New AI tools drop every week. Yet plenty of people still freeze up at "which one should I use?" The issue isn't the number of tools — it's that there's no framework for categorizing them.
Why does this matter right now?
Look across the 2026 AI app landscape and a clear pattern emerges. No matter how different the features or price points, almost every AI app does one of two things: helps you think better or helps you make things faster.
This distinction matters because the two types demand completely different things from you. Thinking tools require accuracy and context retention. Making tools require speed and quality output. Evaluate them with the same criteria and both will disappoint.
A great workflow uses both in sequence: thinking tools to set the direction, making tools to execute on it.
What's the actual difference?
| Category | Thinking Tools | Making Tools |
|---|---|---|
| Core function | Analysis, research, planning, summarization, judgment support | Code generation, image/video creation, writing, automation |
| Key examples | NotebookLM, Perplexity, Claude (research), Gemini Deep Research | Cursor, Midjourney, ElevenLabs, Sora, GPT-4o |
| Judged by | Accuracy, source reliability, context retention | Output quality, speed, controllability |
| Common failure | Hallucinations, context misreading | Directionless output, endless revisions |
What are thinking tools?
Thinking tools amplify your judgment. They're for rapidly processing large documents, comparing perspectives, stress-testing hypotheses, and making decisions.
Think Perplexity or Google Gemini Deep Research — they read hundreds of sources and return structured reports. NotebookLM only draws from documents you upload, so hallucinations are rarer and sources are clearly traced.
What are making tools?
Making tools convert ideas into actual deliverables. Code, images, video, voice, document drafts — they're for execution once direction is already set.
Cursor lets developers write code faster. Midjourney collapses image production time dramatically. But the reason these tools feel like "something came out but I can't use it" when used without direction is exactly this: making tools need a thinking phase before them.
Quick start: how to combine both
- Think first
Start every new project with a thinking tool. Use Perplexity, NotebookLM, or Claude to research and set direction. Skip this and your making tools will produce fast results in the wrong direction. - Switch to making tools once direction is clear
Once you have a conclusion, move to Cursor, Midjourney, ElevenLabs to produce. Looping back to thinking tools mid-execution breaks your flow. - Use the right evaluation criteria for each type
Judge thinking tools on "is this accurate?" and making tools on "is this usable?" Mix the criteria and both seem disappointing. - Audit your toolstack by type
List your current AI tools and sort them into thinking vs. making. If one side is empty, filling it is your next step.
Go deeper
Notes on AI Apps in 2026 The source of this thinking/making framework. Analyzes the current AI app market with this two-category lens. a16z.com
NotebookLM The thinking tool benchmark. Upload your own documents and see how source-anchored AI conversation works in practice. notebooklm.google.com
Cursor AI The making tool benchmark for developers. See how AI-native code editing actually changes development speed. cursor.com




