ChatGPT

ChatGPT is excited to announce a new partnership with the Council of Fashion Designers of America (CFDA) to launch The Innovation Hub, a two-year initiative aimed at empowering American designers and brands.

This program focuses on the creative and business benefits of artificial intelligence (AI) by pairing promising CFDA members with leading AI builders. Together, they will co-create tools that address current real-world challenges in the fashion industry. Key aspects of The Innovation Hub include:

• Collaborative Approach: Pairing fashion designers with AI specialists to develop practical solutions.
• Focus on Real Challenges: Addressing specific issues within the fashion industry, ensuring relevancy to designers' needs.
• Creativity and Business Longevity: Highlighting how AI can enhance creativity while supporting sustainable business practices.

This initiative is significant for small business owners in the fashion sector as it demonstrates the potential of AI in solving industry-specific problems. By participating in The Innovation Hub, designers can gain access to cutting-edge technology tools that could streamline operations, enhance creativity, and ultimately drive positive change in their businesses.

ChatGPT has introduced EVMbench, an innovative benchmark designed to evaluate the capabilities of AI agents in detecting, exploiting, and patching high-severity vulnerabilities in smart contracts. This new tool serves not only as a measurement framework but also as a crucial call to action for developers and security researchers to integrate AI-assisted auditing into their routines.

Key Features of EVMbench:

• Detection: Identifies vulnerabilities in real-world smart contract code.
• Exploitation: Simulates realistic attack scenarios to demonstrate the potential impacts of these vulnerabilities.
• Patching: Provides safe and effective fixes that have been tested for reliability.

EVMbench's release includes detailed tasks, tooling, and an evaluation framework intended to support ongoing research in managing and measuring the evolving capabilities of AI in cybersecurity.

This development is significant for builders and developers, as it underscores the growing necessity of incorporating AI into security auditing processes, thereby enhancing the security landscape of smart contracts.

A recent study published today highlights the impactful role of ChatGPT's LLM in clinical decision support, showcasing significant improvements in patient outcomes within neurosurgery. The research indicates that LLM-assisted interventions potentially prevented major complications in two-thirds of patients treated at low-resource settings.

Key findings of the study include:

• Improved Clinical Plan Quality: The implementation of OpenAI models enhanced the quality of clinical plans in 60% of mock rounds.
• Benefits for Medical Officers: Medical officers exhibited a 21% improvement; nurses experienced a more modest 6.5% increase, likely due to their narrower focus on patient management.
• Error Reduction: Significant error reduction from 33% to 0% was noted during case testing.
• Preventive Interventions: Of 14 clinical interventions across 9 patients suggested by the LLM, 6 interventions were credited with potentially preventing serious complications.

Why does this matter for builders and developers? This study underlines the potential of AI systems like ChatGPT to deliver substantial improvements in healthcare, especially in low-resource environments where expert human resources are scarce. The encouraging results have led to the establishment of the West Africa AI Neurosurgery Consortium, aimed at further validating these findings across five countries in the region.

OpenAI, the organization behind ChatGPT, has announced a commitment of $7.5 million to The Alignment Project, a global fund dedicated to independent research on AI alignment, initiated by the UK AI Security Institute. This significant investment reflects a strong dedication to fostering AI safety amid rapidly advancing technologies.

Key Details:

• Funding Amount: $7.5 million
• Project Name: The Alignment Project
• Initiator: UK AI Security Institute
• Aim: To build a robust, diverse, and independent alignment ecosystem in the AI field.

By supporting independent research, OpenAI aims to enhance the safety and reliability of advanced AI systems. This collaboration underscores the importance of community engagement in addressing the challenges posed by AI advancements.

For developers and builders in the AI space, this commitment signals a growing recognition of the need for ethical guidelines and frameworks that ensure AI technologies are aligned with human values and safety standards. The investment in independent research is crucial for promoting innovation while prioritizing safety in AI developments.

Over the past five months, the ChatGPT team has been experimenting with a revolutionary approach in software development: building and shipping an internal beta product with zero lines of manually written code. This innovative method leveraged Codex to generate app logic, tests, continuous integration, documentation, observability, and internal tooling, drastically cutting development time to about one-tenth of conventional coding methods.

Key Details:

• No Manual Core Code: All components were generated entirely by Codex.
• Reduced Time Investment: Estimated development time was reduced by 90% compared to traditional coding practices.
• Focus Shift: Engineers concentrated more on scaffolding systems, abstractions, and feedback mechanisms rather than on writing code directly.
• Efficiency in Review: The review process has transitioned primarily to agent-to-agent workflows, minimizing the reliance on human reviewers.
• Maintainability: Building for actual users ensured that choices were grounded in maintainability.

This approach is significant for builders and developers, as it highlights a shift in resource allocation within software development. While the future implications of entirely agent-generated systems remain to be studied—particularly concerning long-term coherence and the integration of human judgment—these insights can aid in strategic investments in automation tools and processes.

ChatGPT has just announced a significant update with new primitives aimed at improving the process of building agents. This development is essential for those looking to create more efficient workflows that require extended execution times.

Key Highlights of the Update:

• Introduction of 10 new tips specifically designed for managing multi-hour workflows.
• Enhanced reliability for running long-duration tasks using the new primitives.
• Detailed guidance available through the link provided, which offers practical insights for developers.

These updates are vital for builders and developers, as they facilitate smoother and more dependable operations for complex processes, ultimately leading to more robust applications and services.

ChatGPT has launched GPT-5.3-Codex-Spark, an ultra-fast model specifically designed for real-time coding. This model is currently available as a research preview for Pro users through the Codex app, Codex CLI, and IDE extension.

Key Features:

• Faster Performance: GPT-5.3-Codex-Spark is built on a shared production stack, complementing GPUs for critical low-latency workloads.
• Optimized Infrastructure: Enhancements include improved response streaming, accelerated session initialization, and restructured parts of the inference stack, which will benefit all Codex models in the coming weeks.
• Contextual Capabilities: It is currently a text-only model featuring a 128k context window, with future plans to expand into larger models and multimodal input.
• Early Access Program: A select group of API customers has been granted early access to Codex-Spark for product experimentation and optimization feedback.

This launch is significant for builders and developers who require faster coding solutions and enhanced performance for their applications. As access expands, more users will be able to leverage these advancements for their coding tasks, leading to increased productivity and efficiency in developer workflows.

OpenAI's GPT-5.2 has achieved a significant milestone in theoretical physics, collaborating with researchers from prestigious institutions such as the Institute for Advanced Study, Vanderbilt University, University of Cambridge, and Harvard University. This collaboration has resulted in a groundbreaking preprint that reveals a novel interaction involving gluons, which was previously thought to be impossible under normal conditions.

Key Highlights:

• New Formula: GPT-5.2 proposed a formula for a gluon amplitude that was later validated by an internal OpenAI model.
• Specific Conditions: The preprint demonstrates that gluon interactions can occur under certain circumstances.
• Extension to Gravitons: Following the initial findings, the research has extended to include gravitons, with further generalizations in progress.
• Collaborative Efforts: The results highlight the importance of collaboration in scientific advancement, emphasizing community engagement and feedback.

This breakthrough is not only a testament to the capabilities of AI in advancing scientific research but also showcases the potential for future discoveries in fundamental physics. For developers and researchers in AI and physics alike, this development opens new avenues for exploration and experimentation within the realm of theoretical models and computational predictions.

For more details, refer to the full post here.

ChatGPT has unveiled Lockdown Mode, a groundbreaking advanced setting aimed at bolstering security for organizations against prompt injection-based data exfiltration attacks. This feature is particularly targeted at employees deemed high-risk, such as executives and security personnel, who may be more susceptible to such threats.

Key Features of Lockdown Mode:

• Disables Network-Enabled Tools: To enhance security by minimizing exposure to potential attacks.
• Restricts Capabilities: Limits certain functionalities to ensure predictable behavior in sensitive contexts.
• Designed for High-Risk Users: Tailored for individuals facing increased risks of being targeted by malicious actors.

This update complements recent measures against URL-based data exfiltration, highlighting ChatGPT’s commitment to evolving its safety and security protocols to address novel threats. As AI systems continue to advance, ChatGPT is dedicated to improving safeguards to protect sensitive information effectively.

Importance for Builders/Developers:

Implementing Lockdown Mode provides an essential layer of security for organizations utilizing ChatGPT, ensuring that they can leverage AI tools without compromising sensitive data integrity. This is crucial for maintaining trust and compliance in a landscape of growing cyber threats.

ChatGPT's deep research capabilities are now powered by GPT-5.2. Users can connect to apps, search specific sites, track real-time progress, and view fullscreen reports. This update is rolling out starting today.

OpenAI introduces new primitives in the Responses API designed for long-running tasks. Key features include server-side compaction for extended agent runs, managed internet access for hosted containers to install libraries, and native support for Agent Skills with a new pre-built spreadsheets skill. Discover more about these updates.

Richard Lane in Reno is leveraging ChatGPT to modernize an 86-year-old salvage yard while preserving its core operations. By testing ideas and streamlining decisions, ChatGPT aids in maintaining the viability of one of Reno's fastest metal shops, fostering growth and team collaboration.

OpenAI has announced the availability of GPT-5.3-Codex, enhancing the capabilities within Codex. This new version aims to simplify the development process, allowing users to build applications more efficiently.

OpenAI partnered with Ginkgo Bioworks to develop a lab-in-the-loop system using GPT-5, leading to a 40% reduction in protein production costs. The model suggests experiments that are executed at scale, utilizing over 36,000 reaction compositions. Future plans involve applying this iterative approach to other biological workflows, enhancing efficiency in research and development.

A team of finalists at the OpenAI Codex Hackathon won $10,000 by developing Yolodex, an end-to-end game vision factory. Their pipeline automates video analysis, extracting frames and auto-labeling objects with Codex agents to create a training-ready YOLO dataset in minutes, significantly reducing the time needed for manual labeling. The innovative demo was presented to tech leaders, including Sam Altman. Demonstration and repository links provided.

OpenAI has launched a new developers page aimed at assisting developers in building AI applications in production. The page will feature real workflows, Codex builds, API and OSS updates, and insights from the community. Further updates are expected soon, providing valuable resources and best practices for AI development.

A recent town hall featured 50 AI builders who engaged in a candid discussion about the future of AI with Sam Altman. This insightful conversation provides perspectives from industry leaders on advancements and challenges in the AI field. Watch the full discussion to gain valuable insights into the direction of artificial intelligence.

OpenAI has launched its Codex App, providing new capabilities for AI coding assistance. Users report positive experiences and have already started building projects with it, highlighting its potential impact on productivity and creativity in software development. Early access impressions suggest promising features.

The Codex App Server enables use of Codex agents across IDEs, CLIs, desktops, and web apps via a bidirectional JSON-RPC API. This post includes a detailed breakdown of the protocol design and conversation primitives utilized in real products, enhancing developer experiences across various environments.

OpenAI has introduced Frontier, a platform designed to help enterprises build, deploy, and manage AI agents. Frontier equips agents with essential workplace skills and partners OpenAI engineers with enterprise teams to refine best practices. Major companies like HP, Intuit, and Uber are among its early adopters, with several others piloting the platform. OpenAI is also collaborating with various partners to enhance enterprise solutions.

OpenAI is partnering with Snowflake to integrate OpenAI models into Snowflake's Cortex AI and Intelligence. This collaboration will enable teams to build agents and gain insights from their enterprise data directly in the tools they use, benefiting Snowflake's 12,600 global customers. New features will be developed to support shared workflows, with Snowflake using ChatGPT Enterprise for enhanced decision-making efficiency.

The Codex app is now available on macOS, offering a command center for working with agents. Key features include multitasking with isolated worktrees, creating reusable skills, and automating workflows. For a limited time, it's accessible through ChatGPT Free and Go subscriptions, with doubled rate limits for Plus, Pro, Business, Enterprise, and Edu users. Download the app at openai.com/codex.

OpenAI has launched Prism, a free AI-native environment for scientific writing and collaboration. This tool enhances efficiency by integrating features like editable writing, diagram generation from images, and concurrent chat threads for citations and math checks. Alex Lupsasca, Kevin Weil, and Victor Powell showcase Prism's capabilities with GPT-5.2 within LaTeX projects. Discover more and try Prism at: https://prism.openai.com

Our partnership with NVIDIA is foundational, vital for training and inference. We scaled compute from 0.2 GW in 2023 to 1.9 GW by 2025, with exponential inference demand growth. We aim to anchor on NVIDIA while expanding partnerships with Cerebras, AMD, and Broadcom to enhance performance and reliability. This strategy enables us to support a growing number of real-world use cases at global scale.

OpenAI built an in-house AI data agent that reasons over 600+ petabytes and 70,000 datasets. It uses natural language to help Engineering, Product, and Research teams obtain trustworthy answers with contextual, table-level knowledge. The agent is Codex-powered and integrates product and organizational context.

OpenAI outlines three privacy initiatives: an AI Privacy Hackathon — after a debut in Dublin with 60+ participants, the next hackathon will be held in Dubai in April; a Privacy Filter being built to provide safeguards across the AI lifecycle and planned to be open-sourced for developers; and a Privacy Engineering Centre of Excellence in Dublin, with active hiring.

OpenAI has launched Prism, a free cloud-based workspace designed for scientists, powered by GPT-5.2. Prism allows unlimited projects and collaborators in a LaTeX-native format, reducing version conflicts and streamlining research workflows. It is accessible to anyone with a ChatGPT personal account and will soon be available for Business, Enterprise, and Education plans. OpenAI aims to enhance scientific collaboration and discovery through this innovative tool.

Last night, OpenAI hosted a Town Hall with 50 builders discussing critical questions about AI's impact on software development. Key topics included overcoming go-to-market challenges, adapting to rapid model improvements, and enhancing idea quality with AI. Watch the full conversation on YouTube for valuable insights.

AI is reshaping scientific workflows. As research grows more complex, researchers are using AI to navigate vast literatures, accelerate data analysis, and focus on higher‑value work such as experimental design and planning.

OpenAI CFO Sarah Friar and Khosla Ventures founder Vinod Khosla examine growing demand for compute, why compute is the scarcest AI resource, and approaches to widening access to AI benefits in a new OpenAI Podcast episode.

OpenAI launched the Education for Countries program at Davos to partner with governments and educators on scaling safe AI in learning. The initiative focuses on empowering educators, providing training, conducting longitudinal research on AI’s impact on cognition, and customizing AI tools. First cohort includes Estonia, Greece, Italy (CRUI), Jordan, Kazakhstan, Slovakia, Trinidad & Tobago, and the UAE.

OpenAI is spotlighting "Frontier Builders"—companies built natively on machine intelligence that run OpenAI models in real-world systems across security, biodefense, go-to-market, and customer experience. Featured companies include Decagon, Vanta, Unify, Valthos, and Clay.

Codex CLI is a local, cross-platform agent that orchestrates the user, model, and tools via an "agent loop" to make high-quality, safe code changes on your machine. The post explains how Codex builds and manages context, queries models, and applies reusable patterns for developers using the Responses API.

ChatGPT supports millions daily by breaking down medical information, helping users prepare for doctor appointments, and assisting in managing overall wellbeing, highlighting its role in personal health management.

A significant number of Americans find the US healthcare system complex and challenging. Many are utilizing ChatGPT to better understand and navigate healthcare information. This highlights the growing role of AI tools like ChatGPT in supporting patients, providers, and the public in managing healthcare complexities.

Steve, a heart failure patient, uses ChatGPT to help manage his health by tracking his diet, medications, and inflammation as part of his doctor's care plan, demonstrating practical application of AI in personal health management.

Burt utilizes ChatGPT to manage his life with two forms of cancer, leveraging the AI to understand medical scans, prepare for doctor appointments, and simplify complex medical information for his family. This highlights ChatGPT's role in assisting users with healthcare challenges by providing clarity and support in complex medical communications.

OpenAI announces ChatGPT Health, a new feature enabling users to connect electronic medical records and wellness apps for personalized support such as explaining test results, preparing for doctor visits, advising on diet and workouts, and comparing insurance plans. Developed with physicians, ChatGPT Health prioritizes privacy by isolating health conversations from regular chats and excluding data from model training. Currently in limited rollout, broader access will come soon.

A user shared how ChatGPT assisted her during her teenage son’s cancer treatment by translating medical reports, preparing questions, and facilitating more informed conversations with healthcare providers. This illustrates a practical application of AI in healthcare communication, enhancing patient and caregiver understanding and engagement.

OpenAI has launched OpenAI for Healthcare, an enterprise product suite designed to assist hospitals, clinics, and academic medical centers in delivering consistent, high-quality care. The platform supports HIPAA-compliant usage and includes partnerships with notable institutions such as AdventHealth, Baylor Scott & White Health, and Stanford Medicine Children's Health.

OpenAI highlights ChatGPT's extensive use in health, with hundreds of millions of weekly health and wellness queries. An upcoming OpenAI Forum conversation with CMO Kate Rouch and Dr. Nate Gross will explore how ChatGPT Health assists users in understanding medical info and preparing for doctor visits. Kate shares how personal AI experiences influenced the product's development.

OpenAI announces that ChatGPT Go, a new low-cost subscription tier, is now globally available where ChatGPT is offered. They will begin testing ads in the U.S. for free and Go tiers, ensuring ads are separate, clearly labeled, and do not influence responses. Pro, Business, and Enterprise tiers will remain ad-free. The approach prioritizes user trust, transparency, and conversation privacy from advertisers.