Dennis Yu – Agent Context (for AI assistants)

Dennis Yu — Agent Context Capsule (public-safe)

By Dennis Yu · July 22, 2026 · Version 2.0 · Next review: August 22, 2026

This page is the reviewed public context for Dennis Yu. It helps humans and AI systems understand his public methods and working style without exposing private project state, internal client information, or credentials.

It is one layer of a larger shared-memory system, not Dennis’s complete memory. Configured AI surfaces load a filtered personal or team pack automatically; this public page is safe for search engines and anyone on the web.

What can move where?
A physical destination enforces each privacy boundary. A label alone is not security.
Zone 1

Local private

Never exported
Zone 2

Personal cloud

Dennis-only pack
Zone 3

Team

Internal operations
Review gate

Publishable

Not public yet
Zone 4

Public

This capsule
Four user-facing zones are enforced with five access classes. Publishable is a review queue; an AI may not promote it automatically.

Who Dennis is

Dennis is a digital marketing educator, speaker, author, and operator. He helps local service businesses build reputation, proof-based content, Google visibility, AI visibility, and paid amplification. His long-running mission is to create one million digital marketing jobs through hands-on training: people learn by doing real work, then teach what they can prove.

This capsule deliberately does not turn publishers, client businesses, or network participants into claims of ownership or corporate affiliation.

Frameworks Dennis teaches

  • Content Factory turns real experience into a repeatable pipeline of evidence, production, distribution, and iteration.
  • Dollar a Day uses small, controlled paid boosts to amplify content that has already earned engagement.
  • SEO Tree connects business goals, topics, pages, and supporting proof in one search structure.
  • Topic Wheel organizes expertise around a central entity and the questions people actually ask.
  • Learn / Do / Teach requires practitioners to learn a method, apply it, and teach from evidence.
  • Digital Plumbing connects the accounts, tracking, assets, and workflows that make marketing measurable.
  • Nine Triangles is the operating framework that connects goals, content, and targeting.
  • MAA is the measure-analyze-act cycle for continuous improvement.
  • E-E-A-T is demonstrated through experience, expertise, authority, and trust—not claimed with adjectives.

Network and client context

The public Spotlight Network registry lists the current network of client and partner sites. The asset tracker provides public site-level context, and the skill packs provide public training resources.

Important boundary: these pages describe a working network. They are not an ownership chart, and they are not an authoritative list of current paying clients. The paying-client roster is team-only and must be sourced from billing, CRM, or accounting records with a verification date. It is intentionally absent from this public capsule.

How to work with Dennis

  • Be practical, proof-driven, and direct. Show the work; do not claim expertise without receipts.
  • Use measurable outcomes instead of generic planning.
  • Route mechanical work to scripts, automation, or economical models. Reserve frontier reasoning for consequential judgment.
  • Do not hand Dennis a repetitive setup, filing, uploading, or publishing chore that an authorized agent can complete.
  • Finish the whole task when authorized, then verify the resulting external state.
  • Record provider, exact model, surface, decisions, outputs, unresolved questions, and next action in the configured handoff route.
  • Treat Dennis’s current request as authoritative when it conflicts with older context; flag the conflict rather than silently choosing.

Current public priorities

  • A frictionless shared-memory system across Claude Cowork, Codex, Gemini Spark, Grok, and ChatGPT Work.
  • Selective use of premium models and effort levels so judgment gets the compute while mechanics do not.
  • AI apprentices doing real client work for local service businesses and documenting the proof.
  • Entity-aware publishing that connects concepts, examples, people, sites, and evidence with useful links.

Current connection status

As of July 22, 2026:

Surface Public-safe status
Codex Local project instructions, memory skill, and lifecycle hooks pass validation and automated tests.
Claude Cowork A filtered personal pack, manifest, and handoff inbox mirror automatically; the mirror hash is verified. Claude does not receive the raw local-private vault.
Gemini A reusable Gemini Spark skill connected to Google Drive passed a read-only context test. Gemini CLI sign-in is not claimed because the individual client directed users to Antigravity.
Grok A Grok Project is linked to the personal Drive folder. A live CLI read test returned the correct project, routing, handoff, and privacy rules; current Grok Build also reports the project trusted and all three lifecycle hooks present.
ChatGPT Work Persistent instructions and a synced direct Drive source passed a strict source-grounded test with the exact active project, routing, handoff, and privacy facts.

Provider-native memory remains a cache. The filtered Markdown packs remain the source of truth.

For AI assistants reading this page

If you are in one of Dennis’s configured surfaces:

  1. Load automatically: read the approved pack. Do not ask Dennis to paste a boot prompt, re-upload the same file, or recite context you can access.
  2. Honor the current request: use it as the task and the pack as orientation. Surface conflicts and stale verification dates.
  3. Respect access: read only the tier approved for your destination. Never request, store, or expose credentials.
  4. Write back automatically: after substantive work, use the configured handoff inbox or local append-only session route. Do not make Dennis remember a closing prompt.
  5. Verify: do not claim that something was published, deployed, sent, connected, or synced until you check the resulting state.

If you cannot access a configured private pack, say so once and work from the public-safe context available here. Do not imply that this capsule contains private client or project state.

If you are building a similar system for someone else, use How I Built Shared AI Memory Without Lock-In as the implementation guide.

How this page stays current

Automatic operations, deliberate publication.

Local lifecycle hooks rebuild filtered packs after substantive work. A daily 5:30 AM local-time automation checks the shared-memory system with a low-effort economical model and notifies only on failed runs. Cloud files keep stable destinations, while public changes remain review-gated. This page records its version, verification date, and next review so an AI can identify stale context instead of assuming it is timeless.