Shareable
Twenty posters of the framework.
The stack, the laws, the archetypes, the dynamics, every poster downloads as a watermarked PNG or PDF. Use them in decks, on LinkedIn, on your wall. Citation-ready, no signup.
Back to the frameworkI, The Stack
Three ways to see all ten layers at once.
The Framework · One Image
At-a-Glance Summary
The single image to answer 'what is this framework?', 10 layers with a one-liner each, plus the 4 laws strip. Built for screenshot + paste.
The Framework · At a Glance
Supply Chain of Intelligence™
The 10 layers of the generative AI stack, not logistics, not freight.
The Four Structural Laws
Intelligence commoditizes downward.
Value accrues at bottlenecks.
Surface captures attention; chain captures power.
Generation and verification must be separate.
10 layers · 50 sublayers · 4 laws. The map for every AI strategy conversation.
SupplyChainOfAI.com
Magazine Cover
The Hero Brand Poster
Dark, editorial, scroll-stopping. The single image to anchor a LinkedIn post or open a deck.
A Framework for AI Defensibility
The Supply Chain
of Intelligence.
10 layers · 50 sublayers · 4 structural laws
below the line
above the line, where intelligence compounds
By Anand Arivukkarasu · supplychainofai.com
Reference Sheet
The 10 × 50 Grid
Every layer and every sublayer on one page. The pin-on-the-wall version. ★ marks structurally defensible sublayers.
Supply Chain of Intelligence™ (SCoI)
The 10 layers · 50 sublayers of the generative AI stack.
SCoAI · SHEET 01/01
Not logistics. The AI stack.
★ = structurally defensible sublayer. Below L1 = foundation. Above = where intelligence compounds.
Anand Arivukkarasu · SupplyChainOfAI.com
One Square
The Stack, on a Square
Sized for LinkedIn, X, decks, Pinterest. Ten chips, the tagline, the attribution, nothing else.
Supply Chain of Intelligence™ (SCoI)
The 10 layers of the generative AI stack.
Not logistics. The generative AI stack.
Anand Arivukkarasu
Mental Model
Above / Below the Line
The single most teachable diagram in the framework. Below the line: inputs consumed. Above: intelligence compounded.
The Mental Model
Above the line, intelligence compounds.
Below the line, the inputs get consumed.
▲ Above, value compounds
data · trust · distribution · workflow · context · surface · memory
▼ Below, inputs the chain consumes
power · water · fabs · chips · data centers
Below the line is consumed. Above the line is accumulated. Build above. Hedge below.
Anand Arivukkarasu · SupplyChainOfAI.com
II, The Arguments
The posters that win the disagreement.
Why It Matters
10 Benefits of the Framework
The answer to 'why should I care?' One artifact, ten precise reasons. Best for product leaders, founders, and skeptical investors.
Why The Framework Exists
10 things you cannot say
without Supply Chain of Intelligence™.
A Common Language
Replaces vague words, 'agentic', 'wrapper', 'AI-native', with layer-level precision.
Predicts Who Gets Absorbed
Law I: surface-only products become features inside the model layer below them.
Names the Bottleneck
Law II: durable value sits at the scarce layer, data, trust, distribution, memory.
Decodes 'Agent'
An agent is not a layer. It is L5 + L7 (+L8) packaging. Decode it; never quote it.
Separates Surface From Chain
Beautiful UIs capture attention. Deep chains capture power. The framework draws the line.
Shows Where Intelligence Compounds
L1 → L8 is gravity-fed downward, value-fed upward. Memory (L8) is the strongest moat.
Maps Trust to a Layer
L3 is not a feature, it is a structural gate that the model layer cannot legally cross.
Forces an Honest Moat Conversation
Founders, investors, PMs ask: 'which layers do we actually own?' No more hand-waving.
Replaces Hype With Diagnosis
Every announcement maps to a layer. Every threat maps to a law. Every category maps to an archetype.
Shifts From Generation To Accumulation
Most AI thinking optimizes for output. The framework optimizes for what persists.
Vague words let weak strategy hide. The framework forces it into the light.
Anand Arivukkarasu · SupplyChainOfAI.com
The Physics
The Four Structural Laws
Wrappers get compressed. Bottlenecks compound. Surface captures attention; chain captures power. Memory is the final moat.
The Physics of the Stack
The Four Structural Laws.
Intelligence Commoditizes Downward
Wrappers don't survive. Wrappers become features.
L7-only → absorbed by L2
Value Accrues at Bottlenecks
Find the scarce layer. Own it. Everything else is rent.
L1b · L3 · L8
Surface Captures Attention; Chain Captures Power
Beautiful UIs get users. Deep chains keep them.
L4 + L5 + L6 + L8
Memory Is the Final Moat
What the system remembers about the user, no one else can rebuild.
L8 compounds
Laws describe what the market will do, not what you wish it would do.
Anand Arivukkarasu · SupplyChainOfAI.com
The Moat Shape
The Defensibility Triangle
L1b × L3 × L8. Own two corners and you're defensible. Own three and you're uncopyable. Everything else is rent.
Where Moats Live
The Defensibility Triangle.
Own two corners = defensible. Own three = uncopyable.
The model layer can replicate the surface. It cannot replicate your data, your trust, or your memory.
SupplyChainOfAI.com
Buzzword, Decoded
The Agent Decoder
'Agent' is not a layer. It's L5 + L7 (+L8) packaging. Use this whenever someone announces an agent, and ask which three layers.
Buzzword, Decoded
"Agent" is not a layer.
Without L5
It's a chatbot.
Without L7
It's a script.
Without L8
It's a demo.
When someone says "we built an agent," ask which three layers.
SupplyChainOfAI.com
III, The Models
The frameworks-within-the-framework.
Pattern Recognition
The 6 Archetypes
Every AI company collapses into one of six shapes: Fortress · Refinery · Railroad · Memory · Surface · Agent. Find yours.
Pattern Recognition
The 6 Archetypes of AI companies.
Every AI company collapses into one of six shapes. Find yours.
Fortress
Multi-layer ownership. Built to outlast the model layer.
e.g. Bloomberg · Epic · Tempus
Refinery
Owns the data well. Sells refined intelligence upward.
e.g. Apollo · Scale · Getty
Railroad
Owns the rails. Every workload pays the toll.
e.g. NVIDIA · AWS · Stripe
Memory
Compounds context per user. Switching cost becomes painful.
e.g. Notion AI · Granola · Linear
Surface
Beautiful, exposed, structurally absorbable by L2.
e.g. Jasper · Copy.ai · Gamma
Agent
Packaged as 'an agent.' Decode it: L5 + L7 (+L8) only works with L8.
e.g. Sierra · Lindy · Cognition
Archetype is not a vibe. It is the layer-set you own.
Anand Arivukkarasu · SupplyChainOfAI.com
Three Dimensions
The Intelligence Cube™
Layer × Sublayer × Depth. Defensibility is volume, not feature count.
The Intelligence Cube™
Defensibility has three dimensions.
Layer × Sublayer × Depth. Volume = total defensibility.
SupplyChainOfAI.com
For Non-Technical Readers
The Gold Mining Analogy
The whole stack told as one extended metaphor, from the land itself to the polished jewelry. The poster to hand a board member.
For Non-Technical Readers
Supply Chain of Intelligence,
told as a gold mine.
The Record Book, Compounding Knowledge
The jeweler keeps records: which designs sold, which metals each customer prefers.
Wearing the Jewelry, The Moment of Experience
People see the ring on the finger, the surface, the sparkle, the emotional moment.
The Jewelry Store & Workshop
A single ring is useful.
The Master Jeweler
A jeweler takes refined gold and crafts rings, necklaces, watches, each requiring specialized skill.
The Railroads & Transport
Refined gold needs to move, by rail, armored truck, secure vault.
The Hallmark & Assay Office
Before gold enters the market, the assay office verifies purity and the hallmark guarantees quality.
The Smelter & Refinery
Raw ore becomes pure gold through smelting.
The Raw Gold Ore
The unrefined material pulled from the earth.
The Shovels & Mining Equipment
Before anyone finds gold, someone has to build the pickaxes, drill rigs, and mine shafts.
The Ground Itself, Land, Power, Materials
Before the gold rush, you need land, water rights, ore deposits, and the miners who work the seams.
Every gold rush enriches the shovel sellers, the assayers, and the refiners, long after the miners are gone.
Anand Arivukkarasu · SupplyChainOfAI.com
Self-Diagnostic
Where Does Your Company Live?
A fill-in-the-blank stack chart. Print it, tick the layers you actually own, and answer three honest questions.
Print · Pin · Answer
Where does your company actually live?
Mark the layers you own. Not what you touch. Not what you integrate. What you own.
Retention, learning, compounding context. What the system remembers.
Interface, presentation, experience. How the user meets the intelligence.
Workflow, routing, coordination. How skills compose into outcomes.
Applied skills and capabilities. Doing the actual work.
Connectivity, permissions, integrations, the pipes layer.
Trust, acceptance, approval. Can the system be allowed in?
Intelligence refinement. Rent early, build custom at scale.
The raw input. What data do you have that nobody else can get?
The shovels. Chips, data centers, networking, cloud, edge, what is needed to process intelligence.
What supports the chain. Energy, water, fabs, materials, skilled trades, the inputs the entire stack consumes.
Q1
Which layers do you own?
Tick boxes →
Q2
Which is your bottleneck?
The scarce one, Law II.
Q3
What happens if L2 ships it free?
If nothing, you're inside Law I.
If the boxes you tick are only L7, the model layer below you owns your roadmap.
Anand Arivukkarasu · SupplyChainOfAI.com
IV, The Dynamics
Where the framework becomes an operating system.
Where The Meter Runs
The AI Tax Map
Seven taxes every AI product pays, token, context, orchestration, inference margin, API dependency, model lock-in, distribution. Most founders only price for two.
Where The Meter Runs
The AI Tax Map.
Every AI product pays seven taxes. Most founders only price for two.
Inference Margin Collapse
Tax #01Every call = compute cost. Margin shrinks with usage, not scale.
The Token Tax
Tax #02$/token × every user × forever. The meter never stops running.
Model Lock-In Tax
Tax #03Switch providers = rewrite prompts, re-eval, re-validate. Real cost.
API Dependency Tax
Tax #04Rate limits, deprecations, ToS shifts. Your roadmap = their decisions.
Context Tax
Tax #05Bigger windows = bigger bills. Context cost compounds per turn.
Orchestration Tax
Tax #06Each tool-call = a model hop. Latency × cost × failure surface.
Distribution Tax
Tax #07App stores, ad platforms, search, they price you. You don't price them.
If you can't name your tax stack, your margin isn't real, it's just unbilled.
Anand Arivukkarasu · SupplyChainOfAI.com
Law I, Visualized
The Stack Compression Map
When the model layer ships your feature for free, who holds and where the counter-move sits. The visual proof of Law I.
Law I, Visualized
The Stack Compression Map.
When the model layer ships your feature for free, who holds, and where the counter-move sits.
◌ Compressed
Surface-only★ Survives
Bottleneck ownersThe model layer absorbs anything that doesn't sit underneath it.
Anand Arivukkarasu · SupplyChainOfAI.com
Four Strategic Arcs
The Migration Paths
Companies that survive Law I migrate. Four archetypal arcs, Wrapper to Workflow, Surface to Memory, Tool to System, Data to Platform, cover almost every move worth making.
Four Strategic Arcs
The Migration Paths.
Companies that survive Law I migrate. Four arcs cover almost every move worth making.
Wrapper → Workflow
Stop renting features from L2. Embed inside the work.
Surface → Memory
Trade attention for accumulation. Make leaving painful.
Tool → System of Record
Become the place the workflow runs through, not on.
Data → Platform
Open the API. Let others build the surface on your bottleneck.
Standing still on L7 is the only move that always loses.
Anand Arivukkarasu · SupplyChainOfAI.com
How The Stack Moves
The Gravity Flow
Data flows down. Value flows up. Memory compounds recursively. The single rule that explains where margins go.
How The Stack Moves
The Gravity Flow.
flows down, training, fine-tuning, context fill.
flows up, margin accrues at orchestration & surface.
compounds, every interaction makes leaving more painful.
Build where things accumulate. Not where they pass through.
SupplyChainOfAI.com
The Moat Quadrant
Open vs Closed Surfaces
Plot a surface by openness, gate density, and memory depth. The top-left is where AI cannot legally, technically, or commercially compete.
The Moat Quadrant
Open vs Closed Surfaces.
Plot a surface by openness, gate density, and memory depth. The top-left is where AI cannot easily reach.
Read The Map
- ● X, how open the data/API is
- ● Y, how many trust gates protect it
- ● Size, depth of accumulated memory
The top-left big bubbles are where AI cannot legally, technically, or commercially compete.
Open is cheap. Closed is durable. Gated + closed + memory-rich is uncopyable.
Anand Arivukkarasu · SupplyChainOfAI.com
What Google Actually Said
GEO is Just SEO, The Five Myths
Google's Nov 2025 generative-AI search guide, decoded through the 10-layer stack. The AEO/GEO playbook being sold to founders is mostly noise, here's what actually moves the needle.
What Google Actually Said
GEO is just SEO.
The five myths, debunked.
Google's Nov 2025 generative-AI search guide, decoded through the 10-layer stack. Most of the "AEO/GEO playbook" being sold to founders is noise.
The Myth Being Sold
What Google Says
What Actually Wins · By Layer
Write llms.txt files for AI
“Not a Google signal. No special treatment.”
Ship unique, expert content at L1b.
Chunk your content for AI parsing
“No ideal length. Systems read context.”
Write for humans. L7 is downstream of L1.
Rewrite copy in 'AI-friendly' phrasing
“Models understand synonyms and intent.”
First-hand POV beats keyword fitting.
Farm inauthentic mentions across the web
“Spam systems catch it. RAG ignores it.”
Earn citations through L3 trust gates.
Pile on structured-data schema everywhere
“Not required for generative search.”
Technical hygiene + indexability. That's it.
The Bottom Line
Google's AI search is grounded in core ranking. The moat is still L1b unique data × L3 trust gates × L8 memory. Everything else is rented surface.
Source: Google Search Central · "Optimizing your website for generative AI features" (2025).
Anand Arivukkarasu · SupplyChainOfAI.com
V, For Your Audience
Role-specific posters to share alongside your own POV.
For Product Leaders
The Product Leader's Lens
Five questions to ask of any AI feature on your roadmap, mapped to the layer each one stress-tests. The poster that ends 'just a wrapper' arguments in a standup.
For Product Leaders
Five questions to ask of any AI feature on your roadmap.
Before the demo. Before the spec. Before the OKR. If your team can't answer four of five, you're shipping a wrapper, not a product.
If the model layer ships this for free next quarter, what's left?
If the answer is 'nothing', you're building inside Law I.
Where does the data come from that no competitor can get?
No proprietary L1 → no defensible learning loop.
What does the system remember after the session ends?
Stateless features churn. Memory compounds retention.
Who verifies the output when it's wrong?
Generator and verifier must be separate. That's Law IV.
Is the surface the moat, or just the doorway?
Beautiful UI gets users. Deep chain keeps them.
"A roadmap is a layer claim. Most AI roadmaps are claiming L7."
Supply Chain of Intelligence™, the 10 layers of the generative AI stack.
SupplyChainOfAI.com
For Investors
The Investor's Lens, Rent or Own
A 10-layer diligence map. Which layers compound, which commoditize, and the five questions every AI deal memo should answer before the check clears.
For Investors
Rent or own? A 10-layer diligence map.
Every AI cap-table claim collapses to one question, which layers does this company structurally own, and which is it merely renting from the layer below?
Five questions for the next deal memo
- 1.Which two layers does this company actually own?
- 2.What does the L2 roadmap make free in 18 months?
- 3.Where does proprietary data come from, structurally?
- 4.What compounds the longer customers stay?
- 5.If a 'gate' exists above it, who owns the gate?
Supply Chain of Intelligence™, the 10 layers of the generative AI stack.
SupplyChainOfAI.com
More posters as the framework grows. Suggest one →
Back to the framework