OpenAI Brings Claude Artifacts and Cursor to ChatGPT

This week has been all about OpenAI – from securing $6.6 billion in funding to catapult frontier AI research, to shipping new capabilities, including real-time APIs, vision integration in the fine-tuning API, prompt caching (with 50% discounts and faster processing for recently seen input tokens), and model distillation, among others.?

OpenAI’s on Track

Today, OpenAI introduced canvas, a new interface within ChatGPT designed for writing and coding projects, enabling seamless collaboration with the AI through direct edits, feedback, and targeted revisions. It enhances productivity by leveraging GPT-4o's capabilities for more precise and iterative workflows.

Many developers are dubbing it a Cursor replacement, while others are still questioning its worth. “Somebody check on the Cursor team, please,” said Santiago Valdarrama, while others say that OpenAI is coming for all the VS Code forks.?

Similar to Claude Artifacts, canvas’ interface allows users to work on writing and coding projects side-by-side with ChatGPT. It is integrated with GPT-4o and can be manually selected in the model picker.

The recently released OpenAI o1 is a great model for reasoning, but falls short when it comes to coding abilities. Canvas makes up for it. OpenAI decided to use GPT-4 and GPT-4o for this coding interface, and not o1, which explains that o1 was never meant for coding.

Cursor Killer? Not Yet

With the introduction of canvas, ChatGPT is now able to better understand the context of user projects. Developers can highlight specific sections, allowing ChatGPT to focus on targeted areas and provide inline feedback and suggestions.        

“It can also review code, add logs and comments, fix bugs, and port to other coding languages like JavaScript and Python,” said OpenAI in the blog post. The team said that canvas improves decision-making accuracy for when to trigger coding tasks, reaching 94% compared to baseline models.

Users maintain full control of their projects in canvas, with the ability to directly edit text or code. A menu of shortcuts enables quick actions such as adjusting writing length, debugging code, and performing other tasks. The system automatically opens canvas when it detects complex scenarios and can be triggered manually by including “use canvas” in a prompt.

It seems like with just a few more updates, OpenAI can make a complete Cursor alternative. But since OpenAI is the first backer of Anysphere, the company building Cursor, it may not, after all, want to be in direct competition with them.?

Basically, canvas is a great code editor without any context of the codebase, routes, or logs. On the other hand, Cursor gives users complete control within the IDE, and you have the option to bring in other models.

Moreover, when it comes to killing Cursor, OpenAI is highly unlikely to get into the IDE business as that would mean competing with Microsoft’s VS Code, and not just Cursor. Besides, OpenAI has already partnered with Cursor for providing its o1 models on their AI editing platform.

OpenAI Doesn’t Want to Just be on the Platform Layer

Canvas feels more like a competitor to Claude Artifacts right now, but without the visualisation and the execution of code it writes, which also seems to be just another update away.?

People have been comparing canvas with Claude, Cursor, and GitHub Copilot, and claim that with this update, OpenAI might just make many shift back to ChatGPT for writing and coding, since the UX of ChatGPT is a lot better.

Simultaneously, according to sources, Microsoft has plans to upgrade its capabilities on the VS Code IDE, which would help it compete with Cursor.?        

Moreover, GitHub Copilot is also heavily investing in and introducing new features for its coding assistant platform. A few weeks ago, GitHub CEO Thomas Dohmke posted on X a video of GitHub Copilot in VS Code running with OpenAI’s o1 model, which he called “flat out badass”.

So, apart from providing models for other platforms, the company has decided to allow code editing on ChatGPT, which would make others in the market cry for help. All in all, the race to be the best AI coding tool seems to be getting intense with OpenAI entering the fray.?

Enjoy the full story here.?

LangChain is Great, but Only for Prototyping

“LangChain is not meant for production,” declared Praveer Kochhar, the co-founder of Kogo Tech Lab, at Cypher 2024. He was responding to a query regarding optimising infrastructure using LangChain.

KissanAI founder Pratik Desai went a step ahead when he wrote on X that “if your developer proposes LangChain for production use, fire them”. That was a bold statement, but there are multiple reasons why no one wants to use LangChain, besides for prototyping.?

Read the full story here.?

AI Bytes?

• Juniper Networks recently launched an AI-native local cloud instance in India, powered by Mist AI, to provide secure, resilient, and scalable wireless, wired, and SD-WAN solutions, aligning with India’s data protection laws.
• PwC India recently partnered with Meta to scale and democratise generative AI solutions using Meta’s Llama models, enabling enterprises and citizen services to access advanced AI capabilities, drive innovation, and solve real-world challenges, aiming to create a digitally empowered future for businesses and communities across India.
• StarTree showcased its real-time analytics capabilities at Current 2024, demonstrating how its cloud platform, powered by Apache Pinot, enables seamless observability through Prometheus-compatible time series databases and real-time Grafana dashboards.?