# 0.4.7 How today's tools realize it (/how-ai-works/critical-thinking/both-ai-vendors) --- type: Document title: "0.4.7 How today's tools realize it" description: "Features in Claude, ChatGPT and open-source tools that help you check." status: stable order: 15.7 ksor: owner: team:panaversity audience: [ public ] approval: by: process:panaversity at: 2026-11-11T08:18:16Z chapter: "0.4" part: "0" expert_status: null # as the Part 0 overview: Part 0 is optional on every route concepts: [ "0.4.7" ] last_verified: 2026-11-11 sources: - id: anthropic-model-settings resource: https://support.claude.com/en/articles/8664678-change-the-model-effort-and-thinking-settings title: "Change the model, effort, and thinking settings (Claude Help Center, verified 2026-11-11)" - id: anthropic-web-search resource: https://support.claude.com/en/articles/10684626-enable-and-use-web-search title: "Enable and use web search (Claude Help Center, verified 2026-11-11)" - id: anthropic-project-rag resource: https://support.claude.com/en/articles/11473015-retrieval-augmented-generation-rag-for-projects title: "Retrieval augmented generation (RAG) for projects (Claude Help Center, verified 2026-11-11)" - id: openai-deep-research resource: https://help.openai.com/en/articles/10500283-deep-research-in-chatgpt title: "Deep research in ChatGPT (OpenAI Help Center, verified 2026-11-11)" - id: openai-projects resource: https://help.openai.com/en/articles/10169521-projects-in-chatgpt title: "Projects in ChatGPT (OpenAI Help Center, verified 2026-11-11)" - id: openai-temporary resource: https://help.openai.com/en/articles/8914046-temporary-chat-in-chatgpt title: "Temporary chat in ChatGPT (OpenAI Help Center, verified 2026-11-11)" - id: openai-release-notes resource: https://help.openai.com/en/articles/6825453-chatgpt-release-notes title: "ChatGPT release notes, the entry \"More controls in temporary chat\" of 27 August 2026 (OpenAI Help Center, verified 2026-11-11)" - id: ollama-faq resource: https://docs.ollama.com/faq title: "FAQ (Ollama, accessed 2026-11-11)" generated: at: 2026-11-11T08:18:16Z by: process:panaversity trust_tier: unverified build_id: sha256:2c8363a5436bca5eb3cc81adfe9c5cdb55380e522b3ebd7618270caef15a8351 dirty: true ksor_version: 0.0.60 --- **In everyday life.** Three kitchens can cook the same recipe. The ovens are different, so you look for the controls before you start. **What you will learn.** Which features in today's tools make checking easier, and the one difference that changes how you work. **Why it matters.** The checkpoint is the same on every tool. But some tools show you their sources and their thinking, and some do not. The ideas in [0.4.1](why-judgment-is-the-job.md) to [0.4.6](delegate-not-surrender.md) are true on every tool. This lesson holds the product details, which change often. **Anthropic (Claude).** *As verified: November 11, 2026.* - **Visible thinking.** A Thinking section appears above the answer. You can open it to read a summary of how Claude reasoned.[^anthropic-model-settings] It is an explanation to inspect. It is not proof that the answer is correct. Thinking does not add a truth-checker ([Chapter 0.1](../what-ai-actually-is/the-machine.md)). - **Cited sources.** Answers from web search come with sources.[^anthropic-web-search] Open them and confirm that they say what the answer claims. - **Projects and files.** Give the model the real data, so you can check its claims against that data.[^anthropic-project-rag] - **New chats and model choice.** Start a new chat so the AI does not stay with its first answer. You can also change the model, from the menu beside the send button, and compare the answers.[^anthropic-model-settings] **OpenAI (ChatGPT).** *As verified: November 11, 2026.* - **Deep research with sources.** Reports come with sources.[^openai-deep-research] Open them. - **Projects and file uploads** let you base answers on your own data.[^openai-projects] - **Temporary Chat** is not saved in your history.[^openai-temporary] It has two settings. A **Personalized** temporary chat can still use your saved memories and custom instructions. An **Unpersonalized** one uses neither.[^openai-release-notes] When you want a fresh answer, choose Unpersonalized, and check the setting before you type. **Open source.** *As verified: November 11, 2026.* Ollama runs a model on your own computer, so you can ask the same question to a different model. Its makers say they do not see your prompts or data when you run a model locally.[^ollama-faq] **Security.** That is true for local models only. Ollama also offers cloud models, which send your prompts to its servers, so check which kind you are running. Download models only from the project's official site. Small local models make more mistakes. Use one as a second opinion, never as the final judge. **The difference that changes a decision.** Does the tool show sources and reasoning that you can inspect? If it does not, treat every claim as unsupported until you check it somewhere else. **The invariant.** This is the rule that stays true on every tool. A well-written answer is a draft. Checking it is your job. And a number that a decision depends on comes from code run on the original file, which you then check. [^anthropic-model-settings]: Claude Help Center, "Change the model, effort, and thinking settings." Accessed November 10, 2026. [^anthropic-web-search]: Claude Help Center, "Enable and use web search." Accessed November 9, 2026. [^anthropic-project-rag]: Claude Help Center, "Retrieval augmented generation (RAG) for projects." Accessed November 9, 2026. [^openai-deep-research]: OpenAI Help Center, "Deep research in ChatGPT." Accessed November 9, 2026. [^openai-projects]: OpenAI Help Center, "Projects in ChatGPT." Accessed November 9, 2026. [^openai-temporary]: OpenAI Help Center, "Temporary chat in ChatGPT." Accessed November 10, 2026. [^openai-release-notes]: OpenAI, ChatGPT release notes, "More controls in temporary chat," August 27, 2026. Read on OpenAI's help site on November 11, 2026. [^ollama-faq]: Ollama documentation, FAQ. Accessed November 10, 2026.