Another Crazy Day in AI: ChatGPT Now Does the Heavy Lifting
- Wowza Team
- Jul 17
- 4 min read
Hello, AI Enthusiasts.
The inbox is (mostly) quiet, and the weekend is within reach. But one update’s too big to miss.
OpenAI just turned ChatGPT into an agent that works for you. It can handle real tasks solo: open your GitHub, fill your spreadsheet, check your Gmail. It's like giving ChatGPT a desk and a mouse.
While OpenAI is building agents, Meta is busy building its Avengers, starting with ex-Apple minds, and doubling down on the whole “superintelligence” thing.
Meanwhile, at a summit in Pittsburgh, the President is betting $90B that AI and steel can shape America’s next industrial boom, backed by energy firms and tech CEOs.
The bots may not sleep, but you should. See you next scroll.
Here's another crazy day in AI:
ChatGPT Agent can now do the work for you
Meta adds two former Apple scientists to superintelligence push
Trump pushes $90 billion AI and Energy initiative at Pittsburgh Summit
Some AI tools to try out
TODAY'S FEATURED ITEM: Agent Mode: From Chat to Getting Things Done
Image Credit: Wowza (created with Ideogram)
What happens when your chatbot gets its own computer and starts doing your job for you?
OpenAI has introduced a major update to ChatGPT: a new agentic system that enables the model to complete complex tasks on your behalf using its own virtual computer. This new capability, called agent mode, allows ChatGPT to go beyond conversation, interacting with websites, analyzing data, creating editable documents, and pulling relevant information from connected apps like Gmail or GitHub. It’s the result of combining earlier tools like Operator and deep research into one cohesive experience, making ChatGPT more useful in both everyday life and professional work.
The update is now available to Pro, Plus, and Team users and marks a significant step in making large language models more action-oriented. Rather than relying solely on user-driven prompting, the agent can choose how to complete a task by selecting from a set of built-in tools like a web browser, code terminal, spreadsheet editor, and API access, and intelligently navigating those tools based on the task at hand.
How It Actually Works
End-to-end task completion - Request something like "analyze our top three competitors and create a presentation" or "plan a dinner party for eight people and order the ingredients" and ChatGPT manages the entire process from research to final delivery
Adaptive approach selection - The system chooses the most effective method for each step, switching between visual web browsing, text analysis, code execution, or API calls as needed
Continuous user control - You can step in at any point to redirect the process, take over the browser yourself, or halt the task, with ChatGPT requesting permission before taking consequential actions
Documented performance results - Testing demonstrated 41.6% accuracy on Humanity's Last Exam and 27.4% on FrontierMath, with results often comparable to human performance on complex professional tasks
Direct app connectivity - Integrates with existing ChatGPT connectors and can securely authenticate to websites when access to personal information is required
Multi-layered safety protocols - OpenAI has implemented defenses against prompt injection attacks and requires explicit user approval for actions like purchases or sending emails
The practical implications of this development are worth examining carefully. We're looking at a system that can handle the kind of multi-step, cross-platform work that typically requires significant time and attention from knowledge workers. Many professionals spend substantial portions of their day gathering information from various sources, synthesizing it, and then creating deliverables. Having a system that can autonomously manage these workflows represents a meaningful change in how digital work might be approached, though it also brings questions about oversight, quality assurance, and the evolving nature of professional roles.
What makes this particularly noteworthy is the breadth of its deployment. Unlike previous AI agents that were confined to research settings or limited user groups, this is rolling out to millions of existing ChatGPT users. The safety measures OpenAI has implemented suggest they recognize the responsibility that comes with releasing a system capable of taking real-world actions. The performance benchmarks indicate genuine capability, but the real evaluation will be how well it handles the unpredictable, often messy nature of actual work environments. As users begin to experiment with these capabilities, we'll likely see both creative applications and unforeseen challenges emerge, providing valuable insights into how agentic AI systems perform when deployed at scale in real-world contexts.
Read the full article here.
OTHER INTERESTING AI HIGHLIGHTS:
Meta Adds Two Former Apple Scientists to Superintelligence Push
/Reuters
Meta has hired two Apple AI researchers, Mark Lee and Tom Gunter, for its new Superintelligence Labs team, signaling an aggressive push to attract top talent in artificial intelligence. Their move follows the recent defection of Ruoming Pang, who led Apple’s Foundation Models team, reportedly for a multi-million-dollar package. Meta CEO Mark Zuckerberg is personally driving the company’s AI expansion, pledging massive investment in data centers and talent. The hires reflect a broader trend of tech giants racing to build teams capable of developing AI systems that may surpass human intelligence.
Read more here.
Trump Pushes $90 Billion AI and Energy Initiative at Pittsburgh Summit
/Jessica Guay, Andy Sheehan, Jessica Riley, Michael Guise, on CBS Pittsburgh
At a summit in Pittsburgh, President Trump unveiled over $90 billion in AI and energy investments for Pennsylvania, backed by 20 leading tech and energy firms. The initiative includes building data centers, training workers for AI and energy jobs, and positioning Pittsburgh as a national hub for artificial intelligence. Trump called it the largest investment package in the state's history, highlighting partnerships with companies like Westinghouse and Nippon Steel. Despite protests, the event brought together major political leaders and tech CEOs to outline a vision for America’s AI-powered industrial future.
Read more here.
SOME AI TOOLS TO TRY OUT:
That’s a wrap on today’s Almost Daily craziness.
Catch us almost every day—almost! 😉
EXCITING NEWS:
The Another Crazy Day in AI newsletter is on LinkedIn!!!
Leveraging AI for Enhanced Content: As part of our commitment to exploring new technologies, we used AI to help curate and refine our newsletters. This enriches our content and keeps us at the forefront of digital innovation, ensuring you stay informed with the latest trends and developments.
Comments