Recently, OpenAI unveiled canvas, a new interface for working with ChatGPT on writing and coding projects. Many wonder if it’s better than Claude Sonnet 3.5 Artifacts. The answer is no.
Here’s why: canvas uses GPT-4o, and it is not better at coding than Claude Sonnet 3.5. While canvas has some good features for developers, like users’ collaboration and version control, it lacks critical features like code preview. Many people are now creating their first apps with Claude. These include niche apps that only a few people need, internal tools for small companies, and silly apps just for fun.
For instance, a user recently tried to visualise a dual monitor setup for a desk and compared how the monitors would fit next to each other. Claude, in an instant, created a perfectly usable app for this purpose in just a few interactions.
The app wasn’t particularly mind-blowing, but that was precisely the point. While it might not have been worth the time and effort for the user to code it themselves, having Claude create it in less than 5 minutes made it a valuable tool.
This scenario perfectly illustrates the potential of AI-assisted app creation for quick, personalised solutions.
Canvas is Cute, but Artifacts is the Real Deal
Apart from building software in a few minutes, users have also reported how Claude Artifacts is helping them learn coding. A developer who was struggling to learn how to code said that by being able to see how a project actually comes together more thoroughly, he has started to get a lot of the concepts that earlier struggled to comprehend.
“One of the hardest things, at least for me, is self-learning since I am a very conceptual learner. I need to be able to wrap my head around the whole idea and drill down into it. Alternatively, I start with a part and kind of build like a snowball of knowledge, something that’s usually a collaborative approach, making it difficult to do entirely on my own.
“However, with Claude, I’m having a wonderful time, with my only limitation being the amount of responses I can use at a given time,” he added.
AIM also experimented with Artifacts, and we were able to create a Cricket Quiz game, Temple Run, and Flappy Bird, all with a single line of prompt in English. “On-demand software is here,” said Joshua Kelly, chief technology officer of Flexpa, a healthcare tool company. Using Artifacts, he built a simple stretching time app for his runs in just 60 seconds.
With Artifacts, everyone can now create simple apps, and those apps can be shared with users. This means, we are slowly moving towards a phase where everyone would be an app developer, all thanks to Artifacts.
Even if we don’t consider Artifacts and only think of Sonnet 3.5 for coding, it turns out to be a better choice than a majority of the models.
Claude Sonnet, FTW
A few weeks ago, an electrician who had zero experience with software development built a multi-agent JavaScript application using Claude 3.5 Sonnet. The developer calls this Panel of Experts (POE), which uses multiple AI agents to process queries in a multi-step approach, all of which are done by just defining the prompt in a high-level manner.
This was just one example. There are multiple users who have praised Claude Sonnet 3.5 for its coding capabilities. One developer said on Reddit that it is absolutely phenomenal. “I am so impressed by its coding abilities. Feels like my productivity went up 3.5x this past few days. Really amazed by what I managed to ship, this is mainly due to Claude,” he added.
Developers with decades of experience are also praising Claude’s programming capabilities and how it is helpful in relieving brain strain. One developer with over 40 years of experience said that having something like Claude is a huge benefit.
Users have also reported that Claude Sonnet is one of the very few models which can handle large projects. Due to this, Sonnet is often compared to a mid-engineer with unlimited memory.
Claude Sonnet is not limited to giving a working code in a few shots but is also considered one of the best when it comes to debugging. Liu Grey, a data scientist from Austria, mentioned on X that Sonnet fixed a bug instantly and then went ahead and improved her render pipeline and doubled the update rate without increasing CPU, GPU or memory usage.
What about Reasoning?
OpenAI o1 models are often considered one of the best for reasoning. But things took a turn when, a few days ago, a user who used reflection over Claude Sonnet 3.5 achieved better results in terms of reasoning compared to o1.
It is worth noting that developers are also pushing Claude as much as they can. A good example of this would be the open-source VSCode extension for Claude, known as Cline (previously Claude Dev). This extension gets updated on a regular basis. A user mentioned that with zero coding knowledge, he managed to build a web application in just one day.
Things are clear now. Developers love Claude Sonnet and are developing solutions around it. Sure, OpenAI is also trying to catch up with tools like canvas for developer audiences but we are yet to see the impact.
On the other hand, Anthropic is also focused on bringing the best experience for their developer audience with initiatives like Projects and Artifacts, making it a win-win situation for both developers and the company.
The post Claude 3.5 Brushes Off Canvas with a Stroke of Code appeared first on AIM.
