Openai enhances its AI-driven “deep research” capabilities with the ability to analyze your codebase on GitHub.
On Thursday, Openai announced what was called ChatGpt Deep Research’s first “connector” to search the web and other sources to compile an in-depth research report on the topic. Currently, ChatGpt Deep Research can be linked to GitHub (beta), allowing developers to ask questions about their codebase and engineering documentation.
According to an OpenAI spokesperson, the connector will be available to ChatGpt Plus, Pro and team users in the coming days, with enterprise and EDU support coming soon.
GitHub connectors for ChatGpt Deep Research arrive to make AI-powered chatbots more convenient by building ways for AI companies to link AI-powered chatbots to external platforms and services. For example, humanity has debuted its recently debuted unification. This will allow the app to provide a pipeline to the AI chatbot Claude.
Openai provided a plugin feature for ChatGPT several years ago, but it has been removed in favor of a custom chatbot called GPTS.
“I often hear users think that ChatGpt’s deep search agents want to connect to internal sources in addition to the web,” Nate Gonzalez, head of Business Products at Nate Gonzalez, wrote in a LinkedIn blog post. “(That’s why) we’re introducing our first connector today.”
In addition to answering questions about the codebase, the new ChatGPT deep search GitHub connector allows ChatGPT users to understand how to split product specifications into technical tasks and dependencies, summarise code structures and patterns, and implement new APIs using real code examples.
TechCrunch Events
Berkeley, California
|
June 5th
Book now
Of course, there is a risk that Deep Research will hallucinate. Of course, the AI models that exist do not form things with confidence. However, Openai is casting new abilities as a potential timesaver rather than as a replacement for experts.
An Openai spokesperson said that because ChatGPT respects the organization’s settings, users are only viewing github content and already viewing codebases that are explicitly shared with ChatGpt.
Openai is investing in supporting coding tools, and recently announced an open source coding tool for terminals called Codex CLI, upgrades the ChatGPT desktop app to read code with a developer-centric coding app. The company considers programming as the biggest use case for its model. A good example is that Openai reportedly has reached an agreement to buy coding assistant Windsurf with AI for $3 billion.
In other Openai news on Thursday, the company launched tweak options for developers looking to customize new models for specific applications. Developers can now tweak Openai’s “inference” model. Openai can invoke enhanced fine tunings via techniques that use task-specific grading to improve model performance. The fine tuning is also available on the company’s GPT-4.1 Nano model.
According to Openai, only validated organizations can tweak the O4-Mini. Meanwhile, GPT-4.1 nano tweaks are available to all paid developers.
Openai has begun measuring specific models and developer features. The company argues that abuse needs to be prevented.