I’ve spent a lot of time using and reading about AI tools this year and I’d like to share that knowledge with others. I’ve compiled a list of categories and tools within them, so you can choose the right tool for your situation and how get the most out of the tool. I’ve also added any limitations on functionality or free usage that you should know.

Coding assistants

These tools streamline the coding process by offering code generation, auto-completion, and chat. All the tools in this section bring genAI to your IDE, so you don’t have to switch to your browser and back.

Use these tools as assistants to automate tasks and augment your abilities, but remember to question and judge the quality of results instead of blindly accepting everything.

Note that while GitHub Copilot does not have a free tier (just a trial), the others (IDX, Cursor and Codeium) have free tiers that do not expire.

GitHub Copilot

The famous AI coding assistant from a few years ago which is still one of the most famous ones.

Uses:

  • Extension for VS Code with code-generation functionality.
  • Available on the GitHub website as an AI assistant such as in Pull Requests.

Unfortunately, this is only available as a 30-day free trial and a paid subscription.

See the Billing page or Subscriptions plans page.

Cursor

image

Uses:

  • Local IDE with code-generation functionality.
  • Provides code generation in the terminal.

An AI-focused IDE for creating and updating code, tab auto-completion, and for chats with persistent history. This runs on VS Code internally and looks a little different, so switching is easy. The advantage is that Cursor is free, unlike VS Code paired with GitHub Copilot.

Here is an example of using code editing to apply changes inline:

image

Pricing:

  • Free tier - a limited number of queries per month for “Premium models” and “gpt-4o-mini or cursor-small”. Then it falls back to unlimited requests on the GPT-3.5 model. Go to your profile page on the Cursor website to see your usage so far.
  • Paid tier - subscribe to a paid plan for higher usage limits. Or put in your OpenAI key if you already have a ChatGPT subscription. It also supports keys for other providers.

See the Pricing page for prices and specific usage allowances.

See my Cursor cheatsheet for usage tips.

IDX

An online IDE by Google, released in 2024. It is based on VS Code and runs on Google Cloud.

image

  • Online IDE.
    • Create and run projects for free. Syncs with GitHub repos. This is a limit on how many active projects though and they get archived with inactivity.
    • AI chat, AI tab auto-complete, and AI code suggestions.
    • Includes Bash terminal for running code.
    • Supports free use of the Gemini API for genAI. Note if you use your credentials locally instead of in their environment, you get an error.

Functionality via the VS Code command prompt or keyboard shortcuts:

image

Chat with Gemini:

image

Code suggestions:

image

They have sample templates which are great for getting started - with Node, Python, Go, etc.

image

See also the extension in the sidebar for more options.

Codeium

image

Uses:

  • IDE extension with code-generation functionality - Codeium extension in the VS Code extensions marketplace.
  • Browser-based chat - Live chat
  • Code review assistant - Forge

Pricing:

  • Free tier - unlimited usage for the basic features, which covers prompts, chat, and auto-complete.
  • Paid tier - more advanced models like Llama 405B and GPT-4, guaranteed zero-data retention, and some other advanced features.

See the Pricing page.

Links:

Amazon Q

image

Uses:

  • Chatbot
  • Extension for VS Code with code-generation functionality.
  • Deploy AI-generated solutions.

I haven’t use this one yet but it was recommended by a colleague. Amazon lists 5 products on their Amazon Q page, including:

  • Amazon Q Business - Answer questions, provide summaries, generate content, etc. It can also generate and deploy apps based on the requirements you give it in natural language.
  • Amazon Q Developer - chat, debug, add tests, upgrade.

See their Getting Started page for how to use their options.

Pricing:

  • Free tier - this is limited
  • Paid tier - full features.

See Pricing page.

Webpage development

Tools to make webpages and preview them, without coding.

Note that these tools are powerful but if you do not specify exactly what you want, you may get unexpected results. And regardless of what you specify, there’s a chance the results will not look good and they are sometimes not even interactive or visible.

v0.dev

image

Uses:

  • AI-powered code editor in the browser.
  • Webpage preview.

You can enter a prompt and get results for webpages with previews, then tweak the results including targeting specific elements. It generates results with React and Tailwind which you can download.

It is convenient and powerful (including making animations and games). Thought it is notas advanced as an IDE in terms of control and functionality and you start with an empty starting point rather than your codebase.

The v0.dev tool is also limited in that is only gives a single page.

You can achieve similar results if you use an IDE extension as a sidepanel or another window for your browser so you preview your app in realtime. Though, you’ll have to interact with your code (possibly across several files) to get your code assistant to generate changes, while v0.dev allows you to right-click on an element on a page to edit it or add to it.

Websim

Uses:

  • Website creation.
  • Image creation.

This is one of my favorite and most-used AI tools and it is entirely free.

Websim. is a no-code AI-powered website builder that allows you to generate webpages and AI-generated images based on natural language prompts. It offers a unique “hallucinated” internet experience, where pages are dynamically generated as you navigate, creating an initial page with links that don’t lead anywhere until you visit them.

I wrote a full post on this here: Websim

Chatbots

Tools that you can have a dialogue with in the browser or access via API. To understand the token limits and the costs of API queries for each, see my Token Translator website.

ChatGPT

Uses:

  • Chatbot
  • Image creation (paid tier)
  • API available
  • Available in the browser, as a desktop app, and as a mobile app.

OpenAI’s conversational AI model for natural language processing and dialogue.

The free tier is getting more powerful. After you have exceeded your allowance for a few hours, you need to use a less powerful model

Gemini

Uses:

  • Chatbot
  • Search your Google Drive, YouTube, and more with extensions.
  • API available
  • Available in the browser and as a mobile app.
  • Can be used on Android as your Google Assistant.

The free tier works great and has a lot of functionality. Including uploading images and a button to read the text aloud.

Bing Copilot

Uses:

  • Chatbot
  • Search assistant.

This tool is integrated into the Bing website and can be found on the Copilot tab, linked above. It is great for searching as it has access to the internet and can provide links as references.

Bing also offers a browser-based assistant, which can be found in sidepanel of Microsoft Edge. It has context about the page you’re on so you can ask it questions or get it to summarize the page.

This is free and available as GPT-4 model. If you want more control such as keep your data private, then you can opt for the enterprise option.

Claude

Uses:

  • Chatbot
  • API available

Anthropic’s AI assistant for natural language processing and dialogue, which can assist with webpage development. If you enable the new artifacts feature, it can generate multiple code files at once and preview them in the sidepanel, such as HTML, CSS, and JS.

The free tier is limited. After a couple of requests, you need to use a less powerful model.

See also the anthropic.com homepage.

Local AI

Tools to download and run LLM models locally, without an API key. Run either as a CLI tool or with a desktop user interface. There are plenty alternatives out there but I like these so far.

LocalAI

Uses:

  • Chatbot.
  • Local AI (CLI)
  • Drop-in replacement for OpenAI API - wrap your mode in an API server.

This free, open-source alternative to OpenAI provides a drop-in REST API for local inferencing with LLMs, supporting image and audio generation on consumer hardware.

LM Studio

image

Uses:

  • Chatbot.
  • Local AI (desktop app)
  • Drop-in replacement for OpenAI API - wrap your mode in an API server.

This user-friendly desktop app allows you to download and run models locally, offering a convenient way to experiment with LLMs without having to use a CLI.

LM Studio lets you search for models on the Hugging Face Hub.

image

If you want more trusted models, you can search for models by the lmstudio-community account. Here are the most downloaded ones on that account:

image

Then you can load a downloaded model and chat with it.

image

The same prompt with another model:

image

Ollama

Uses:

  • Model hosting
  • CLI-based local AI for running LLMs. Supports various model formats like HuggingFace, LLaMA, etc.

This free, open-source CLI tool supports various model formats and allows you to run LLMs locally. It provides access to a collection of models for download.

On their website, there are open-source AI models available to download for free.

Fabric

Uses:

  • CLI-based tool for running LLMs.

This CLI-based tool allows you to run LLMs locally and can be configured with API keys to access external genAI APIs.

Hugging Face

Uses:

  • Model hosting
  • SaaS available on paid tier.

The Hugging Face Hub on the websites offers a vast collection of AI models for download for free. These can be used with other local AI tools such as the ones listed above.

Multiple models

Explore and compare different LLM and image models within your browser.

Replicate

Uses:

  • Chatbot
  • Image creation
  • Web-based playground for the above
  • API access available

This platform provides a browser-based playground for experimenting with various models, offering API access and documentation. Notably, check out the Flux models (for realistic fingers and text in images) and Meta’s famous 104b model (which is impractical to run locally at over 200GB):

The Replicate site is free up to a point, but then you run out of credits (they do not top up automatically) and then need to pay to access in the playground or API. Some models on the site have instructions for local development too.

Poe

Uses:

  • Chatbot
  • Image creation

On the Poe website, explore models by category and try them out. For example, you can use Claude, GPT-4, Flux, and Dall-E 3.

The site is free but there are additional benefits for premium members.

Image creation

See a detailed comparision in my post, AI art tools for 2023.

Music

Udio

Create and extend musical pieces, with or without words. You can also explore my Udio playlist here.