
  • Deterministic I/O model
  • Formal verification exploration
One language substrate, two execution levels.
05 Phase 5 – Native Substitutions & Performance

Replace browser bottlenecks with GPU-native primitives.

  • Chart library interception (Recharts / Chart.js)
  • Map rendering substitution
  • PDF rendering
  • Media pipeline integration
  • GPU widget toolkit
Faster than browser + structured agent access.
06 Phase 6 – Infrastructure Layer

Rune as platform, not runtime.

  • Rune Server dual-mode serving
  • Remote IR packages
  • IR over network protocol
  • Multi-device synchronization
  • Wayland compositor maturation
  • Native desktop shell apps
Full execution substrate beyond web.
Exploratory

Long-term direction.

These are research-stage directions – not committed roadmap items. They represent where the architecture could lead as the core stabilizes.

Formal determinism proofs
Agent-native UI protocols
LLM → HIR pipelines
IR as universal execution graph
WebMCP hybrid integration
FAQ

Frequently asked questions.

Is Rune a browser?

No. It is a deterministic IR runtime that can ingest web applications. It replaces the browser execution layer with structured intermediate representation.

Is this scraping?

No. Rune intercepts DOM mutations and translates them into structured IR. It is not a scraping layer – it is a runtime substrate.

Does it require site changes?

No, in web compatibility mode. Existing apps run without modification. No SDK, no opt-in required.

Is it production ready?

It is a working prototype being hardened. IR rendering is stable, JS execution works, and coverage is expanding.

Why not use Playwright or RPA?

Playwright automates a browser. Rune replaces the browser execution layer with structured IR. Agents operate on typed nodes, not pixels or DOM selectors.

Is Arth required?

No. Arth is optional – it is the deterministic execution tier for systems that need execution guarantees. Web compatibility mode remains available for existing apps.

More questions? Reach out at rune@splenta.com.