AI Tools, and What Each One Is Actually Good At

OpenAI's assistant, the one most people start with.

Best for

Everyday writing, explanation and images in the same place.

Watch out

Will state something wrong with complete confidence. Check anything factual.

Anthropic's assistant. Strong at long documents and code.

Best for

Working through a long file or a messy document without losing the thread.

Watch out

Message limits on the free tier arrive quickly in a long session.

Google's assistant, wired into Search and Workspace.

Best for

Anything that needs current information, and work inside Docs and Sheets.

Watch out

Quality varies more than the others between task types.

A search engine that answers and cites its sources.

Best for

Research where you need to click through and check the source.

Watch out

Summarises well but will sometimes cite a page that does not say what it claims.

Autocomplete for code, inside your editor.

Best for

Boilerplate, tests, and the second half of a line you have already started.

Watch out

Suggests confidently in code it has not understood. Free for verified students.

A code editor built around an AI that can see your whole project.

Best for

Changes that touch several files at once.

Watch out

Easy to accept a large change you have not read. Review every diff.

An assistant that works in your terminal and edits files directly.

Best for

Long refactors and repository-wide tasks with tests to check them.

Watch out

Needs a clear instruction. Vague asks produce confident, wrong work.

One API key for models from every provider.

Best for

Building an app without locking yourself to one company's pricing.

Watch out

You are paying per token, so a loop with a bug gets expensive fast.

Runs open-weight models on your own machine with one command.

Best for

Learning what a model actually is, offline, at no cost.

Watch out

A laptop with 8 GB of RAM will run small models slowly.

The same idea as Ollama with a window and buttons.

Best for

Trying local models without touching a terminal.

Watch out

Less convenient once you want to call the model from your own code.

Where the open models, datasets and demos live.

Best for

Finding a model for a specific job — translation, speech, classification.

Watch out

Overwhelming without a clear idea of what you are looking for.

Image generation with the most distinctive look.

Best for

Illustration and concept art where style matters more than accuracy.

Watch out

No free tier, and hands and text still come out wrong often enough to matter.

Design tool with AI generation and editing built in.

Best for

Posters, social posts and anything where you need a template to start from.

Watch out

Everything made in it looks like it was made in it.

Cuts the background out of a photo.

Best for

Product photos for a shop listing, in about two seconds.

Watch out

Struggles with hair, glass and anything semi-transparent.

Speech to text, open source, runs on your own machine.

Best for

Transcribing recorded lectures or calls without sending audio anywhere.

Watch out

Punjabi and mixed Hindi-English speech come out noticeably worse than English.

Text to speech that sounds close to a person.

Best for

Voiceover for a product video or a course module.

Watch out

The free tier runs out quickly, and cloned voices raise consent questions you should take seriously.

Workflow automation you can host yourself.

Best for

Connecting a form, a database and a model without a monthly per-task bill.

Watch out

Self-hosting means you maintain it. Budget an afternoon to set up.

The same idea, hosted, with far more integrations.

Best for

Joining two services that already exist, in ten minutes, with no server.

Watch out

Priced per task, so a busy workflow gets expensive.

A place to test prompts against Gemini and get an API key.

Best for

Prototyping a prompt before you write any code around it.

Watch out

Tied to Google's models only.