Fulford Island, an OpenAI researcher, had the feeling that Deep research It would be a success even before it was launched.
Fulford had helped build the artificial intelligence agent, which autonomously explores the website, deciding for itself what links click, what to read and what to collect in an in -depth report. Openai first made deep research available; Every time he fell, says Fulford, was flooded with consultations of colleagues anxious to recover it. «The number of people who were Dming excited me quite,» says Fulford.
Since he went to the public on February 2, Deep Research has also proven to be a success among many users outside the company.
«Deep research has written 6 reports so far today,» Patrick Collison, CEO of Stripe Posted in x A few days after its launch of the product. «In fact, it’s excellent.
«Deep research is the AI product that really obtained a significant part of the DC policy formulation community to start feeling the AGI.» wrote Dean Ball, a member of the George Mason University who specializes in AI policies.
Deep research is available as part of the Chatgpt Pro plan, which costs $ 200 per month. A consultation is needed, such as «Write me a report on the Massachusetts Health Insurance Industry» or «Tell me about Wired coverage of the Government Efficiency Department», and then presents a plan, looking for relevant websites, combing through its content and deciding which links click and what information deserves a greater investigation. After sometimes exploring dozens of minutes, synthesizes your findings in a detailed report, which may include appointments, data and graphics.
Many tools currently marked as AI agents are essentially chatbots connected to simple programs without much sophistication. The deep research model itself goes through a type of artificial reasoning before devising a plan and advancing with each step. The model provides details of this reasoning behind your research in a side window.
«Sometimes it’s like ‘I need to go back, this does not seem so promising,» says Josh Tobin, another operai researcher involved in the construction of deep research. «It’s great to read some of those trajectories, just to understand how the model is thinking.»
Operai evidently sees deep research as a tool that could assume more office work. «This is something we can climb,» says Tobin, and adds that the agent could be trained to complete a specific white -collar job. An agent with access to the internal data of a company could quickly prepare a report or presentation, for example. Tobin says that the longest objective is «to build an agent that is not just good to build reports through the search for the web, but is also good in many other types of tasks.»
Because a deep investigation was trained to analyze and summarize the text written by humans, Tobin says that his team was surprised to see many people who used it to generate code. «It’s an interesting thread to throw,» he says. «We are not totally sure what to do with that.»
