Frequently asked questions

Do I need to install anything on my computer?

It depends which platform you choose and how you plan to use it. Claude, ChatGPT, and Gemini all run in a browser with no installation needed. Claude Cowork or OpenAI's Codex are desktop applications you install on your Mac or Windows computer, and they are the recommended options on this site because they enable access to your files and a persistent memory system. The setup guide walks you through the installation for Claude Cowork step by step.

Is it free?

Most AI platforms have free tiers. They are sufficient for getting started, testing the use cases on this site, and doing occasional professional tasks. Each has a daily usage limit — when you hit it, you wait until the next day or upgrade to a paid plan (around $20/month for each). For regular professional use, a paid plan is worth it; for occasional use, the free tier is fine.

Which platform should I start with?

This is up to you. I personally like and use Claude Cowork over Codex (they have different strengths, and strikingly different company ethos if this matters to you). I have not yet used Gemini Spark extensively. If you are ready to install a desktop application, it gives you the most powerful and persistent working environment. Otherwise, you can opt for Claude in the browser if you want to try Claude first without installing anything. ChatGPT if your organisation already uses OpenAI tools, or if you want web search included for free. Gemini if your work is based in Google Workspace. The comparison page covers the differences in more detail.

What are the markdown files in the repository for?

They are personalisation files — plain text documents you place in your working folder that tell your AI who you are, how you work, and how you write. Once in place, your AI reads them at the start of every session, so it always has your professional context without you having to repeat yourself. The repository has downloadable templates for all four files, with guidance on what to put in each one.

Is it safe to use these tools for work?

Safe for non-confidential work — drafting, analysis, literature review, campaign prototyping, and so on. You should NOT paste organisational documents, personal patient data, identifiable information, or documents marked confidential into a consumer AI account. If your organisation needs AI tools for sensitive data, ask your IT or data protection officer about enterprise options, which offer stronger data isolation and compliance.

Will my conversations be used to train the AI?

By default, all three platforms may use conversations to improve their models. Each has an opt-out in its settings; ask the platform directly how to opt out. For any work-related content, opting out is good practice.

My organisation handles health data. Can I use these tools with it?

Not on standard consumer plans. Anthropic, OpenAI, and Google all offer enterprise accounts with proper data protection — no training on your content, data isolation, and compliance with GDPR and HIPAA where relevant. These require an organisational purchase. Speak to your data protection officer before using any AI tool with data that falls under your organisation's information governance policies.

The output doesn't sound like me. What do I do?

This usually means the AI hasn't been briefed on your voice and style. Set up your WRITING RULES.md and ABOUT ME.md files from the repository — these tell the AI exactly how you write and what to avoid. Once in place, outputs will need far less editing. You can also give direct feedback in the conversation: "This is too formal — make it more direct" or "Cut the third paragraph, it repeats what was just said."

The AI produced something that seems factually wrong. What do I do?

Never use AI outputs without checking them first. AI tools can produce confident-sounding statements that are incorrect — this is particularly common with specific statistics, citations, dates, and recent policy developments. Always verify key claims against primary sources before using AI-produced content in formal work or publications. If you ask the AI where a figure came from and it cannot give you a verifiable source, treat the figure as unverified.

How do I get better results from the AI?

Three things make the biggest difference. First, set up your personalisation files so the AI knows your context. Second, be specific in your requests — "write a one-page briefing note on vector-borne disease in southern Europe, using a systems thinking approach" will produce a far better result than "write about climate and health." Third, iterate — if the first draft is not right, say exactly what needs changing and ask again. Three or four rounds of feedback usually gets you to something usable.

Can I use AI tools in languages other than English?

Yes. All three platforms work well across many languages. You can write to the AI in your own language and it will respond in kind. You can also ask it to produce outputs in a specific language — useful for multilingual health information or reports that need to go to non-English-speaking audiences. The health websites use case covers multilingual content in more detail.

How do I add pages to this site, or share my own tools with others?

New pages can be added at any time — they are just HTML files placed in the same folder as the existing ones. To share a skill, template, or tool you have developed, use the submission form on the repository page. Reviewed submissions are added within two weeks and credited to you.

Who made this site?

This site was developed by Daniel Cauchi, a public health medicine doctor from Malta. It grew out of a personal process of learning how to use agentic AI tools as a non-coding professional, and adapting that knowledge for public health practice. Full credits — including the YouTube creators and academic frameworks that informed the site — are on the About page.

Is this site affiliated with Anthropic, OpenAI, or Google?

No. This is an independent site. The platforms mentioned are third-party products included for practical guidance. Pricing and features change — always check the provider's own website for current information.