Public roadmap

Where Sailfin is heading

Sailfin is young, moving quickly, and not being driven toward an arbitrary launch date. This is a reviewed view of the work and ideas that matter—not a promise that every item will land in a fixed order.

Public roadmap last synchronized July 24, 2026.

CompletedIn progressPlannedPaused

Delivered Foundations

Completed projects that the rest of the roadmap builds on. Shipped behavior remains documented in the language reference and status matrix.

Initiative

Build & Toolchain

  • CLI Follow-Ups — Phase D

    Complete the remaining CLI behavior and diagnostics needed for a coherent native Sailfin toolchain.

  • CLI Modularization

    Keep compiler command orchestration modular, testable, and separate from compilation policy.

  • Native sfn bench Command & sfn/bench Harness

    Provide a native sfn bench command and Sailfin benchmark harness for repeatable compiler and runtime measurements.

  • Native Toolchain Pinning & Dispatch

    Let a project declare its required sfn toolchain version natively in capsule.toml ([toolchain]) and, in phase 2, auto-fetch/dispatch to it — retiring .seed-version/Makefile plumbing for downstream projects.

Initiative

Effect System & Language Core

  • Effect Enforcement Foundations

    Deliver hierarchical effects, conservative detection, cross-capsule enforcement, and effect-transparent concurrency foundations.

  • Generics & Higher-Order Functions

    Provide generics and higher-order collection operations with monomorphized, statically checked behavior.

Toward 1.0

Work that currently defines Sailfin's stable 1.0 maturity boundary. This is direction, not a launch deadline or an immutable checklist.

Initiative

Accelerated ML

  • Autodiff & Training Graphs

    Add differentiation and training-graph semantics on top of the typed tensor IR and numeric substrate.

  • Capability-Sealed Model Artifacts

    Package models with verifiable provenance, declared authority, compatibility metadata, and enforceable execution boundaries.

  • Distributed Accelerator Concurrency

    Coordinate accelerator work and collectives through Sailfin’s structured-concurrency and capability model.

  • ML Acceleration — Foundations

    Track-A foundations for SFEP-0052: low-precision numeric substrate, a shape-typed tensor IR + fusion, a first substrate exit, and the first enforced ![gpu]/![model] checks.

  • sfn/ai Model Engines & Adapters

    Define native model-engine interfaces and adapters with explicit model effects, capabilities, and tensor boundaries.

Initiative

Build & Toolchain

  • Build Performance & Production-Readiness

    Make self-hosted builds faster, more deterministic, and production-ready across supported development and release workflows.

  • Compiler Frontend/Backend Capsule Boundary

    Split the compiler into stable frontend and backend capsule boundaries without prematurely fragmenting every compiler pass.

  • Compiler Hardening

    Harden parsing, type checking, captures, and diagnostics so the self-hosted compiler fails predictably on real programs.

  • Per-Module Compile-Time Peak RSS

    Measure and reduce peak compiler memory per module so production builds stay within a predictable resource envelope.

  • Production-Grade Self-Host Determinism Enforcement

    Guarantee deterministic compiler output and reproducible builds across supported hosts and repeated self-hosting runs.

  • sfn LSP & IDE Integration

    Provide a compiler-backed language server for diagnostics, navigation, completion, formatting, and editor integration.

  • Workspace Manifest & Multi-Capsule Policy

    Grow workspace.toml into a production-grade multi-capsule manifest (Cargo/Go/Zig parity) plus the workspace-tier capability envelope no peer has. Design: SFEP-0051.

Initiative

Capability-Sealed Runtime

  • End-to-End Capability Seal

    Seal executable capability declarations end to end so runtime authority matches compiler-checked effects.

  • Native Crypto + TLS Stack — Zero OpenSSL

    Remove the OpenSSL (-lssl/-lcrypto) dependency by building a pure-Sailfin crypto + TLS 1.3 stack — a prerequisite for the capability seal (SFEP-0016).

  • Native Runtime Ownership & Deterministic Destruction

    Complete deterministic destruction and ownership-aware runtime primitives needed for a memory-safe self-hosted runtime.

  • Owned Syscall Layer

    Route operating-system access through owned, capability-aware syscall boundaries implemented in Sailfin.

  • Per-Task Capability Context

    Carry capability and runtime context per task so concurrent work cannot silently inherit ambient authority.

  • Runtime Strings & Memory Safety

    Make string operations efficient and memory-safe across slices, concatenation, aggregate reads, and ownership-sensitive cleanup.

  • Seal-Sufficient Native Backend

    Complete the backend and syscall layer needed for capability sealing without hidden native-runtime escape hatches.

  • sfn/capability — Runtime Object-Capability Model

    Ship reusable runtime object capabilities that make authority explicit, bounded, and testable in Sailfin programs.

Initiative

Effect System & Language Core

  • Capsule-Defined Decorators

    Let capsules define statically resolved decorators with explicit typing, expansion, and effect behavior.

  • First-Class Function Values

    Make functions first-class typed values that can be stored, passed, returned, and called without losing effect information.

  • Information-Flow & Taint Enforcement

    Track sensitive data through model and capability boundaries so policy checks can prevent unintended disclosure.

  • Model Effect Enforcement

    Enforce ![model] declarations and call boundaries as part of Sailfin’s explicit effect system.

  • String Interpolation Syntax

    Add typed string interpolation with clear evaluation, formatting, allocation, and diagnostic semantics.

Initiative

Structured Concurrency

  • Concurrency Frontend Typing

    Complete frontend typing for await results, typed channels, task bodies, and structured-concurrency call sites.

  • Concurrency Maturity Phase 1 — Cancel-on-Fault Nursery

    Cancel sibling tasks when a nursery child fails, with bounded shutdown and deterministic error propagation.

Initiative

Test Infrastructure

  • Sailfin-Native Test Framework

    Provide the native sfn test framework, discovery model, assertions, fixtures, and compiler-owned execution contract.

  • Test Runner Performance & Reliability

    Harden test-runner reliability and improve measured performance before introducing parallel scheduling.

Later

Important approved directions that are not currently inside the 1.0 boundary. Later does not mean rejected, and work may move as the language evolves.

Initiative

Build & Toolchain

  • aarch64-Linux Target Support

    Bring the self-hosted compiler and installer to 64-bit ARM Linux, including ABI support, release artifacts, and on-device self-host verification.

  • Fine-Grained Compiler Sub-Capsules

    After the frontend/backend boundary is stable, split compiler passes into smaller internal capsules where evidence shows a maintenance or performance benefit.

  • Makefile Retirement — sfn Owns Build/Check/Bootstrap

    Retire the root Makefile by moving bootstrap, build, test, bench, and check orchestration into native `sfn` commands.

  • Native Windows Self-Host (MSVC ABI)

    Self-host Sailfin natively on Windows with the MSVC ABI, including bootstrap, linking, tests, and release packaging.

  • Runtime Diagnostics & Allocation Telemetry

    Add opt-in runtime diagnostics for allocation, ownership, and task behavior without changing release semantics.

  • WebAssembly Target

    Add a WebAssembly compilation target with an explicit runtime, capability, and packaging contract.

Initiative

Capability-Sealed Runtime

  • Full Unsafe Capability Enforcement

    Extend unsafe enforcement across the full call graph and runtime boundary after the current local checks are dependable.

  • Perf-Parity Native Backend

    Close measured performance gaps in the fully native backend after correctness and capability sealing are established.

  • Transitive Capability Audit & SBOM

    Audit transitive capsule authority and produce machine-readable capability provenance for builds and dependencies.

Initiative

Effect System & Language Core

  • Currency & Time Literals

    Design exact currency and time literal syntax with explicit units, precision, overflow, and interoperability rules.

  • Effect Polymorphism & Effect-Aware HOFs

    Let generic and higher-order code express, propagate, and constrain caller effects without erasing safety.

  • Pipeline Operator

    Design a pipeline operator that composes cleanly with function calls, generics, effects, and diagnostics.

  • User-Facing Ownership & Borrow Checking

    Expose ownership and borrowing rules in the language after deterministic runtime ownership is stable.

Initiative

Product & Adoption

  • Benchmark v2 — Adoption Readiness and Controlled Learnability

    Separate real-world adoption readiness from controlled learnability, then test external pull only after a credible technical signal.

  • External Credibility — Evidence Before Promotion

    Build trust through truthful claims, reproducible evidence, and one time-boxed outreach test before deciding whether to promote Sailfin more broadly.

  • MCP-Proxy Enablement

    Runtime and stdlib gaps blocking a pure-Sailfin MCP proxy. Track A (stdio, 4 pure-Sailfin wrappers) is the critical path; Tracks B and C4 serve the deferred HTTP/control-plane work.

  • REST-Server Dogfooding Gaps

    Close the language, runtime, and standard-library gaps needed to build and operate a basic REST server entirely in Sailfin.

Initiative

Structured Concurrency

  • Bridge the WebSocket Runtime

    Connect WebSocket operations to the native runtime and structured-concurrency model without hidden blocking behavior.

  • Concurrency Maturity Phase 2 — Async-I/O Reactor

    Replace thread-per-task blocking I/O with a portable reactor while preserving structured cancellation semantics.

  • Rich Synchronization & Atomics

    Add the synchronization and atomic primitives justified by measured concurrent workloads after core task semantics stabilize.

Initiative

Test Infrastructure

  • Advanced Test Modes

    Add opt-in property, fuzz, snapshot, and other advanced modes once the core native runner contract is stable.

  • Parallel Test Execution

    Execute independent tests concurrently with deterministic reporting, isolation, and configurable resource bounds.

  • Test Runner Ergonomics

    Improve test discovery, filtering, output, failure reporting, and day-to-day command ergonomics.

Exploration

Concepts worth preserving while their design, sequencing, or fit remains open. These are not delivery promises.

Initiative

Build & Toolchain

  • Notebook & Interactive Tooling

    Explore interactive execution, notebooks, and REPL-like workflows once compiler APIs and incremental behavior can support them honestly.

Initiative

Effect System & Language Core

  • Algebraic Effect Handlers

    Explore whether algebraic handlers complement Sailfin’s explicit capability effects without weakening predictability.

How this page works

Linear plans the work; this page explains the direction

Maintainers plan Initiatives and Projects in Linear. A Project appears here only after it has been deliberately marked for public publication and assigned a reviewed horizon. Horizons can change as designs mature, dependencies move, or development accelerates. Linear links and private planning context are never exposed on this page.

Contributing

Help improve Sailfin

Public GitHub issues are the right place for bug reports and concrete proposals. Significant language, runtime, ABI, or toolchain changes should be grounded in an SFEP before implementation.

Open an issue