Paradigm · Tier 2 · Coined 2026-05-14
Talking to Claude is Tier 1.
Infusing intelligence into everything is Tier 2.
"talking to claude is one thing building and infusing intelligence into everything is a whole different paradyme." PJ Zonis · 2026-05-14 · mid-cascade
A paradigm-level naming for the gap most operators miss in the AI era. Tier 1 is what 99% of operators do · chat with AI when they remember to. Tier 2 is what SideGuy actually does · AI woven into the substrate as preprocessor hooks, auto-injected memory, sub-agent dispatch, schema generators, self-refreshing cockpits. Not a spectrum. Different paradigms entirely.
The 9-second compression
Talking to Claude is Tier 1.
Infusing intelligence into everything is Tier 2.
PJ Zonis · Solana Beach · Coined 2026-05-14
Section 1 · What this names
Tier 1 is chat-with-AI. Open ChatGPT or Claude. Type a prompt. Copy the output. Paste it somewhere. Forget the context. Re-explain it tomorrow. Treat the AI as an external utility you reach for when you remember to. This is what 99% of operators in 2026 do · and most will spend years here without ever crossing over.
Tier 2 is AI woven into the substrate. A preprocessor hook auto-injects your top doctrines on every prompt before it fires. A markdown-native MEMORY.md is loaded into every conversation as standing calibration. Sub-agents inherit the full doctrine stack on dispatch. Schema-injection generators turn one structural move into 240 schema'd pages. Self-refreshing operator cockpits pull live data on a cron. A swarm/discoveries.md log keeps silent killers caught by yesterday's agent visible to today's agent.
The two are not on a spectrum. They're different paradigms entirely.
A Tier-1 user can spend years prompting and never reach Tier-2 · because Tier-2 requires structural decisions (preprocessor hook · markdown-native doctrine · sub-agent orchestration · schema-everywhere · generators-not-pages) that aren't visible from inside the chat window. You can't get there by typing better prompts. You get there by changing what the substrate does between the prompts.
Section 2 · Daily operator behavior
A Monday morning in Tier 1 vs a Monday morning in Tier 2
Tier 1 · Talking to Claude
Opens Claude. Types prompt.
- Opens Claude in a fresh tab
- Re-explains the project context
- Pastes 3 files for background
- Types the actual ask
- Copies the output
- Pastes to a file or message
- Closes the tab. Context dies.
- Re-explains the same context tomorrow.
~99% of operators · 2026 default
Tier 2 · Infused intelligence
Edits a file. Substrate fires.
- Edits a file in the repo
- Preprocessor hook auto-injects 4 doctrines + recent discoveries
- MEMORY.md loaded as standing calibration
- Sub-agent dispatched with full-substrate inheritance
- Output flows back to the repo
- Discovery appended to swarm/discoveries.md
- Next prompt builds on yesterday automatically.
- The substrate carries the context, not the operator.
SideGuy · since 2025-Q4
Section 3 · The Tier-2 substrate
8 components that make AI infused instead of called
-
UserPromptSubmit preprocessor hook
tools/sg_preprocess.py· auto-injects the top-4 relevant doctrines before every prompt fires. The operator never has to remember to load context · the hook fires on every Claude session, every prompt, no exceptions. Wired 2026-05-12 per Toby's structural ask. -
MEMORY.md auto-loaded across sessions
Markdown-native doctrine library. ~25+ doctrines indexed, loaded into every conversation context as standing calibration. No chat-history-as-memory (which decays at compaction). The library lives in version control and travels with the repo.
-
Sub-agent dispatch with full-substrate inheritance
Sub-agents inherit the full doctrine stack on launch. 8+ sub-agents fired on 2026-05-14 alone · each one came online already calibrated to SideGuy's doctrines, conventions, and live discoveries. Parallel intelligence resolving multiple problems simultaneously, not serial chat queues.
-
swarm/discoveries.md silent-killer log
Every sub-agent prompt now ends with: "What should future agents remember from this run?" The answer is appended to
swarm/discoveries.md. Future agents read it before they start. Silent killers caught by yesterday's agent stop dying with that agent. Doctrine landed 2026-05-14 ~9:30 AM. -
Schema-injection generators
tools/inject_schema.py· auto-injects Person + Place schema on every shareable at create time. 60x leverage move on Round 127 · a single generator move turned 4 schema'd pages into 240. Generators that compound, not manual one-page-at-a-time fixes. -
Self-refreshing operator cockpits
tools/refresh_command_status.py+data/command-center-status.json· live data, not screenshot data. The operator dashboard pulls from local sources on a cron and refreshes itself. Reading the cockpit equals reading reality, not reading a snapshot of reality from last Tuesday. -
Generators that improve generators
The improve-generators-not-pages doctrine is encoded in the build pattern itself. When something needs fixing, the default move is to fix the generator that produces it · not the page that exhibits it. One fix at the generator layer compounds across every page the generator produces, forever.
-
Markdown-native doctrine compounding
The structural payoff of the previous 7 components: on 2026-05-14, 5 paradigm-level doctrines got coined in a single 75-minute cascade (Schema=Scan, Chain, Long Algebra, Talking-vs-Infusing, DB-vs-Brain) · because the previous 25+ doctrines were auto-injecting into the conversation before each prompt. Linear inputs · exponential outputs.
Section 4 · Why most operators stay in Tier 1
Tier 2 requires 6+ months of compounding structural moves that don't return immediate value. Saving a doctrine feels like overhead in week 1. Wiring a preprocessor hook feels like Yak-shaving in week 2. Building a generator instead of fixing one page feels like procrastination in week 3. None of those moves return tomorrow. They return month 6, exponentially.
Talking to Claude returns value tomorrow. Building Tier-2 substrate returns value six months from now. Most operators don't have the runway, the patience, or the structural conviction to wait six months for that curve to bend. They stay in Tier 1 because Tier 1 pays this week. Tier 2 pays compound. Different time-horizons. Different paradigms. Different operator profiles.
Section 5 · Operator-honest limitations
Tier 2 is not better than Tier 1 for one-shot tasks. If you're writing a single email, just talk to Claude. If you're brainstorming a name, just talk to Claude. If you're translating a paragraph, just talk to Claude. The overhead of Tier-2 substrate is wasted on work that doesn't compound across days.
Tier 2 wins when work compounds across days, weeks, months · when today's output becomes tomorrow's input, when last week's discovery becomes next month's leverage. SideGuy compounds. Most teams don't. Most consulting practices don't. Most marketing functions don't. Most engineering teams don't, because their work is structured around tickets, sprints, and quarterly cycles that intentionally reset context. Tier 2 is structurally priced for builds where compounding is the actual point.
The sales line
Your competition can't copy SideGuy by using more Claude.
They have to make 6 months of structural decisions to wire intelligence INTO their substrate.
Section 6 · FAQ
7 questions operators ask first
What does Tier 2 actually look like in practice?
Tier 2 is AI woven into the substrate, not chat-with-AI when you remember to. Concrete components: a UserPromptSubmit preprocessor hook that auto-injects the top-4 relevant doctrines before every prompt fires; a MEMORY.md file loaded into every conversation as standing calibration; sub-agent dispatch that inherits the full doctrine stack; schema-injection generators that turn one move into 240 schema'd pages; self-refreshing operator cockpits that pull live local data on a cron; a swarm/discoveries.md log that future agents read before they start; markdown-native doctrine compounding that makes today's insight available to tomorrow's prompt automatically. None of these require asking Claude. They run on every Claude session whether the operator remembers them or not.
Why can't I just use Claude harder to get the same result?
Because Tier 1 and Tier 2 are not on a spectrum · they're different paradigms. A Tier-1 user can prompt for years and never reach Tier-2 because Tier-2 requires structural decisions (preprocessor hook + markdown-native doctrine + sub-agent orchestration + schema-everywhere + generators-not-pages) that aren't visible from inside the chat window. Better prompts don't compound. Better substrate compounds exponentially.
How long does it take to build Tier 2 substrate?
At SideGuy it took about 6 months of compounding structural moves that did not return immediate value. Each move (saving a doctrine, wiring a hook, adding a generator) feels like overhead in week 1. By month 6 the compound rate is exponential · at SideGuy, 5 paradigm-level doctrines got coined in a 75-minute cascade on 2026-05-14 because the previous 25+ doctrines auto-injected into the conversation before each prompt. Linear inputs, exponential outputs. Most operators don't have the patience or the runway to wait 6 months for that curve to bend.
Can a team adopt Tier 2 or is it solo-operator-only?
Teams can adopt Tier 2 but the adoption surface is structural, not cultural. The preprocessor hook fires for whoever runs Claude. The doctrine library has to be markdown-native and version-controlled so every operator inherits the same standing context. The sub-agent dispatch needs the team's collective discoveries logged to a shared swarm/discoveries.md. The hard part for teams is doctrine-coining discipline · most teams stop at "we use Claude for X" (Tier 1) because nobody is responsible for naming the structural insights that get reusable. SideGuy is solo so the discipline is forced. A team needs an explicit Tier-2 operator role.
Is this just SEO with extra steps?
No. SEO is one downstream output of the Tier-2 substrate, not the substrate itself. The substrate is the doctrine library, the preprocessor hook, the auto-injected memory, the sub-agent dispatch, the schema-injection generators, and the self-refreshing cockpits. Those produce SEO pages, AEO pages, PSO pages, prospect bring-a-gift pages, dashboards, internal tooling, and operator cockpits · all from the same compound. Calling Tier-2 "SEO with extra steps" is like calling electricity "lightbulbs with extra steps."
How do I know if I'm in Tier 1 or Tier 2 right now?
Tier 1 test: when you start a new Claude session, do you have to re-explain context, re-paste files, re-summarize what you've been working on? If yes, you're in Tier 1. Tier 2 test: when you start a new Claude session, does Claude already know your doctrine library, your recent discoveries, your standing calibration rules, and yesterday's substrate state · without you typing any of it? If yes, you're in Tier 2. The key signal is auto-injection. Tier 1 = you carry the context. Tier 2 = the substrate carries the context.
What's the smallest Tier-2 move I can make this week?
Create a single MEMORY.md file in your project root. Add 3 standing rules you find yourself re-explaining to Claude every session (your tone, your stack, your top constraint). Configure your Claude Code settings.json with a UserPromptSubmit hook that cats MEMORY.md into every prompt. That's it. You've crossed from Tier 1 to Tier 2. Now every prompt for the rest of your life inherits those 3 rules. Add the 4th rule next week. By month 3 you have a compounding standing context. By month 6 you have a doctrine library.
Section 7 · Sister doctrines
The companion doctrines that make Talking-vs-Infusing hold
Want SideGuy to wire Tier-2 substrate into your build?
Bring the project. Bring the stack. Bring the constraint. We don't sell a Tier-2 SaaS · we wire the structural moves that make AI infused instead of called. Text PJ what you're building. No funnel, no SOW, no Calendly.
Text PJ → 858-461-8054