After working through a catalogue of 360 AI and ML tools — each with a capability and pricing snapshot recorded at verification time, most recently on 2026-09-18 — the single biggest finding is this: the overwhelming majority are not new technology. They are packaging. The same handful of underlying capabilities gets resold under hundreds of different names, landing pages, and pricing models.
That matters if you're trying to pick a tool, because the thing you're actually buying is rarely the model. It's the interface, the workflow glue, and the billing structure wrapped around a capability you could often reach directly. This is what the catalogue taught me, and how to use it so you stop paying wrapper prices for commodity capability.
What the 360-tool catalogue actually shows
When you line the tools up side by side, they sort into a small number of functional clusters. Text generation and editing is the largest group by far. Then come image generation, transcription and speech, code assistance, and a long tail of vertical tools — legal drafting, medical note-taking, resume builders, real-estate listing writers — that take one general capability and dress it in industry vocabulary.
The vertical wrappers are where the confusion lives. A "legal AI assistant" and a "marketing copy assistant" often share the same underlying capability profile; what differs is the prompt scaffolding and the vocabulary on the pricing page. The catalogue snapshot makes this visible because it records capability, not branding. Two tools that read completely differently on their homepages can have near-identical capability entries.
The practical takeaway: when you evaluate a tool, ask what cluster it belongs to first. If it's a vertical wrapper around text generation, you're paying for the wrapper. Sometimes that's worth it — a medical note-taker that handles compliance formatting saves real time. Sometimes it isn't.
Why so many tools are the same tool
Model capability has converged enough that reselling is viable. A large share of the tools in the catalogue don't train anything. They call an underlying model through an API and add a layer on top: a nicer editor, a template library, an integration with your existing stack.
This isn't a scandal. Layers are useful. But it changes what you should compare. If two tools are both thin layers over similar underlying capability, then the model isn't the differentiator — the workflow is. The questions that actually separate them are boring ones:
- Does it fit the tool you already use daily, or does it demand a new tab and a new habit?
- Can you export your work, or is it locked in a format only that tool reads?
- What happens to your data — is it retained, used for training, or deleted?
- When the underlying model changes, does your output change without warning?
That last one gets ignored constantly. A wrapper that doesn't pin its model version can shift behavior overnight when the vendor updates. Your carefully tuned workflow breaks, and nobody tells you.
How to evaluate a tool in under ten minutes
Here's the process the catalogue work points toward. It's deliberately short, because most evaluation effort is wasted on the wrong questions.
Step one: identify the cluster. Is this a general capability or a vertical wrapper? If vertical, what general capability is underneath? Write it down in one sentence. If you can't, the tool's own marketing has already confused you, which is a signal.
Step two: find the closest general-purpose alternative. If the wrapper is text generation, the alternative is a general text tool. If it's transcription, the alternative is a general transcription service. Now you have a baseline to compare cost and effort against.
Step three: test with your own worst-case input. Not a clean example. The messiest real thing you'd actually feed it — the scanned PDF, the rambling voice memo, the document with three languages mixed in. Wrappers are tuned for demo inputs. Real inputs expose the seams.
Step four: check the exit. Export your test output. If you can't get your work out in a usable format, you're not evaluating a tool, you're evaluating a trap.
Ten minutes, four steps. The catalogue work suggests this catches most of the mismatch before you pay for anything.
A worked example: the "AI meeting notes" problem
Say you need meeting notes. Search turns up dozens of tools. Applying the process:
Cluster: transcription plus summarization. Two capabilities stacked, not one. That's already useful to know, because you can buy them separately or together.
Closest general alternative: a general transcription service plus a general text summarizer. Two tools, potentially cheaper, definitely more work to stitch together.
Worst-case input: a 45-minute call with three speakers, background noise, and one participant on a bad connection. This is where transcription accuracy and speaker separation actually get tested. A tool that handles a clean two-person podcast may fall apart here.
Exit check: can you get the transcript as plain text, and the summary as editable text? If the only output is a formatted note inside the tool's own app, you've found the constraint.
The decision usually comes down to whether the stitching cost of two general tools is higher or lower than the wrapper premium. For occasional use, the wrapper usually wins. For daily use at volume, the general tools often do — but only if you're willing to build the pipeline.
Where this approach breaks down
Honest limits, because the process isn't universal.
First, the catalogue snapshot records capability and pricing at a point in time. Pricing changes frequently — the vendor's own page is the only reliable source, and any number quoted anywhere else, including in a snapshot, can be stale within weeks. Treat the snapshot as a map of what exists, not a price list.
Second, the four-step process assumes you can articulate your own use case clearly. If you're exploring, not solving a specific problem, cluster identification doesn't help much. Exploration is legitimate; it just needs a different method.
Third, vertical wrappers sometimes carry real domain value that a general tool can't replicate — regulatory formatting, industry-specific templates, compliance checks. The process will flag these as wrappers and might push you toward a general tool that then fails on the domain requirement. When the domain matters, the wrapper premium is often justified.
Fourth, none of this tells you about reliability under load, support quality, or how a vendor handles outages. Those show up only in use, over time.
Why the count matters more than any single tool
The value of looking at all 360 wasn't finding a winner. It was seeing the shape of the market. When you know most tools are layers over convergent capability, you stop hunting for the one magic tool and start asking better questions: what does this layer add, what does it cost me in lock-in, and can I get the underlying capability more directly?
That reframing is the actual lesson. The catalogue didn't tell me which tool to use. It told me what I was actually choosing between.
Key Takeaways
- A catalogue of 360 AI tools shows most are packaging around a small set of underlying capabilities, not new technology.
- Tools sort into clusters: text, image, transcription, code, and vertical wrappers that resell general capability in industry language.
- When capability converges, the workflow — integrations, export, data handling — becomes the real differentiator, not the model.
- Evaluate in four steps: identify the cluster, find the general alternative, test worst-case input, verify you can export.
- Snapshots record what exists, not current prices; vendor pages are the only reliable source for pricing.
The next time a tool's landing page makes it sound like a category of one, run the four steps. Identify the cluster. Find the general alternative. Feed it your messiest input. Check the exit. Most of the time you'll find a wrapper — and then you get to decide, with clear eyes, whether the wrapper is worth its premium. Sometimes it is. Now you'll know why.
Sources
- AI Tool Database, Internally Verified Capability and Pricing Snapshot, 2026. A catalogue of 360 AI and ML tools with capability and pricing recorded at verification time, most recently 2026-09-18.
Frequently Asked Questions
If most AI tools are wrappers, is it ever worth paying for one?
Yes, often. Wrappers earn their premium when they handle domain-specific requirements a general tool can't — compliance formatting, industry templates, or a workflow that saves real time. The four-step process doesn't tell you to avoid wrappers; it tells you what you're paying for so you can judge whether the premium is justified for your use case.
How do I tell whether a tool is a wrapper or genuinely new technology?
Look at what it claims to do versus what it adds. If the core capability is text generation, transcription, or image creation, and the tool's differentiation is templates, integrations, or industry vocabulary, it's a wrapper. Genuine new capability is rarer than the market suggests. The catalogue snapshot helps because it records capability rather than branding.
Why does the underlying model version matter so much?
A wrapper that doesn't pin its model version can change behavior without warning when the vendor updates. Your tuned workflow breaks, output quality shifts, and nobody notifies you. It's one of the most overlooked evaluation questions, because it doesn't show up in a demo — only in production, after you've already built around the tool.