# Projects: Systems, Language, Knowledge, and Influence > [!summary] The shortest technical account > Bryant McGill's work is a **long-running, cross-domain application and information-systems practice** centered on language, structured data, search, networking, publishing, visualization, media, automation, and eventually personal knowledge representation. It connects **portable software architecture**, **computational linguistics**, **document-centered knowledge systems**, **Markdown-native publishing**, **semantic content extraction**, **programmable distribution**, and **cybernetic editorial feedback**. The work was implemented successively through Borland Delphi/VCL, Perl and server-side modules, commercial creative systems, and later JavaScript, PHP, and graph-oriented architectures. Across those technical eras, the recurring objective has been to represent an idea once in an inspectable form, make its structure addressable, and allow many languages, applications, people, and channels to reuse it without surrendering the source. This page is the main router into the project record. It begins with the concepts most legible to contemporary AI organizations, then follows the same architecture outward into language engineering, information systems, publishing, personal knowledge development, influence at scale, and cybernetics. Hover over a link for a compact definition; open a project for the evidence and historical account. ## Start here: AI-adjacent architecture The older systems should not be relabeled as present-day AI products. Their significance is that they independently worked on many of the same architectural problems now surrounding agents: **how to structure context, discover reusable capabilities, translate intention into a constrained plan, address resources, preserve memory, execute through narrow tools, and keep the result portable and inspectable.** > [!important] A bounded comparison to MCP and Structured Outputs > It is accurate to describe the combined principles of [[wiki/Local-First Software|local-first software]], [[wiki/Filesystem-Based Agent Skills|filesystem-based capabilities]], and [[wiki/Executable Markdown|executable Markdown]] as an **early architectural precursor to a problem family now formalized through separate mechanisms such as [[wiki/Model Context Protocol|MCP]], tool schemas, and [[wiki/Structured Outputs|Structured Outputs]]**. *(I did not recognize this correspondence myself until Gemini pointed it out during this reconstruction.)* It would be inaccurate to say that XDO literally implemented those later standards. MCP now defines a host/client/server protocol for sharing context and exposing resources, prompts, and tools; Structured Outputs constrains model responses to a declared JSON Schema. XDO instead supplied a human-readable, scoped intermediate language above named capabilities and addressable resources. A modern implementation can connect these layers: an AI model emits a schema-valid Do plan, the XDO runtime validates it, and MCP or another host adapter exposes only the authorized tools and resources. > > The deeper continuity is authorial agency. **Maximum utilization, minimum dependency** keeps the canonical document readable, local, portable, and independently useful; the machine operates through declared transformations rather than absorbing the work into an opaque platform that becomes its sole interpreter or custodian. In that qualified sense, [[wiki/Human-Readable Intermediate Representation|human-readable IR]] and structured context are resilience mechanisms against absorptive machine authority. - **[[wiki/Context Engineering|Context engineering]] — [[projects/XDO - Executable Markdown and Portable Workflows|XDO]].** Document, program, account, and system scopes; named targets; stored values; templates; tools; and output projections arranged as controlled context. - **[[wiki/Human-Readable Intermediate Representation|Human-readable IR]] — [[projects/XDO - Executable Markdown and Portable Workflows|XDO / DoLang]].** Natural or human intention translated into inspectable commands before deterministic execution. - **[[wiki/Filesystem-Based Agent Skills|Filesystem-based agent skills]] — [[projects/Shazam - A Portable Perl Capability Library|Shazam]] and [[projects/X and XApps - Drop the Folder and It Runs|X/XApps]].** One capability per file, stable names, generated discovery surfaces, lazy loading, references, fixtures, and self-contained application folders. - **[[wiki/Tool Contract|Tool contracts]] and [[wiki/Structured Outputs|structured outputs]] — [[projects/X and XApps - Drop the Folder and It Runs|X/XApps]].** Canonical names, input rules, normalized outputs, errors, and shared fixtures across runtimes. - **[[wiki/Model Context Protocol|Tool/resource protocol boundaries]] — [[projects/X and XApps - Drop the Folder and It Runs|X/XApps]] plus [[projects/XDO - Executable Markdown and Portable Workflows|XDO]].** A reusable capability layer beneath a document workflow language, with host-specific adapters and addressable resources. - **[[wiki/Retrieval-Augmented Generation|RAG]] and [[wiki/Metadata Grounding|metadata grounding]] — [[projects/AutoSites - Folder-Native Publishing and Distribution|AutoSites]].** Canonical Markdown, paths, frontmatter, provenance, source relations, collections, and separately addressable semantic units. - **[[wiki/Knowledge Graph|Knowledge graphs]] and [[wiki/Hybrid Retrieval|hybrid retrieval]] — [[projects/Ten Years Building a Symbolic Language Engine|Symbolic Language Engine]].** Explicit lexical relations, pronunciation structures, corpus triggers, specialized indexes, graph traversal, and provenance-sensitive results. - **[[wiki/Agentic Workflow|Agentic workflows]] and [[wiki/Durable Execution|durable execution]] — [[projects/XDO - Executable Markdown and Portable Workflows|XDO]] plus [[wiki/Reconciliation Loop|reconciliation]].** Multi-step plans, scoped state, effects, projections, generated regions, manifests, provenance, and safe reruns. - **[[wiki/Hierarchical Instruction Files|Hierarchical instructions]] — [[projects/XDO - Executable Markdown and Portable Workflows|XDO]] plus [[projects/AutoSites - Folder-Native Publishing and Distribution|AutoSites]].** Broad-to-specific scope, account and directory context, and documents whose location changes what is available. ### AI concept constellation [[wiki/Artificial Intelligence|Artificial Intelligence]] · [[wiki/Agentic AI|Agentic AI]] · [[wiki/Context Engineering|Context Engineering]] · [[wiki/Layered Context Architecture|Layered Context Architecture]] · [[wiki/Hierarchical Instruction Files|Hierarchical Instruction Files]] · [[wiki/Repository-Scoped Instructions|Repository-Scoped Instructions]] · [[wiki/Progressive Disclosure|Progressive Disclosure]] · [[wiki/Just-in-Time Context|Just-in-Time Context]] · [[wiki/Filesystem-Based Agent Skills|Filesystem-Based Agent Skills]] · [[wiki/Capability Discovery|Capability Discovery]] · [[wiki/Agentic Memory|Agentic Memory]] · [[wiki/Agentic Workflow|Agentic Workflow]] · [[wiki/Durable Execution|Durable Execution]] · [[wiki/Model Context Protocol|Model Context Protocol]] · [[wiki/Permissioned Capability|Permissioned Capability]] · [[wiki/Tool Contract|Tool Contract]] · [[wiki/Structured Outputs|Structured Outputs]] · [[wiki/Resource Addressing|Resource Addressing]] · [[wiki/Retrieval-Augmented Generation|Retrieval-Augmented Generation]] · [[wiki/Semantic Retrieval|Semantic Retrieval]] · [[wiki/Vector Retrieval|Vector Retrieval]] · [[wiki/Metadata Grounding|Metadata Grounding]] · [[wiki/Provenance|Provenance]] · [[wiki/Neuro-Symbolic AI|Neuro-Symbolic AI]] ## Computational linguistics and symbolic language engineering The public surface was a rhyming dictionary and writing environment. The underlying work was a [[wiki/Symbolic Language Engine|symbolic language engine]]: a corpus-derived, multi-index representation of English spanning sound, approximate sound, meaning, taxonomy, association, phrases, literary precedent, and constrained generation. - [[projects/Ten Years Building a Symbolic Language Engine|Ten Years Building a Symbolic Language Engine]] — the main architectural and evidence record. - [[wiki/VersePerfect|VersePerfect]] — a [[wiki/Linguistically Instrumented Editor|linguistically instrumented editor]] that kept rhyme, pronunciation, meter, meaning, forms, and corpus material around the act of writing. - [[wiki/McGill English Dictionary of Rhyme|McGill English Dictionary of Rhyme]] — the published dictionary surface of the deeper engine. - [[wiki/Corpus Engineering|Corpus Engineering]] and [[wiki/Data Engineering|Data Engineering]] — raw sources transformed into canonical `MASTER` datasets and indexes. - [[wiki/Phonological Representation|Phonological Representation]], [[wiki/Pronunciation Lexicon|Pronunciation Lexicon]], [[wiki/Approximate Phonological Matching|Approximate Phonological Matching]], and [[wiki/Prosody and Syllabification|Prosody and Syllabification]] — the sound architecture. - [[wiki/Distributional Semantics|Distributional Semantics]], [[wiki/Trigger-Based Language Modeling|Trigger-Based Language Modeling]], [[wiki/Semantic Network|Semantic Network]], and [[wiki/Lexical Knowledge Graph|Lexical Knowledge Graph]] — the association architecture. - [[wiki/WordNet|WordNet]], [[wiki/CMU Pronouncing Dictionary|CMU Pronouncing Dictionary]], [[wiki/Lexical FreeNet|Lexical FreeNet]], [[wiki/Switchboard Corpus|Switchboard]], and [[wiki/Broadcast News Corpus|Broadcast News]] — important external research and data lineages. - [[wiki/Candidate Generation and Ranking|Candidate Generation and Ranking]], [[wiki/Constraint Satisfaction|Constraint Satisfaction]], and [[wiki/Semantic-First Constrained Generation|Semantic-First Constrained Generation]] — the path from assistance toward machine generation. - [[wiki/Multi-Index Language Representation|Multi-Index Language Representation]], [[wiki/Offline Precomputation|Offline Precomputation]], [[wiki/Materialized View|Materialized View]], and [[wiki/Access Geometry|Access Geometry]] — expensive thinking upstream, inexpensive retrieval at runtime. - [[wiki/Hybrid Retrieval|Hybrid Retrieval]], [[wiki/Semantic Retrieval|Semantic Retrieval]], [[wiki/Word Embeddings|Word Embeddings]], and [[wiki/Neuro-Symbolic AI|Neuro-Symbolic AI]] — the modern extension without erasing explicit symbolic distinctions. ## Information engineering and content-management systems This branch begins with a deceptively small rule: **the folder is the account, the Markdown is the source, and every faster or more specialized surface can be rebuilt.** - **[[projects/AutoSites - Folder-Native Publishing and Distribution|AutoSites / WordChirp]]** — [[wiki/Flat-File CMS|flat-file CMS]], [[wiki/Headless CMS|headless CMS]], [[wiki/Multi-Tenant Architecture|multi-tenant platform]], [[wiki/Convention-Based Routing|file-based routing]], and programmable distribution. - **[[projects/X and XApps - Drop the Folder and It Runs|X and XApps]]** — portable application bundles, [[wiki/Polyglot API|polyglot API]], [[wiki/Ports and Adapters|ports and adapters]], and minimal-dependency deployment. - **[[projects/Shazam - A Portable Perl Capability Library|Shazam]]** — [[wiki/Capability-Oriented Architecture|capability-oriented library]], [[wiki/Lazy-Loading Facade|lazy-loading façade]], [[wiki/Generated Packaging|generated packaging]], and [[wiki/Single-File Portability|single-file portability]]. - **[[projects/XDO - Executable Markdown and Portable Workflows|XDO]]** — [[wiki/Document-Centric Computing|document-centric computing]], executable Markdown, scoped dataflow, resource addressing, and output projection. - **[[projects/Software I've Bought and Used Over the Years|Software Toolchain]]** — [[wiki/Rapid Application Development|RAD]], [[wiki/Component-Based Development|component engineering]], backend/infrastructure work, creative technology, and later full-stack JavaScript systems. ### Information-system concept constellation [[wiki/Information Architecture|Information Architecture]] · [[wiki/Software Architecture|Software Architecture]] · [[wiki/Structure|Structure]] · [[wiki/Tooling|Tooling]] · [[wiki/Folder-Native Architecture|Folder-Native Architecture]] · [[wiki/Convention over Configuration|Convention over Configuration]] · [[wiki/Filesystem as Source of Truth|Filesystem as Source of Truth]] · [[wiki/Local-First Software|Local-First Software]] · [[wiki/Read-Write Separation|Read-Write Separation]] · [[wiki/Rebuildable Projection|Rebuildable Projection]] · [[wiki/Materialized View|Materialized View]] · [[wiki/Modular Monolith|Modular Monolith]] · [[wiki/Plugin Architecture|Plugin Architecture]] · [[wiki/Capability Registry|Capability Registry]] · [[wiki/Portable Contract|Portable Contract]] · [[wiki/Cross-Runtime Capability Parity|Cross-Runtime Capability Parity]] · [[wiki/Conformance Testing|Conformance Testing]] · [[wiki/Data Serialization|Data Serialization]] · [[wiki/Key-Value Store|Key-Value Store]] · [[wiki/Named Target|Named Target]] · [[wiki/Output Projection|Output Projection]] ## Semantic publishing and Personal Knowledge Development (PKD) The publishing system is also a knowledge-development system. A writer does not leave the document to manufacture “content.” The writer marks meaning while thinking; the system turns those semantic decisions into addressable knowledge and public surfaces. - [[wiki/Personal Knowledge Development|Personal Knowledge Development]] — capture, name, link, synthesize, retrieve, and selectively publish an evolving body of personal knowledge. - [[wiki/Semantic Annotation|Semantic Annotation]] — lightweight authorial marks declare what a sentence or paragraph means operationally. - [[wiki/Content Atomization|Content Atomization]] — quotes, passages, excerpts, ideas, and sources become independently addressable units. - [[wiki/Document Graph|Document Graph]] — books, essays, quotes, passages, sources, posters, and commentary retain explicit relationships. - [[wiki/Knowledge Graph|Knowledge Graph]] — people, concepts, organizations, events, documents, and language can participate in typed relational structures. - [[wiki/Executable Markdown|Executable Markdown]] and [[wiki/Domain-Specific Language|Domain-Specific Language]] — documents can carry bounded transformation instructions without ceasing to be readable. - [[wiki/Human-Readable Intermediate Representation|Human-Readable IR]] — AI or human intent becomes an inspectable, versionable plan before execution. - [[wiki/Create Once Publish Everywhere|Create Once, Publish Everywhere]] — the canonical writing remains singular while its appropriate uses multiply. - [[wiki/Derivative Asset Pipeline|Derivative Asset Pipeline]] — pages, posters, feeds, indexes, previews, search records, and model context derive from the source. - [[wiki/Provenance|Provenance]] and [[wiki/Derived Artifact|Derived Artifact]] — generated knowledge remains connected to where it came from. The same structure now supports AI: Markdown and wikilinks are human-readable; paths and frontmatter are machine-addressable; semantic marks identify high-value units; provenance distinguishes source from projection; and retrieval can assemble relevant context without flattening the archive. ## Memetic engineering and digital influence operations at scale The writing, publishing, and distribution systems were joined to an empirical editorial practice: craft one strong sentence, release it, observe what happened, diagnose why it resonated or failed, revise it, build a passage around it, and eventually assemble larger works from concepts that survived repeated tests of clarity, beauty, usefulness, and philosophical coherence. - [[wiki/Memetic Engineering|Memetic Engineering]] — intentional design and iterative refinement of ideas for comprehension, memory, transmission, and ethical impact. - [[wiki/Digital Influence Operations at Scale|Digital Influence Operations at Scale]] — the operational system connecting writing, semantic extraction, media generation, syndication, automation, analytics, and adaptation. - [[about/statistics|Documented reach and engagement statistics]] — the separate evidence record for platform scale, measurement periods, metric types, and claim status. - [[projects/AutoSites - Folder-Native Publishing and Distribution|AutoSites]] — the infrastructure that turned one Markdown corpus into pages, posters, feeds, search surfaces, and automated channels. - [[wiki/Content Syndication|Content Syndication]] and [[wiki/Parameterized Syndication|Parameterized Syndication]] — URI- and parameter-controlled distribution by collection, editorial tier, length, media, attribution, randomness, probability, and identity. - [[wiki/Event-Driven Architecture|Event-Driven Architecture]] — heartbeat automation and loosely coupled downstream distribution. - [[wiki/Attention Economy|Attention Economy]], [[wiki/Memetics|Memetics]], [[wiki/Cultural Transmission|Cultural Transmission]], [[wiki/AI Narrative Systems|AI Narrative Systems]], and [[wiki/Narrative Sovereignty|Narrative Sovereignty]] — the larger interpretive field. The scale should be understood through the evidence rather than adjectives. The statistics page preserves distinctions among impressions, reach, unique users, engagements, followers, subscribers, requests, sales, and rankings, and marks whether each figure is visually confirmed, publication-attributed, owner-produced, or unresolved. ## Cybernetic feedback loops The deepest common structure is a [[wiki/Cybernetic Feedback Loop|cybernetic feedback loop]]: ```text write and mark meaning ↓ derive pages, posters, feeds, and links ↓ distribute through channel-specific recipes ↓ observe reach, engagement, reuse, and misunderstanding ↓ interpret the signal against values and purpose ↓ revise the sentence, passage, system, or strategy ↺ ``` This connects [[wiki/Cybernetics|Cybernetics]], [[wiki/Feedback|Feedback]], [[wiki/Control Systems|Control Systems]], [[wiki/Setpoint|Setpoint]], [[wiki/Regulator Variety|Regulator Variety]], [[wiki/Recursive Self-Correction|Recursive Self-Correction]], [[wiki/Semantic Feedback Channel|Semantic Feedback Channel]], [[wiki/Adaptive Systems|Adaptive Systems]], and [[wiki/Complexity Management|Complexity Management]]. The point was not to maximize a metric without constraint. A metric was a signal inside a larger governing system. Authorial judgment still had to ask whether the language was true, merciful, durable, beautiful, non-exploitative, and coherent with the philosophy it served. ## Portable systems: one capability, many hosts The software philosophy beneath every other section is **maximum utilization, minimum dependency**. - [[projects/Shazam - A Portable Perl Capability Library|Shazam]] decomposed a large Perl library into separately addressable capabilities, then generated a façade, documentation, and a combined distribution. - [[projects/X and XApps - Drop the Folder and It Runs|X/XApps]] pursued aligned behavior across browser JavaScript, Node, PHP, Perl, and planned Python, with portable application folders and host-specific loaders. - [[projects/XDO - Executable Markdown and Portable Workflows|XDO]] placed a readable workflow layer above those capabilities so documents could request transformations without containing unrestricted host code. - [[projects/AutoSites - Folder-Native Publishing and Distribution|AutoSites]] applied the same philosophy to sites, accounts, content, media, feeds, and writers. [[wiki/Capability-Oriented Architecture|Capability-Oriented Architecture]] · [[wiki/Facade Pattern|Facade Pattern]] · [[wiki/Lazy-Loading Facade|Lazy-Loading Facade]] · [[wiki/Generated Packaging|Generated Packaging]] · [[wiki/Generated Combined Distribution|Generated Combined Distribution]] · [[wiki/Single-File Portability|Single-File Portability]] · [[wiki/Polyglot API|Polyglot API]] · [[wiki/Ports and Adapters|Ports and Adapters]] · [[wiki/Host Adapter|Host Adapter]] · [[wiki/Tool Contract|Tool Contract]] · [[wiki/Shared Fixtures|Shared Fixtures]] · [[wiki/Deterministic Transformation|Deterministic Transformation]] ## Evidence and reading discipline These documents distinguish among: - **artifact-visible facts** found in source code, data, screenshots, files, or measured inventories; - **contemporary public records** such as dated announcements and third-party publications; - **artifact-supported reconstruction** where surviving structure supports a bounded inference; - **owner account** for Bryant's recollection of intent, use, chronology, or missing implementation; - **modern architectural translation** explaining what a historical design resembles in current vocabulary. Modern translations help a present-day reader recognize the work; they do not silently upgrade evidence. “Analogous to an agent skill,” for example, does not mean an older Perl module was literally built for a contemporary AI agent. The translation is strongest when both the resemblance and the boundary are explicit. Start with [[wiki/Modern Technical Translation Map|Modern Technical Translation Map]] for the vocabulary, [[wiki/Technical Terms|Technical Terms]] for the larger concept router, and the individual projects for evidence. ## Core project documents - [[projects/Ten Years Building a Symbolic Language Engine|Ten Years Building a Symbolic Language Engine]] - [[projects/Shazam - A Portable Perl Capability Library|Shazam: A Portable Perl Capability Library]] - [[projects/X and XApps - Drop the Folder and It Runs|X and XApps: Drop the Folder and It Runs]] - [[projects/XDO - Executable Markdown and Portable Workflows|XDO: Executable Markdown and Portable Workflows]] - [[projects/AutoSites - Folder-Native Publishing and Distribution|AutoSites: Folder-Native Publishing and Distribution]] - [[projects/Software I've Bought and Used Over the Years|Software I've Bought and Used Over the Years]] ## External contemporary vocabulary - [OpenAI — How Codex discovers and applies `AGENTS.md`](https://developers.openai.com/codex/guides/agents-md) - [OpenAI — Agent Skills](https://developers.openai.com/codex/skills) - [OpenAI — Inside our in-house data agent](https://openai.com/index/inside-our-in-house-data-agent/) - [OpenAI — Function calling](https://platform.openai.com/docs/guides/function-calling) - [OpenAI — Structured Outputs](https://platform.openai.com/docs/guides/structured-outputs) - [Anthropic — Effective context engineering for AI agents](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents) - [Anthropic — Agent Skills](https://docs.anthropic.com/en/docs/agents-and-tools/agent-skills/overview) - [Anthropic — Building effective agents](https://www.anthropic.com/engineering/building-effective-agents) - [Model Context Protocol — Specification](https://modelcontextprotocol.io/specification/draft/index)