Peel Back the Landing Page: A Field Guide to Spotting Wrapper Startups

Let's be honest with each other for a minute.

You've seen the pitch. Sleek landing page, gradient hero section, a demo video with satisfying typing sounds, and a claim like 'AI-powered workflow transformation for enterprise teams.' You sign up for the beta. You poke around. And somewhere around minute four, you realize the entire product is one system prompt and a prayer.

Nothing wrong with shipping fast. But there's a difference between a company and a Chrome extension with a Series A, and early adopters deserve better pattern recognition. Here's the field guide we use internally.

The Red Flags

1. The prompt is the product. Ask a hard question about how the system handles an edge case. If the answer is 'the model figures it out,' you're not looking at engineering — you're looking at vibes with a subscription tier.

2. Zero evals, zero telemetry. We got an anonymous walkthrough of a well-funded YC-batch company recently (we'll spare them the embarrassment). When we asked how they measured output quality across model versions, the founder literally shrugged. If a model update can silently break your product and you wouldn't know for a week, you don't have a startup. You have a dependency.

3. The roadmap is a demo reel. Watch for roadmaps where every item is a feature a user could already get by typing into ChatGPT slightly differently. 'Coming soon: custom personas.' Cool. That's a text box.

4. No data flywheel. The wrapper's dirty secret is that it learns nothing. If user interactions don't feed back into something proprietary — a dataset, a fine-tune, a routing policy — the moat is the founder's charisma. That's a real moat, sometimes, but let's call it what it is.

5. Pricing that only makes sense as arbitrage. A classic wrapper tell: they charge you $30/month for something that costs them $2 in API calls, and their entire business model is that you haven't checked the token math. Arbitrage dies the moment the model providers ship the feature natively — and they always ship the feature natively.

6. 'Built on GPT-5' is the deepest technical disclosure. Real infrastructure teams can't stop talking about the boring stuff: latency budgets, fallback chains, context management. Wrappers can't, because there's nothing under the hood to talk about.

7. The founder demos beautifully and answers architecture questions badly. Trust the pattern. If someone can riff for ten minutes on vision but goes vague on 'how do you handle context windows over long documents,' the demo was the whole show.

The Real Signals

1. A proprietary eval pipeline. This is the big one. Teams that build their own evaluation harnesses — scoring outputs against domain-specific test sets, tracking regressions across model releases — are playing a different game. Evals are unsexy, expensive, and impossible to fake. A company with a real eval pipeline knows things about model behavior that its competitors don't, and that knowledge compounds.

2. A novel orchestration layer. Some problems genuinely require custom machinery: multi-agent routing, long-horizon memory, cost-aware model switching. If a team built a nontrivial orchestration layer to solve something the raw API can't do, that's engineering, not wrapping. Bonus points if the orchestration itself generates data no one else has.

3. Domain-specific feedback loops. The best AI companies we've looked at recently are embedded in a vertical — legal discovery, medical billing, logistics — where their users generate labeled, high-value interaction data that no general-purpose model can replicate. The model is a commodity. The feedback loop is the asset.

The Uncomfortable Truth

Here's where we lose half our readers: wrappers aren't automatically bad businesses. Some of the most profitable software companies in history were, functionally, thin layers over commodity infrastructure. Distribution and UX are real.

The problem is the framing. A wrapper dressed up as a moonshot wastes everyone's time — the founders', the investors', yours. A wrapper that's honest about being a fast, focused product on top of commodity intelligence? That can be a perfectly good little company.

So peel back the landing page. Ask about the evals. Ask what they learned that their competitors couldn't look up. The answers will tell you everything.

And if the answer is a shrug? You already know what you're looking at.