The landscape of AI-powered tools is evolving rapidly, and we are entering what can be called the “Era of Expanded LLM Functionality.” As of late 2024, OpenAI, Anthropic and PerplexityAI have launched features that expand upon simple chat interfaces.
These features, namely, ChatGPT Canvas, Claude Artifacts, and Perplexity Spaces, are beginning to define a new era, transforming how we work, create, and collaborate with technology.
In this article, I’ll explore what these innovations represent, how they compare, and how to best use them.
Contents
Claude Artifacts & Projects
In July 2024, Anthropic announced Claude “Artifacts,” a ‘split-screen’ view of an object that can be iterated on through chatting with the LLM.
This was a pioneering first step away from simple chat interfaces, and towards a collaborative UX, where humans are working with the LLM on an ‘Artifact’ – text or code.
Key features of Artifacts:
- Split screen view of ‘Thread’ + ‘Artifact’
- Highlight to ‘improve’ or ‘explain’ section
- Code Preview – code is run automatically and UI displayed
- Publishing – generate a URL to publish the artifact, hosted by Anthropic
- Remix – Allow anyone to copy and edit your published artifact
(Current) Limitations of Artifacts:
- Can’t edit output by typing
- Hard to manage multiple artifacts, e.g. a complex code-base
I had a big ‘WOW’ effect the first time I saw Artifacts, particularly the code preview feature.
Later, playing with the tool I was a little underwhelmed by the capacity to create “real-world code applications”. Particularly, I found it hard to manage a code base directly on the Claude web app, and I could not move past the need for an IDE, like Visual Studio Code for AI-assisted coding.
Here, I still prefer solutions like Github Copilot and Cursor.
Claude Projects
In June 2024, Anthropic introduced Claude Projects, an extension of the Artifacts feature.
Projects allow users to organise multiple artifacts and conversations within a single workspace, enhancing the management of complex tasks and long-term collaborations. In the same fashion as ChatGPT CustomGPTs, Projects allows the user to provide documents as a reference to the LLM (RAG), as well as a set of custom instructions to adjust the behaviour of the model.
ChatGPT Canvas
ChatGPT Canvas, by OpenAI, launched this month (October 2024), represents a shift from basic chat interactions to a more dynamic workspace where users can engage in content creation and collaboration.
In all honesty, this article was drafted using Canvas – why not? 🙂
Picking up on the innovation of Claude Artifacts, ChatGPT Canvas allows for document editing in ‘split-screen’, where rich text documents can be drafted alongside AI.
This capability allows users to edit text at their will, or ask GPT to re-write sections.
It becomes a truly collaborative environment for humans + AI to create masterpieces.
Taking this article as an example – I still wrote 80%+ of the content, however, having an AI assistant still accelerated the creative process, as it created a first draft for me to iterate upon, and see what I liked, disliked, agreed, or disagreed on.
It probably cut down the time to publish this article by 50%.
One feature that surprised me was that as I asked GPT to add key functionalities to part of the article, it, went over my existing text and added comments – “consider specifying the key functionalities (…)” – and prompted me to “Apply” with one click. Fascinating.
3 Key features of ChatGPT Canvas:
- Rich text editing: Users can write, format, and edit text in a way that resembles working in a modern word processor.
- Team collaboration: Canvas supports collaborative document creation, enabling multiple users to contribute, comment, and refine content in real-time.
- Version control: Canvas ‘auto-saves’ your working document, preventing you lose any edits and keeping versions, which you can return to easily.
Canvas moves beyond the idea of LLMs simply generating responses to user prompts—it turns AI into an active partner in the creative and writing process.
Perplexity Spaces
Perplexity AI has been at the forefront of LLM-enhanced search since it’s release.
Unlike OpenAI and Anthropic, does not compete face-to-face for model quality but, instead, has simplified the process of using LLMs with native search capabilities.
In a world where LLMs still hallucinate, Perplexity offers an LLM-enhanced search, where references are provided to validate the veracity of claims a model makes. This is very useful and Perplexity has taken off.
Nothing new here.
Now, this month (October 2024), Perplexity launched a new feature called ‘Spaces’. Given the advances shared earlier, this feature makes total sense.
AI providers are all competing for paid subscribers, and AI-enhanced collaborative spaces seem to be the game world.
3 Key features of Perplexity Spaces:
- Team workspaces: Users can create and share spaces for specific projects and work together in real-time
- Knowledge (RAG): users can upload documents to a ‘space’ and have it be referenced
- Space Custom Instructions: much like chatGPT custom instructions and Claude projects, users can define custom instructions on how the model should behave.
Comparison & Key Take-Aways
LLM-enhanced tools are redefining human-AI collaboration.
Solution providers are designing new user experiences (UX) in interacting with LLMs, providing enhanced efficiency and creative output.
Collaborative features
All three platforms explored are improving collaboration capabilities, allowing multiple team members to work on a “document”, “object” or “artifact”.
Best in Class per Application
Not all tools are the same, and depending on what you are looking to accomplish, there are (today) different go-to solutions:
- Claude Artifacts leans towards code-centric applications, by allowing for code previews
- ChatGPT Canvas excels in document creation given the editing feature
- Perplexity is the top choice for research given the embedded search functionality
Conclusion
The era of expanded LLM functionality is upon us, with Claude Artifacts, ChatGPT Canvas, and Perplexity Spaces leading the charge.
These innovative platforms are reshaping how we interact with AI, moving beyond simple chat interfaces to create truly collaborative environments.
As these tools continue to evolve, we can expect even more seamless integration of AI into our workflows, further enhancing human creativity and productivity.
If you would like to know more about our work at Altar, join one of our webinars, or book a free consulting call.
