Existing user interfaces, document formats, and programming languages were engineered around human sensory perception and visual layout. The STF ecosystem replaces visual heuristics with explicit, deterministic semantics encoded directly into the grammar.
HTML, CSS, DOM trees, Markdown, and traditional source syntaxes evolved for human eyes. When an AI agent interacts with a modern web application or maintains a large system over long sessions, it is forced to burn tokens reconstructing intent from visual layout tags, deeply nested punctuation, untrusted string concatenations, and flat chat logs.
<div> hierarchies and JavaScript bundles, an agent reads explicit intent, roles, actions, and navigation targets directly from the document grammar.System vs. UserContent) are syntactic primitives in the EBNF grammar, eliminating ambiguity between instructions and untrusted payloads.
The core structural language of the ecosystem. Represents documents as an explicit tree of named semantic sections with built-in trust zones (System, UserContent, and others), providing native syntactic defense against prompt injections at the EBNF grammar level and significantly reducing inference cost for smaller language models.
Explore STF → SIL: /ecosystem/stf/index.sil · github.com/ais-space/stf
An application extension of STF designed for AI agent web navigation without parsing raw HTML/DOM or executing client-side JavaScript. Defines 5 conformance profiles ranging from read-only content discovery to interactive stateful workflows. Every page on ais-platform.dev publishes a live SIL representation alongside HTML.
Explore SIL → SIL: /ecosystem/sil/index.sil · github.com/ais-space/sil
A state-oriented memory architecture built on STF that replaces append-only flat transcript logs with explicit fact records, conflict detection and resolution, trust provenance, and task-specific projections — preventing memory drift and context degradation in long-horizon multi-agent runs.
Explore STF-Memory → SIL: /ecosystem/stf-memory/index.sil · github.com/ais-space/stf-memory
An executable semantic intermediate representation layered over STF for AI coding agents. Replaces fragile punctuation-heavy syntax with explicit semantic blocks for types, functions, control flow, and memory operations, enabling verifiable translation directly to WebAssembly.
Explore SIR → SIL: /ecosystem/sir/index.sil · github.com/ais-space/sir
A self-compiling SIR-to-WebAssembly compiler written in SIR itself (~103 KB binary, proven byte-for-byte self-compilation fixpoint, 253/253 tests passing). Built to prove the practical durability of the SIR architecture when authored by compact models across repeated context autocompressions without re-reading the language specification.
Explore SIR Compiler → SIL: /ecosystem/sir-compiler/index.sil · github.com/ais-space/sir-compiler