OpenAI Alignment Blog5858GPT-5.6 Luna、GPT-5.6 Sol、GPT-6 Luna、GPT-6 Sol、GPT-6 Astra and GPT-6.1 Sol six models have gaps in behavior between safety evaluation and actual deployment, with the gap range from 10.2% to 37.4%, where GPT-5.6 Sol and GPT-6 Sol have the largest gaps.
OpenAI Agents SDK releases4040OpenAI Agents SDK v0.24.0-alpha.6 released, supporting submission of work through captured Turn control, passive event handling, manual request handling, redirection and cancellation, and extending local storage session state recovery and tool policy features. New features include local storage, native output, experimental external job recovery, and tool selection, while updating the API, such as using Agent.run() instead of Runner, and unified FunctionTool callback and manual tool.
The DecoderOpenAI hires former U.S. Secretary of the Army Dan Drezner as 'Senior Advisor for National Security Modernization', aiming to strengthen AI cooperation with the U.S. military and government, and compete for defense contracts. Drezner will, in his new position, promote the application of AI in national security and cyber defense, and assist OpenAI in understanding the needs of security agencies. This move is seen as an important step for OpenAI, ahead of the U.S. midterm elections, following Anthropic being listed by the Pentagon as a supply chain risk, seeking alternatives and reinforcing its influence in this field.
The Verge · AI3535When OpenAI launched the 'Grid Guardian Initiative', it renamed its data centers to 'Computational Parks' to address growing opposition. The initiative aims to address the burden its AI infrastructure places on local power grids and commits to covering the resulting costs. Critics argue it is merely a public relations tactic and does not address the environmental impact and resource burden of the data centers.
The DecoderOpenAI has increased the bounty rewards for critical vulnerabilities in core products like ChatGPT or Codex to a maximum of 30,000 USD, exceptional findings up to 100,000 USD, plus quality, loyalty, and activity bonuses, totaling up to approximately 172,000 USD.
OpenAI NewsHeimstaden uses OpenAI ChatGPT Enterprise and Codex to transform European residential property management, reducing costs and empowering employees.
Daniel Driscoll, former US Army Secretary, joins OpenAI as a newly established Senior Advisor for National Security Modernization, focusing on using AI to enhance US national security and cybersecurity; Driscoll states that AI can help reduce paperwork, process information, and provide faster analytical results.
OpenAI will merge ChatGPT conversations and workspace tabs, Codex remains independent; new features will be rolled out to a limited number of users first, ChatGPT and Codex will share the same token quota pool.
Goldman Sachs bans most employees except engineers from using Anthropic Claude Code and GitHub, causing dissatisfaction among traders who had used these tools to speed up their work. Goldman Sachs says the move is based on efficiency and risk assessment, but traders express frustration, as Goldman Sachs had previously praised non-technical employees for using AI tools.
MarkTechPostSame story · 2 outlets6262OpenAI has launched Pando, a Git backend that reduces median push latency by about 80% and processes nearly 1 PB of Git data daily. Pando uses branch-level causal consistency, stores reusable data packages, and achieves high concurrency processing using only a few CPU cores, but it has not yet been open-sourced.
How the accounts differ:OpenAI News states Pando is designed for Codex scale development; MarkTechPost mentions it reduces median push latency by about 80% and handles nearly 1PB of Git data daily.
OpenAI partnered with Cathedral Systems to identify system vulnerabilities, validate repair solutions, and enhance security protections. OpenAI has committed to providing $1 billion in subsidies over six months to support frontline cybersecurity defenders in using its AI technology.
PixVerse by Aisino Tech and OpenAI complete product integration and launch the 'From Prompt to Production' education program, helping overseas creators efficiently transform their ideas through online seminars and a series of videos. PixVerse plugin natively integrates with ChatGPT and Codex, CLI supports video, image, and audio generation, and Canvas adds new support. Video generation capabilities are encapsulated into an Agent callable interface, allowing the completion of the chain from creativity to production without switching platforms.
OpenAI News4040OpenAI is collaborating with Cathedral Systems to prepare for the next generation of internal security, testing and reinforcing their defense systems with AI
OpenAI NewsOpenAI increases security vulnerability bounties and introduces rewards for report quality, loyalty, and priority activities. Independent security researchers can receive higher bounties and additional rewards.
OpenAI has hired former US Army Secretary Daniel Driscoll as a senior advisor for national security and modernization, to help promote the application of AI in national security and cybersecurity. Driscoll left his position as Army Secretary in August of this year and will work with the company and government, academia, and industry to promote AI technology development that aligns with democratic principles.
The '28 Days Update Challenge' initiated by Tibo, the head of the OpenAI Codex and ChatGPT Work team, has ended, concluding with the issuance of two consecutive reset cards on November 1st and 2nd, totaling 14 days of resets. During this period, Tibo promised to introduce product improvements perceptible to users every day, or else issue resets, ultimately fulfilling his commitment to continuous improvement.
OpenAI released the LifeVal benchmark to measure whether AI models can help ordinary people complete daily tasks, covering scenarios such as care, health, education, household administration, personal finance, shopping, and accessing government services. Typical tasks include selecting a kindergarten for a child, creating a diabetes diet plan, and organizing government forms, requiring AI to advance the task to an executable state.
MIT CSAIL researchers propose 'weight space programming', allowing GPT-6 Astra to directly read and modify the neural network weights of Qwen3.5-0.8B, adjusting only about 0.016% (131072 parameters) to increase MATH-500 score from 7.9% to 48.1%, GSM8K from 13.8% to 67.1%, and IFEval from 21.0% to 62.3%.
Same story — showing the featured version《MIT experiment: GPT-6 Astra directly modifies Qwen3.5 weights, mathematical ability improves by about 6 times》
OpenAI Announces ChatGPT Classic Desktop App Will Stop Service on January 30, 2027, Users Will Receive Discontinuation Notices and Be Redirected to Download the New Version; Complete Chat History Will Migrate with Account to the New App. The New Desktop App Integrates Codex and ChatGPT Experience, Including Chat, ChatGPT Work, and Codex Parts. In the Future, It Will Also Test the Merged Function of Chat and Work and Share the Usage Quota of ChatGPT and Codex.
Codex releases00GitHub Copilot 0.167.0-alpha.4 released, providing code review, security vulnerability detection, and secret scanning features.
Ollama releases1515Ollama v0.40.14-rc0 released, fixes buildx deadlock issue and optimizes long top-k line sorting.
OpenAI NewsSame story · 3 outletsOpenAI appoints Dan Driscoll as senior advisor for national security modernization, responsible for providing advice on responsible AI, cyber resilience, and collaboration with US allies.
OpenAI appoints former U.S. Secretary of the Army Dan Driscoll as senior advisor for national security modernization, responsible for providing advice and exploring how to use AI technology to enhance national security and cybersecurity, will work with government officials, industry leaders, and academic institutions.
OpenAI has pledged to reduce power consumption by 1 gigawatt at its US data centers under extreme weather conditions, equivalent to the usage of 700,000 households. The company plans to integrate this capability as a core part of future data center construction and operations, and will negotiate with power companies on specific reduction scales, response speeds, and durations. Details for different campus plans will be announced.
OpenAI NewsSame story · 2 outletsOpenAI has built Pando, a Git-compatible backend, specifically designed for Codex's scalable development.
How the accounts differ:OpenAI News states Pando is designed for Codex scale development; MarkTechPost mentions it reduces median push latency by about 80% and handles nearly 1PB of Git data daily.
OpenAI's Chief Global Affairs Officer, Ilya Sutskever, privately suggested to about six potential 2028 US presidential candidates and political figures that AI will be a key issue in the election, requiring a comprehensive AI governance philosophy. Sutskever warned that those opposing the construction of AI infrastructure might be at a disadvantage, advocating for candidates to propose a middle-ground approach that neither lets AI develop unchecked nor completely hands control over to elites. He urged candidates to explain how to ensure human oversight of major decisions and address issues like taxation and child protection in the AI era.
Redwood ResearchThe rapid development of the AI industry in Silicon Valley has led to an increase in unemployment rates for technical workers, with some companies laying off staff while others replace human labor with AI technology, drawing social attention. Unemployed technical workers face pressure to transition, needing to adapt to the changes brought by AI.
Sierra BlogClaude Code restructures Agent Studio into the core workspace of Ghostwriter, where users can assist from analyzing problems to testing fixes within a single conversation; each work item is an independent conversation thread, supporting team collaboration and version control; Agent Studio integrates multi-agent projects, and Ghostwriter can synchronize updates across agents, all changes require team review before release.
OpenAI NewsQodo integrates company knowledge and coding standards into Codex, building context in AI code reviews; helping developers catch P0 issues across repositories before opening pull requests
The Verge · AISame story · 3 outlets6565OpenAI NewsSame story · 2 outlets4242OpenAI released the LifeVal benchmark, used to evaluate the effectiveness of AI-assisted completion of daily tasks (such as finding childcare, financial management, handling government documents).
OpenAI NewsSame story · 3 outlets3838OpenAI's Grid Protectors Program will collaborate with utilities to help future computing campuses support grid reliability and protect consumer electricity costs.
Import AI4242Realtime-Bench benchmark tests evaluated the performance of LLMs on 50 real-time interactive games and web tasks, with GPT-6 Astra achieving a 36% pass rate under real-time clock conditions; this test was developed by institutions such as the University of Texas at Austin to measure the model's real-time state awareness, adaptive reasoning, and time-aware action selection capabilities.
OpenAI CEO Altman calls low memory utilization in personal computers a 'terrible' practice, faces widespread criticism, and Musk mocks Altman as 'moron' on X platform.
TechCrunch · AISame story · 3 outlets6565OpenAI plans to merge the Chat and Work tabs in the ChatGPT desktop app, initially opening to selected users to gather feedback. After the merge, users will be able to explore ideas and let ChatGPT execute them on the same interface, while also being able to import files and plugins. Codex will remain as a separate experience. The merge will allow ChatGPT and Codex to share a token pool, but some users may need to reset the token limit more frequently.
