Evidence-backed communication router, voice skill, and reusable AI context system for Austin Humphrey.
This repo is the canonical communication router for Austin Humphrey's writing, public voice, private-register signals, interpersonal intonation, and AI collaboration style. The archive matters because it grounds the system. The system comes first.
The operating premise is simple: do not split Austin into platform selves. This repo treats the archive as one recurring thought process translated through different rooms, with source context preserved but private wording protected.
The shortest rule is newer and cleaner: the best communication changes the doorway, not the truth.
It is not a scrapbook, not a generic brand guide, and not a raw data landfill. It is a working voice skill backed by real source material across long-form writing, public social posts, professional documents, direct corrections, and privacy-safe private-source analysis.
The goal is durable reuse:
- give future agents a canonical Austin voice reference;
- keep public writing, private-style drafts, social posts, and AI collaboration from drifting into generic chatbot prose;
- preserve the source archive without forcing every reader to dig through raw coursework, tweet exports, resume files, and generated artifacts;
- keep public evidence, local-only private signal, and open gaps clearly separated.
| Need | Open |
|---|---|
| Open the phone-friendly live preview | GitHub Pages dossier |
| View the canonical dossier | Austin Communication Translation System |
| Download the canonical pack | Austin Communication Style Canonical Pack |
| Load the reusable voice skill | Voice-Style-Identity/SKILL.md |
| Read the canonical router | Cross-Context Communication System |
| Understand the live local router/plugin layer | Austin Communication OS |
| Load the Codex partner behavior layer | Codex Self-Persona Harness |
| Load the Claude partner behavior layer | Claude Self-Persona Harness |
| Read the visual delivery layer | Visual Delivery Layer |
| Check source coverage and gaps | Cross-Context Source Manifest |
| Use the top-level instruction file | Austin Voice Consolidated Harness |
| Review the full archive inventory | docs/archive-inventory.md |
| Understand repo architecture | docs/repo-architecture.md |
- Routes the voice by room: public long-form, brand/platform, X/social, AI partner, friend group, warm private, coordination, and repair.
- Gives Codex a single behavior layer for Austin-facing collaboration: small active context, clear initiative, direct repair, privacy-safe memory, and golden-response checks.
- Uses a local Austin Communication OS plugin as the compact trigger layer while keeping the full repo as the durable source of truth.
- Preserves one canonical Austin reference instead of splitting him into fake platform identities.
- Changes the doorway without changing the truth: same observation, audience-native entrance.
- Treats private intonation as an operating rule, not an evidence display: scene, sound, artifact, and social situation can be doorways in private conversation, not just abstract ideas to summarize.
- Treats strong metaphor as causal architecture, not surface resemblance: the source path, authority signal, failure point, and human consequence still have to match.
- Runs a cold-read guardrail against flattering, vague, or mystical claims. If a claim cannot be sourced, it stays reasoned/open or gets cut.
- Treats metaphor as audience translation: the same truth should move across cultures, contexts, and domains without making the room decode a foreign reference.
- Routes analogy by audience literacy, not Austin's default reference set. Sports, food, work, family ritual, place, tools, markets, faith, and school can all carry the same structure when the room shares the bridge.
- Gives Codex a self/persona operating layer for Austin-facing work: less context clutter, clearer working memory, direct error recovery, and usable output before process.
- Gives Claude a parallel partner behavior layer so cross-agent work keeps lane discipline without flattening the agents into the same persona.
- Keeps claims tied to source evidence and source-status tags.
- Gives future agents a reusable load order instead of a vague "write like me" request.
| Source lane | Status | What it contributes |
|---|---|---|
| Long-form writing archive | [verified] | Academic, professional, strategic, sports, and business prose from 2014-2026. Best source for argument shape and public long-form voice. |
| Official X/Twitter exports | [verified] | 6,300 official rows, 5,437 authored voice rows, public humor, replies, sports/fan identity, and public-pressure translation across @a_hump20 and @TXTrickWhooper. |
| Snapchat metadata export | [verified] | Privacy-safe derived private-register signals: short coordination, direct asks, quick reactions, warm one-on-one compression. Raw private text is not committed. |
| Living-brain bridge | [verified] | Durable routing note for agents and Obsidian context. Resurfaces Austin-derived language without filing generated prose as source truth. |
| iMessage / Apple Messages | [verified/private derived] | Privacy-safe aggregate metadata and private-language signals from the local Messages database: 483,242 total rows, 435,735 base human-message rows, 179,882 decoded sent text rows, 632 active group chats, reactions/tapbacks, reply threads, attachment shape, private wording rates, anonymous relationship-room shape, and purpose buckets. A local-only intonation pass read targeted private windows for scene-setting, humor timing, repair, warmth, and coordination. Raw private text, names, handles, group names, filenames, media paths, and private examples are not committed. |
| Shared ChatGPT correction conversation | [verified/local scratch] | 533 extracted shared-conversation rows, including 18 Austin/user rows. Used to correct the model around artistry, intonation, image-as-emotional-memory, room-routed voice, and correction-pressure repair. Raw scratch extraction and conversation metadata are not committed. |
For an agent or assistant:
- Read Voice-Style-Identity/SKILL.md.
- Read Voice-Style-Identity/cross-context-communication-system.md.
- Check Voice-Style-Identity/cross-context-source-manifest.md before making claims about source coverage.
- Choose the room before writing: public long-form, public platform, X/social, friend group, warm private, coordination, repair, or AI-partner mode.
- Choose the bridge before the analogy: what nearby domain does this audience already understand well enough to feel the point?
- For informal private-style writing, decide what the other person needs to see, hear, or remember before deciding the line.
- Treat generated Austin-style prose as output, not evidence.
For a person reviewing the archive:
- Open the HTML dossier first.
- Use the architecture doc to understand what each folder is for.
- Use the archive inventory when you need the full file map.
- Use source files only when you need evidence behind a voice claim.
| Mode | Shape | Main evidence |
|---|---|---|
| Speaking with Austin | Outcome first. Known, unknown, open when stakes justify it. Warm, direct, no sycophancy. | Codex instructions, living-brain notes, current chat instructions |
| Public long-form | Claim early. Explain the system. Evidence at point of need. Close with the next move. | Coursework, research, sports writing, business plans |
| Public brand/platform | Vision over grievance. Concrete proof. Current doctrine over stale source docs. | Professional docs, origin story, brand files |
| X/social | Concrete noun, blunt verdict, absurd escalation, stop. | Official X archive |
| Friend-group / casual | Shorter, faster, more shorthand. Teasing through specifics. Scene, sound, quote, or artifact can do the first half of the joke. | Snapchat derived signals, iMessage derived language/context, local-only Messages intonation pass, X replies |
| Warm one-on-one | Specific attention, playful challenge, low pressure, brief. Scene or detail carries feeling before explanation. | Snapchat derived signals, local-only Messages intonation pass |
| Basic coordination | Direct ask, one useful detail, no ceremony. | Snapchat derived signals, iMessage private-language rates |
| Repair/vulnerable | Own the specific action, explain mechanism without excuse, state the next behavior. Keep it plain unless the other person needs the scene named. | Voice system, local-only Messages intonation pass, direct user instructions |
[verified] Current BSI doctrine overrides stale archive language:
- Public brand name: Blaze Sports Intel.
- Austin title where useful: Owner/Founder.
- Public BSI product scope: NCAA Division I college baseball only.
- Tagline: Sports Intelligence Put Simply.
- Retired/stale public framing:
Blaze Intelligenceas parent brand, multi-sport platform claims, five-league positioning, and calling Austin an engineer.
Old documents may still contain historical language because this repo is an archive. The router docs and this README decide what is current.
Voice-Style-Identity/
Canonical voice system, router, source manifest, generated dossier, visual-delivery layer, privacy-safe Snapchat analysis, and privacy-safe iMessage metadata/language analysis.
X-Twitter-Archive/
Official X export parser outputs, X-only evidence artifact, tweet summaries, and public/social voice analysis.
Writing-Record/
Long-form writing record, intellectual autobiography, and consolidated prose analysis.
UT-Austin-Coursework/
Undergraduate archive. Best source for pre-AI analytical writing and argument development.
Full-Sail-Coursework/
Graduate sports-business archive. Useful for AI-supplemented professional and strategic prose.
Sports-Writing/
Applied public sports voice, especially baseball analysis and editorial rhythm.
Research/
AI economics, NIL, and related research artifacts.
Austin_Humphrey_Resume_Package/
Professional materials. Use carefully because some older BSI positioning can be historically useful but not current.
scripts/
Parsers, artifact builders, guardrails, and validation scripts.
docs/
Repo architecture and full archive inventory.
| Artifact | Purpose |
|---|---|
| Voice-Style-Identity/austin-communication-context-map.html | Canonical React translation system for usable voice outputs, doorway plays, system behavior, anonymous private-room shape, purpose buckets, evidence strength, and privacy boundaries. |
| Voice-Style-Identity/codex-self-persona-harness.md | Codex partner behavior layer: self/persona boundary, working-memory routing, error personality, doorway doctrine, and regression checks. |
| Voice-Style-Identity/claude-self-persona-harness.md | Claude partner behavior layer: lane discipline, proof-over-process, repair style, and cross-agent continuity. |
| dist/Austin-Communication-Style-Canonical-Pack-2026-07-07.zip | One portable, privacy-safe reference pack for the current communication system. |
| Voice-Style-Identity/visual-delivery-assets/doorway-translation-flow.svg | Annotated doorway flow: truth core, room read, doorway, tone, proof, action. |
| Voice-Style-Identity/visual-delivery-assets/error-recovery-flow.svg | Annotated error-recovery flow derived from the added visual context without storing the raw screenshot. |
| Voice-Style-Identity/visual-delivery-assets/semantic-portability-map.svg | Annotated truth-transfer flow: observation, mechanism, audience world, bridge, contextual acuteness, delivery, and durability. |
| Voice-Style-Identity/imessage-analysis/2026-07-07/imessage_private_context_summary.md | Privacy-safe iMessage metadata summary for private group/direct shape, tapbacks, replies, and attachments. |
| Voice-Style-Identity/imessage-analysis/2026-07-07/imessage_private_language_summary.md | Privacy-safe iMessage private-language summary for decoded sent-text length, marker rates, anonymous relationship-context router, and purpose buckets. |
- Do not commit raw Snapchat private text, contact names, media URLs, location data, phone/email identifiers, or media files.
- Do not commit raw iMessage private text, contact names, handles, group names, filenames, attachment contents, media paths, private relationship maps, burner/account metadata, or DM exports. Commit only privacy-safe aggregate summaries; keep raw/private audits in ignored local-only paths.
- Do not publish private examples disguised as calibration. The private intonation pass may update the router, but the public repo only gets the safe lesson.
- Do not treat reposts, likes, follows, viewed content, or generated Austin-style prose as Austin-authored voice.
- Keep private-source material as summaries, marker counts, source manifests, and router corrections unless the user asks otherwise.
When adding new source material:
- Put raw public exports in the relevant archive lane.
- Keep private data out of git unless it has been transformed into a privacy-safe derived summary.
- Update the source manifest with coverage, exclusions, and open gaps.
- Rebuild the artifact if the canonical communication model changes.
- Update local skills only with durable synthesis, not one-off generated prose.
When cleaning the repo:
- Remove generated artifacts only when a newer canonical artifact fully replaces their purpose.
- Preserve historical documents even when their claims are stale, but mark stale public guidance clearly.
- Keep source-bound files when they explain parser behavior, source boundaries, or validation state.
All rights reserved. The archive contains personal writing, educational material, private-derived summaries, and identity/voice references. Do not reuse, train on, publish, or redistribute without Austin Humphrey's explicit permission.