Welcome to 9to5Neural. AI is rapidly evolving, and we help you stay informed. In this first edition, we delve into OpenAI’s new direction, Anthropic’s innovative approach to common AI critiques, and the back-and-forth over presidential AI executive orders. Let’s unpack the latest in AI news.
ChatGPT Engages with Operator
OpenAI has unveiled the 18K gold Apple Watch Edition of ChatGPT. ChatGPT Pro is a $200/month subscription that certainly makes Tim Cook wish Apple had such a revenue stream per user.
Starting now, ChatGPT Pro also provides AI fans with a compelling new incentive to subscribe, aside from enhanced request limits.
Introducing Operator. OpenAI describes it as “a research preview of an agent that can browse the internet to accomplish tasks on your behalf.” From crafting memes to grocery shopping and completing forms, Operator is among the first agents designed by OpenAI to action user tasks.
We are excited to introduce Operator, an agent that autonomously navigates the web to fulfill your requests. It can access webpages and interact with them through typing, clicking, and scrolling. This is currently in a research preview phase, indicating it has limitations but will expand through user feedback.
Operator won’t always come with a $200/month fee. OpenAI has plans to make this AI tool accessible to Plus, Team, and Enterprise subscribers in the future. For now, Operator is available to all ChatGPT Pro users in the U.S. at operator.chatgpt.com.
OpenAI claims Operator is powered by its innovative Computer-Using Agent (CUA) technology.
At the core of Operator lies Computer-Using Agent (CUA), a model that synergizes GPT-4o’s vision capabilities with advanced reasoning thanks to reinforcement learning. CUA is adept at engaging with graphical user interfaces (GUIs)—the buttons, menus, and text fields humans utilize on screens—allowing it to perform digital tasks efficiently without relying on OS- or web-specific APIs. […]
Though CUA is still evolving and has certain limitations, it has achieved impressive benchmark results, with a 38.1% success rate on OSWorld for full computer tasks, 58.1% on WebArena, and 87% on WebVoyager for web-based operations. These outcomes showcase CUA’s proficiency in navigating and functioning across various platforms using a unified action framework.
This might be a good moment to mention that I am resigning from 9to5Neural to dedicate more time to my family.
All forthcoming editions of 9to5Neural will be powered by Operator. I trust the Computer-Using Agent to effectively relay AI news to people in the future.
But wait, I may have been premature in that announcement. Apparently, there’s a problem with our ChatGPT Pro subscription. I’m back at it!
In all seriousness, Operator represents a significant step forward. We might look back at January 2025 as a pivotal moment for AI development. The Computer-Using Agent technology could also address concerns from AI skeptics pondering when ChatGPT-5 will emerge.
Another significant story from OpenAI this week? Stargate. Or, as Sam Altman quipped on X, “big. beautiful. buildings.”
What’s Stargate? Essentially, a colossal computing brain based in Texas. OpenAI elaborated on the initiative this week:
The Stargate Project is a new initiative aiming to invest $500 billion over the next four years in building advanced AI infrastructure for OpenAI within the United States. We will initiate the deployment of $100 billion immediately. This infrastructure is designed to solidify American leadership in AI, generate hundreds of thousands of jobs, and provide significant economic benefits globally. This endeavor will not only facilitate the re-industrialization of the U.S. but also enhance national security for America and its allies.
The initial equity contributors for Stargate include SoftBank, OpenAI, Oracle, and MGX. SoftBank and OpenAI are the primary collaborators, with SoftBank handling financial obligations and OpenAI managing operations. Masayoshi Son will act as chairman.
Essential technology partners in this initiative include Arm, Microsoft, NVIDIA, Oracle, and OpenAI. Construction is presently underway in Texas, with potential sites being evaluated nationwide for additional campuses as we finalize agreements.
Through Stargate, Oracle, NVIDIA, and OpenAI will closely collaborate to design and operate this computing system.
Of course, behind every ambitious AI company lies an equally ambitious billionaire, and discussions about Stargate’s funding have ignited rivalry on X.
Elon Musk, whose xAI venture isn’t involved in Stargate, reacted to the announcement on X, stating “they don’t actually have the funds.” Musk claimed to have reliable information indicating that SoftBank has “less than $10B secured.”
In contrast, Altman expresses confidence that the involved parties have secured necessary funding.
Meanwhile, the OpenAI leader noted that he got caught in a trap regarding Trump (now that Trump has made his character available to play, referencing Stargate).
Truthfully, I have greater optimism regarding the ChatGPT Operator’s future than the billionaire dynamics in play.
Claude Introduces Citation Features
In parallel, Anthropic, known for its cautious approach to AI safety, is launching an exciting new feature for its Claude chatbot called Citations.
Today marks the launch of Citations, a new API capability enabling Claude to ground its responses with source documents. Claude can now reference precise sentences and sections utilized to formulate its replies, resulting in more verifiable and reliable outputs. […]
Previously, developers needed complex prompts to instruct Claude to incorporate source information, often leading to inconsistent results and extensive time spent on prompt engineering and testing. With Citations, users can add source documents into the context window, and during queries, Claude automatically cites claims in outputs based on those sources.
Our internal evaluations indicate that Claude’s built-in citation capabilities outperform most custom implementations, enhancing recall accuracy by up to 15%.
Anthropic identifies key use cases, including customer service inquiries and document summarization.
Best perspective? Kyle B. Russel on X, no citations required:
Claude 3.5 Sonnet and Claude 3.5 Haiku are available for Citations starting today, and Anthropic has provided documentation for exploration.
New AI EO Surpasses Previous AI EO
After a brief intermission from presidential themes, let’s return to the U.S.’s policy on AI.
President Trump resumed his rapid-fire executive order signing on Thursday, revoking the Biden administration’s AI policy executive order and instituting his own.
In case you need a reminder, Biden’s executive order addressed AI safety, infrastructure standards, job mitigation strategies, and watermarking AI-generated content for transparency. In essence, Biden’s order:
- Stressed the need for the secure, responsible, and trustworthy evolution of artificial intelligence (AI).
- Required standards for crucial infrastructure, security upgrades, and oversight of federally supported projects.
- Focused on societal issues, including job disruption concerns, promoting equity, and safeguarding civil rights.
- Mandated transparency in AI-generated content through watermarking to differentiate it from human-made material.
According to AP reports, Trump’s AI executive order eliminates existing governmental policies perceived as hindrances to U.S. AI innovation, emphasizing that the U.S. must “develop AI systems that are unbiased and free from socially engineered agendas,” as outlined in the order.
Beyond broad policy directives, President Trump’s AI EO allows for the “development of an AI action plan within 180 days,” per the AP, led by David Sacks, Special Advisor for AI and Crypto, previously an executive at PayPal appointed by Trump.
Moving forward, technology firms will no longer be obligated to inform the government about their AI model developments that exceed a specific capability threshold.
Intense Competition from DeepSeek R1
Meanwhile, the race in AI advancements isn’t limited to U.S. companies. This week, Chinese AI company DeepSeek launched its R1 model family.
What makes R1 noteworthy is its capability to function locally, delivering performance comparable to OpenAI’s ChatGPT 4o model. Typically, local models lag behind those operating remotely, rendering this development and DeepSeek a company to monitor closely.
The caveat? R1 inherently operates from a government-sanctioned perspective on historical events like the 1989 Tiananmen Square protests and Taiwan’s independence. Just in case the stakes in the ongoing AI competition weren’t clear enough.
Stay tuned for the next edition of 9to5Neural for more updates on AI advancements—exclusively on DMN!
Top iPhone Accessories
FTC: We utilize income-generating auto affiliate links. More.
