OpenAI is enhancing its Codex platform with new plugins and skills designed to improve task execution. Users can now browse a library or create custom plugins accessible via the top-left interface. Skills function as customizable workflows that teach Codex specific procedures, such as formatting newsletters or maintaining a brand voice. By pressing the dollar sign, users trigger pre-defined processes that allow Codex to follow team rules while accessing data from connected tools like Google Drive.
Codex customization options have expanded significantly, allowing users to manage settings through the bottom-left corner of the interface. Key features include preventing the computer from sleeping during long tasks and choosing between a detailed coding mode or a cleaner default output. Users can also set a preferred tone, add custom instructions, and place an avatar on the screen to assist with multitasking. New users are advised to start with core settings before making broader adjustments.
Concerns over AI safety and ethics are growing globally. Baltimore County Councilman Julian Jones condemned an AI-generated image depicting him in a compromising situation, which he claims weaponizes anti-Semitic tropes against his election campaign. Meanwhile, the Guardian questions who controls the internet's future after Anthropic's Claude Mythos model found 40 vulnerabilities in widely used software. The US government has designated Anthropic a critical technology provider, raising questions about the concentration of power among top AI creators.
Practical applications of AI are reshaping commerce and daily life. Ulta Beauty is deploying a Google-powered AI shopping assistant across Search and the Gemini app, utilizing insights from 46 million members for personalized recommendations. Conversely, scammers in Washington state used fake AI-generated trading signals to steal nearly $9.9 million from investors through fraudulent cryptocurrency platforms. These developments highlight both the utility and the risks of generative AI in modern society.
Key Takeaways
['OpenAI upgraded Codex with plugins and skills to handle specific workflows like formatting newsletters and maintaining brand voice.', 'Codex users can now customize settings to prevent computer sleep, choose coding modes, and add avatars for better multitasking.', 'Baltimore County Councilman Julian Jones condemned an AI-generated image depicting him in a compromising, anti-Semitic manner.', "Anthropic's Claude Mythos model identified 40 vulnerabilities in widely used software, prompting US government designation as a critical technology provider.", 'Ulta Beauty launched a Google-powered AI shopping assistant using Gemini Enterprise to provide personalized guidance to 46 million members.', 'A Washington state fraud ring used fake AI-generated trading signals to steal nearly $9.9 million from investors.', 'Humanities educators argue courses must shift from essay-based grading to in-class discussion to combat AI-assisted bypassing of learning.', "Samsung's labor union is demanding 15% of operating profits from its chip division, potentially worth over $400,000 per worker.", 'Wrongful arrest of Angela Lipps in Tennessee highlights the dangers of relying on AI facial recognition without basic investigative verification.', "Jeff Bezos' 'Day 1' thinking and the 70% information rule are being applied to reduce AI-induced procrastination and improve productivity."]OpenAI Codex Adds Plugins and Skills for Better AI
OpenAI is upgrading its Codex platform with new plugins and skills to improve how the AI handles tasks. Users can manage these features through the top-left corner of the interface, where they can browse a library or create their own plugins. Skills act as customizable workflows that teach Codex specific company procedures, such as formatting newsletters or maintaining a brand voice. To use a skill, users press the dollar sign and select the desired option to trigger a pre-defined process. This combination allows Codex to follow team rules while accessing data from connected tools like Google Drive.
OpenAI Codex Lets Users Customize Settings Easily
OpenAI Codex now offers detailed customization options for personalization, task management, and the user interface. These settings are found in the bottom-left corner of the interface and allow users to control how the AI behaves. Key options include preventing the computer from sleeping during long tasks and choosing between a detailed coding mode or a cleaner default output. Users can also set a preferred tone, add custom instructions, and place an avatar on the screen to help with multitasking. New users are advised to start with a few core settings before adjusting everything.
Councilman Condemns Misleading AI Image of Him
Baltimore County Councilman Julian Jones strongly condemned an AI-generated image that showed him in a compromising situation. The image, which depicted him wearing a Keffiyeh and a mustache like Hitler, circulated on social media and was described as deeply disturbing. Jones stated that the image spreads false information and urged people to be careful about what they share online. He called on social media platforms to take stronger measures to stop the spread of AI-generated misinformation. The incident has sparked a broader conversation about the ethical use of AI and the need for regulations to prevent its misuse.
Jones Targets Negative AI Memes in Election
Baltimore County Councilman Julian Jones is running for county executive during an election year and has become a target of offensive AI images. The circulating meme portrays him in a way that seems to attack the Jewish community, which Jones says is being used to weaponize against his campaign. Howard Libit, the Executive Director of the Baltimore Jewish Council, called the meme despicable and stated that voters are too sophisticated to be swayed by fake photos. Jones hopes that seeing these negative political memes will draw attention from politicians on both sides of the aisle to denounce such attacks.
Guardian Questions AI Control Over Internet
The Guardian discusses Anthropic's Claude Mythos, an AI model that found 40 vulnerabilities in widely used software and 10 times more flaws in Firefox than before. The US government has embraced Anthropic by designating it a critical technology provider and awarding it a contract to secure systems. This success raises questions about who controls the most powerful AI models and how to ensure they are used responsibly. The article argues that whoever creates the most powerful AI will gain significant power, so robust regulations and oversight are needed to protect the internet's future.
Humanities Classes Need New Structure for AI Age
A student at WashU argues that humanities departments must change their course structures because AI makes it too easy to bypass genuine learning. Currently, many students take humanities classes only to fulfill graduation requirements, using AI to write essays and summarize readings without actually engaging with the material. This structure allows students to succeed without doing the hard work, reinforcing the prejudice that humanities are less rigorous than STEM fields. The author suggests that courses should place more weight on in-class discussion and real-time thinking rather than relying heavily on essays.
Keeper Security Protects Humans, Machines, and AI
Keeper, a leader in password and secrets management, is launching a next-generation privileged access management solution. This new system deploys in minutes and integrates with any tech stack to prevent breaches and ensure compliance. The company is backed by Insight Partners and is trusted by millions of individuals and thousands of organizations. Darren Guccione, CEO and Co-Founder, discussed the platform at the RSAC Conference 2026 in San Francisco.
Scammers Use AI and Social Media to Steal Millions
Washington state regulators allege that a network of fake investment foundations and cryptocurrency platforms stole nearly $9.9 million from investors. The group used social media to lure victims with fake AI-generated trading signals and promised free investment training. At least 38 investors lost money, including three Washington residents who invested about $49,000. The fraud involved fake trading platforms, advance-fee scams, and unregistered third parties to convert funds into cryptocurrency. The Washington State Department of Financial Institutions filed charges and plans to order fines totaling $1 million.
Wrongful Arrest Highlights AI Facial Recognition Risks
Angela Lipps was wrongly arrested in Tennessee based on an AI-powered facial recognition match that claimed she resembled a woman in bank surveillance footage. She sat in jail for several days while no one from the Fargo Police Department checked her records or verified her identity. The charges collapsed on Christmas Eve after officials found she had never traveled to North Dakota. The case highlights the danger of relying on unreliable software without basic investigative work like phone calls or record checks. The author argues that states must pass strong standards to ensure AI is used responsibly in law enforcement.
Bezos Day 1 Rule Stops AI Procrastination
An author describes using Jeff Bezos Day 1 thinking to fix a ChatGPT workflow that was causing procrastination. The author realized they were spending too much time planning perfect prompts and projects instead of starting work. They applied the 70% information rule, which means making decisions with incomplete data to avoid overthinking. This approach helped them stop spinning their wheels and actually begin tasks. The author notes that this method killed their procrastination and improved their productivity with AI tools.
Ulta Beauty Launches Google AI Shopping Assistant
Ulta Beauty is deploying a Google-powered AI assistant and agentic commerce across Google surfaces like Search and the Gemini app. The new Ulta AI shopping assistant uses insights from 46 million members to provide personalized guidance to shoppers. This feature is built with Google Gemini Enterprise for Customer Experience and was introduced in January. Shoppers can receive product recommendations, compare options, and complete checkout within Google's conversational interfaces. The rollout will happen over the next month.
Samsung Rally Demands Share of AI Profits
About 30,000 people gathered in Pyeongtaek, South Korea, to demand that Samsung employees get a greater share of profits from the AI boom. Samsung's labor union wants 15% of operating profit from the chip division to be handed to workers. This figure represents more than 40 trillion won, which could mean over $400,000 per worker on average. The rally took place outside Samsung Electronics Co.'s main chip hub to highlight the need for fairer profit sharing.
Sources
- OpenAI Codex Plugins Expand AI Capabilities
- OpenAI Codex: Tailor Your AI Assistant
- Baltimore County councilman condemns A.I. generated image of him
- Baltimore County Councilman Julian Jones speaks out against A.I.-generated picture of him
- The Guardian view on Anthropic’s Claude Mythos: when AI finds every flaw, who controls the internet?
- In the age of AI, the humanities department should consider changing their course structures
- Keeper Security: The Identity Security Platform For Humans, Machines, And AI Agents
- Washington alleges scammers used AI, social media to help steal $9.9M in cryptocurrency
- A grandmother lost everything because a cop trusted AI
- I used Jeff Bezos’ 'Day 1' rule to fix my ChatGPT workflow — and it actually killed my procrastination
- Ulta Beauty Deploys Google-Powered AI Assistant and Agentic Commerce
- Samsung Rally Draws 30,000 to Demand Greater Share of AI Profits
Comments
Please log in to post a comment.