Nobody Cares About the Model Now… | The AI Corner Summary

1. Key Themes

Theme 1: The Model Is No Longer the Competitive Advantage

The window when frontier model access was scarce — and therefore defensible — has closed. Open models have closed the quality gap on most tasks, commoditizing what was once the prize.

"By the middle of this year, families like GLM, Qwen, and MiniMax had pulled close enough to the closed frontier that, for most production work, the choice of model no longer decides the result."

"The closed labs still win the hardest reasoning problems, and that lead is real. But winning the top 10% of problems is a thin reed to build a company on when an open model handles the other ninety at a tenth of the price."

Theme 2: Geopolitical Risk Is a Real Infrastructure Risk for AI Builders

Model access through a third-party API is a dependency with an external off-switch — a risk that went from theoretical to concrete on a single day in June 2025.

"On June 12, the United States ordered Anthropic to cut off foreign access to its two strongest models. The company pulled both offline worldwide to comply."

"A capability that can vanish on a Friday cannot sit on your critical path without a backup. Choosing that backup is a business decision long before it is a technical one."

Theme 3: The Five Real Moats in AI — Workflow, Context, Depth, Trust, and Routing

With the model commoditized, defensibility migrates to what surrounds it. The article identifies five durable moat families: Accumulation, Depth, Trust, Orchestration, and Cost Routing.

"What protects a company now lives everywhere except the model: the workflow it owns, and the context it compounds."

"The model answers the question. Everything that decides which question to ask, with what context, and what to do with the answer, is yours to own."

Theme 4: "We Have Data" Is Usually a Fake Moat

Data as a moat claim is the most common dressed-up weakness in AI pitch decks. The article dissects why data often fails all three moat tests.

"The trap is that data tends to behave backwards. The first records you gather are cheap and cover the common cases. The ones after grow more expensive to find and add less, because the easy ground is already taken and the long tail is rare by definition."

"A pile of data is not a moat. It is raw material that becomes one only when more of it visibly improves the product in a way customers can feel."

Theme 5: Enterprise Cost Pressure Is Accelerating Model Routing as a Category

Corporate AI spending discipline is no longer optional. Token billing made costs visible, and large enterprises moved fast to implement tiered routing strategies.

"Uber burned through its entire annual AI budget in roughly four months, then capped what any engineer could spend. Walmart, Cisco, Amazon, and Meta pulled in the reins the same quarter."

"Send each task to the cheapest model that clears the bar, and save the expensive ones for work that truly needs them. Teams that route this way report cutting their bills sharply, with nobody noticing a drop in quality."

2. Contrarian Perspectives

Contrarian 1: "AI Wrappers" Were Never the Death Sentence Everyone Said They Were

The conventional wisdom — that building on top of a model with no proprietary model of your own was a doomed strategy — turned out to be wrong. The fear was misplaced because it assumed the model was the prize.

"For two years, the cruelest word in AI was 'wrapper'... Serious founders walked away from good ideas over it. They were wrong, and 2026 made that impossible to deny."

The evidence: open weights caught up, prices fell, and model access proved to be revocable — making the wrapper critique irrelevant because the underlying premise (model = moat) was false.

Contrarian 2: Raw Infrastructure Scale Is Not a Permanent Advantage — Even for the Labs

The market has assumed that whoever spends the most on compute wins. Even OpenAI, the largest spender in the field, is quietly walking this back.

"Late last year, OpenAI was citing around $1.4 trillion in infrastructure commitments. By February, it was telling investors the real target sat closer to $600 billion, and it had started renting capacity from the same clouds it once planned to outbuild."

"When the largest spender in the field halves its own ambition under the cold eye of the public markets, the belief that raw scale is a permanent advantage looks a good deal weaker than it did a year ago."

Contrarian 3: Going Deep in a Vertical Tends to Manufacture Multiple Moats as a Side Effect

The conventional builder instinct is to stay horizontal and capture a larger TAM. The article argues vertical depth is a moat-compounding strategy, not a limiting one.

"Vertical depth is the same instinct aimed at one industry. Harvey beats a general model on legal work because it has absorbed a domain a generalist never sees, and going deep in one field tends to manufacture the other moats — data and trust and switching cost — as a side effect."

3. Companies Identified

Granola

Description: AI meeting note-taker that stores and surfaces context from past meetings
Why mentioned: Case study for the Accumulation Moat — the product's moat is the user's own memory, not the underlying model
Quote: "Granola, the note-taker that remembers every meeting, does not compete on model quality. It competes on the fact that leaving means walking away from your own memory."

Cursor

Description: AI coding assistant that learns from individual developer behavior
Why mentioned: Case study for the Accumulation Moat — every keystroke generates private training signal competitors cannot access
Quote: "Cursor turns ordinary use into private training signal. Every keystroke, and every accepted or rejected suggestion, is data no rival ever sees."

Sierra

Description: AI customer service platform for enterprise
Why mentioned: Case study for the Depth Moat — owns the full resolution workflow, not just a slice of it
Quote: "Sierra builds for that whole flow rather than a slice of it, which is why replacing it would mean rebuilding an operation, not switching a vendor."

Harvey

Description: AI platform built for legal professionals
Why mentioned: Case study for vertical Depth Moat — domain absorption creates compounding defensibility
Quote: "Harvey beats a general model on legal work because it has absorbed a domain a generalist never sees."

Glean

Description: Enterprise AI search and knowledge management platform
Why mentioned: Case study for the Trust Moat — wins in regulated enterprises through permissions, audit logs, and security controls
Quote: "Glean wins inside large regulated companies on permissions, audit logs, and the unglamorous controls that let a tool survive a security review at all. None of that can be faked or rushed."

Hugging Face

Description: Open-source AI model hub and community platform
Why mentioned: Case study for community-based Trust Moat — developer default behavior built over years, not purchasable with capital
Quote: "The reason developers reach for one place by reflex took years to earn and cannot be bought with a funding round."

Cognition (Devin)

Description: AI software engineering agent
Why mentioned: Case study for the Orchestration Moat — competes on reliability and guaranteed outcomes, not model quality
Quote: "Cognition, maker of the Devin agent, competes here on reliability rather than on the model, to the point of guaranteeing the engineering value a customer can hold it to."

OpenRouter

Description: Model routing infrastructure layer that sends tasks to the cheapest capable model
Why mentioned: Case study for the Cost Routing Moat — a business that earns its value precisely because models are interchangeable
Quote: "OpenRouter sends each task to the cheapest capable model and updates that judgment as the frontier moves. It is a business that earns its keep precisely because models have become interchangeable."

GitHub Copilot

Description: AI coding assistant embedded in GitHub's development environment
Why mentioned: Case study for Distribution Moat — wins by being the default inside existing developer surfaces, regardless of model quality
Quote: "The default destination was GitHub Copilot, not because the model underneath was better, but because it already lived in the editor and the billing those teams used."

OpenAI

Description: Frontier AI lab and API provider
Why mentioned: Two contexts: (1) its API ecosystem as an example of switching-cost moat; (2) its infrastructure spending reversal as evidence against scale-as-moat thesis
Quote (ecosystem): "Once tens of thousands of companies build on the OpenAI API, the moat is no longer the model. It is everyone else's switching cost."
Quote (scale): "Late last year, OpenAI was citing around $1.4 trillion in infrastructure commitments. By February, it was telling investors the real target sat closer to $600 billion."

MiniMax

Description: Chinese open-weight AI model developer
Why mentioned: Quantitative proof that open models now rival closed frontier models on real production benchmarks
Quote: "MiniMax's M3 reached roughly 59% on SWE-Bench Pro, competitive with closed systems on real coding work, at a sliver of the cost and under a license you can run on your own hardware."

Anthropic

Description: AI safety company and developer of the Claude model family
Why mentioned: The June 12 incident — U.S. government order to cut off foreign access — became the concrete proof of API dependency risk
Quote: "The United States ordered Anthropic to cut off foreign access to its two strongest models. The company pulled both offline worldwide to comply."

4. People Identified

Ruben Dominguez

Description: Author of The AI Corner newsletter
Why mentioned: Author of this article; frames the thesis that moats now live in workflow and context, not the model
Quote: "The model it runs on is rented. The moat is everything around it."

Sam Altman

Description: CEO of OpenAI
Why mentioned: Referenced in connection with OpenAI's ecosystem switching-cost moat and the infrastructure spending reversal; described as having an "honest moment" about OpenAI's position
Quote (contextual reference): "Once tens of thousands of companies build on the OpenAI API, the moat is no longer the model. It is everyone else's switching cost." (Altman referenced via image caption: "Sam Altman's honest moment.")

5. Operating Insights

Insight 1: Use a Three-Question Filter to Stress-Test Any Claimed Moat

Before investing in or building around a claimed moat, apply three tests. If it fails any one of them, it isn't a moat.

"1. Would it survive a model swap? 2. Does it compound, or merely accumulate? 3. Could a stranger rebuild it over a weekend? What survives is whatever took years to gather and cannot be conjured on demand."

Insight 2: Implement Tiered Model Routing Now — It's a Cost and Resilience Strategy

Task-to-model routing is both an immediate cost lever and a hedge against single-model dependency. The operational pattern is clear and already validated at scale.

"Send each task to the cheapest model that clears the bar, and save the expensive ones for work that truly needs them. Teams that route this way report cutting their bills sharply, with nobody noticing a drop in quality. Paying frontier rates for routine work now reads as carelessness rather than caution."

Insight 3: Design for Switching Cost from Day One — Don't Rely on Lock-In as an Afterthought

Accumulation moats require intentional architecture: the product must make departure painful by making user data, memory, or behavior signals inherently non-portable.

"Both clear the swap test easily. The memory and the signal belong to the product, never to the model running behind it."

6. Overlooked Insights

Overlooked Insight 1: Data Also Rots — Making Static Data Advantages Liabilities Over Time

The article briefly raises a point that receives little attention in the broader data-moat debate: a corpus that isn't continuously refreshed becomes a liability, not an asset. This has significant implications for any AI business that acquired a data advantage in an earlier period and hasn't invested in keeping it current.

"Data also rots. Patterns change, and a corpus you do not constantly refresh quietly turns into a liability."

Overlooked Insight 2: Vertical Depth Is a Moat Manufacturing Strategy, Not Just a GTM Choice

The article mentions almost in passing that going deep in a single industry doesn't just create one moat — it generates multiple compounding moats simultaneously. This reframes vertical AI as a strategic moat-stacking approach rather than simply a narrower market entry point.

"Going deep in one field tends to manufacture the other moats — data and trust and switching cost — as a side effect."