How we built OWL, the new architecture behind our ChatGPT-based browser, Atlas
OpenAI published a technical deep dive into OWL, the architecture underpinning ChatGPT Atlas, a new Chromium-based browser. The post covers how OWL decouples Chromium internals to enable fast startup, rich UI rendering, and agentic browsing capabilities integrated with ChatGPT. This represents OpenAI's move into the browser layer as an agentic interface, extending ChatGPT's reach into web navigation and task execution.
Related guides (3)
Related events (8)
Continuously hardening ChatGPT Atlas against prompt injection
OpenAI is applying automated red teaming trained with reinforcement learning to harden ChatGPT Atlas, its browser agent, against prompt injection attacks. The approach creates a proactive discover-and-patch loop to identify novel exploits before they can be weaponized. This work is framed as part of broader efforts to secure increasingly agentic AI systems against adversarial manipulation of external content.
OpenAI Introduces Canvas: A New Writing and Coding Interface for ChatGPT
OpenAI announced Canvas, a new collaborative interface for ChatGPT designed to support writing and coding workflows. Canvas provides a side-by-side editing environment where users can work with ChatGPT on longer-form documents and code projects. The feature represents a shift from purely conversational interaction toward a more structured, document-centric collaboration model.
Introducing ChatGPT Agent
OpenAI has launched ChatGPT agent, a new capability that combines reasoning with tool use to autonomously complete multi-step tasks such as research, bookings, and presentation creation. The agent operates under user guidance, integrating thinking and acting in a unified workflow. This represents OpenAI's move to bring agentic capabilities directly into the ChatGPT product for general consumers.
WebGPT: Improving the factual accuracy of language models through web browsing
OpenAI fine-tuned GPT-3 to answer open-ended questions more accurately by giving it access to a text-based web browser. The system, called WebGPT, uses reinforcement learning from human feedback to learn to search the web, read pages, and cite sources. This work represents an early demonstration of retrieval-augmented generation and tool-use in large language models.
OpenAI Spring Update: GPT-4o Announced, Expanded Free ChatGPT Capabilities
OpenAI announced GPT-4o, a new flagship model, alongside an expansion of capabilities available to free-tier ChatGPT users. GPT-4o represents a new omnimodal architecture capable of handling text, audio, and vision in a unified model. The announcement was made via a live demo event and marks a significant shift in OpenAI's product and model strategy.
Introducing ChatGPT Search
OpenAI has launched ChatGPT Search, a feature that provides fast, timely answers with links to relevant web sources directly within ChatGPT. This integrates real-time web retrieval into the ChatGPT interface, moving the product closer to a search engine replacement. The announcement comes from OpenAI's official blog, indicating a significant product expansion.
Introducing ChatGPT and Whisper APIs
OpenAI announced the release of dedicated APIs for ChatGPT (gpt-3.5-turbo) and Whisper, enabling developers to integrate conversational AI and speech-to-text capabilities into their applications. The ChatGPT API offered significant cost reductions compared to existing GPT-3.5 endpoints. This marked a major step in OpenAI's platform strategy, opening programmatic access to its most widely used consumer models.
smol-ai/GodMode: multi-model AI chat browser aggregating ChatGPT, Claude, Bard, and others
GodMode is an open-source TypeScript desktop app that provides unified browser-based access to multiple AI chat interfaces including ChatGPT, Claude, Bard, Bing, and Llama 2. The project has accumulated 5,536 GitHub stars with modest recent momentum (+14 today). It functions as a thin wrapper enabling side-by-side or rapid switching between frontier chat products.