AI’s Next Shift: How DeepSeek R1 Is Redefining the Market

You can now work natively with AI in Google and Microsoft Workspace. Plus: How to use ChatGPT Tasks, President Trump’s investment in AI, and more.

Welcome to Stay Ahead #61 from FlexOS!

Every Wednesday, we share the latest tools and tips to help you work smarter, not harder.

In this edition, we will discuss:

1. News of the Week: DeepSeek R1 is making waves.‍
2. Prompt and Guide:  How to build a weekly content ideas bank with ChatGPT Tasks in 5 steps.‍
3. Must-read News:  DeepSeek R1 and OpenAI Operator reviews. 

Let’s dive in!

This ChatGPT’s rival is making waves

On Monday, major tech stocks—including Nvidia, Microsoft, and Meta—experienced significant declines, and it was due to the entrance of a new player - DeepSeek R1. Chinese startup DeepSeek has unveiled an advanced AI model that’s not only rivaling existing giants but also redefining the industry’s dynamics.

Here are why AI giants are concerned about DeepSeek R1

• Reasoning and Problem-Solving: DeepSeek R1 demonstrates strong capabilities in complex reasoning tasks, including mathematical problem-solving and natural language inference. It achieves a 97% accuracy rate in mathematics and surpasses 96% of humans in programming tests. 
• Benchmark Performance: In evaluations like the American Invitational Mathematics Examination (AIME) and MATH benchmarks, DeepSeek R1 matches or exceeds the performance of OpenAI’s o1 model.  
• Cost-Effectiveness: At this capacity, DeepSeek R1 is currently free (compared to the $200 subscription of ChatGPT Pro). The R1 API is 27 times more affordable than ChatGPT’s o1 API.

Should you try DeepSeek R1? This question has sparked a lively discussion in our Lead with AI community. Here’s what’s standing out about DeepSeek R1:

• It’s particularly strong for tasks that require precision and reasoning.
• Users as beneficiaries: Being open source and low-cost makes AI more available to people and smaller organizations, creating healthy competition and benefits for users.
• Censorship and Data Privacy: Like all AI models, it has safety measures to avoid discussing sensitive topics, but the ​censorship​ on DeepSeek R1 is particularly notable, especially regarding history and geopolitics. For large US companies, DeepSeek might be a hard “no”. 
• The Nvidia’s drop is temporary: The DeepSeek model is comparable to ChatGPT o1 but comes at a fraction of the price. Imagine its potential with more resources, driving greater demand for intelligence rather than lowering the need for compute power.‍
• Other AI giants are catching up: While they have already known about reinforcement learning internally, the launch of DeepSeek AI could accelerate the pace at which their new models are developed and improved.

What’s more interesting to me is how DeepSeek R1’s emergence is keeping OpenAI in the “free AI” game for longer than they might have expected. In response to this growing competition, Sam Altman recently announced that OpenAI’s new o3-mini model will be available for free to all users.

It’s exciting to explore new AI tools like DeepSeek R1, but it’s important to be mindful of the data you share and critically assess the information you receive. Whether using DeepSeek, OpenAI, or others, understanding how your data is used and ensuring the accuracy of outputs should always be a priority.

OpenAI’s first AI Agent is here

Turns out, Sam kept his words when he said 2025 would be a big year for AI agents. OpenAI is starting the year strong with the launch of Operator, the company’s first agent product that can surf the web and complete tasks for you.

By now, this concept is no longer mind-blowing. We’ve already seen similar capabilities with Claude’s Computer Use and Microsoft’s Orchestrator. However, OpenAI’s participation might signal a year of mass adoption, turning AI agents—having AI do the tasks for you instead of just telling you how—into the new norm.

Here are what you need to know about Operator:

• Autonomous Web Interaction: Operator is an AI agent that can use its own browser to perform tasks on the web. It interacts with webpages by typing, clicking, and scrolling.
• Task Automation: Operator can handle repetitive browser tasks such as filling out forms, ordering groceries, and creating memes, and automating tasks like booking travel, making restaurant reservations, and shopping online.
• Self-Correction and User Collaboration: If Operator encounters challenges or makes mistakes, it can self-correct using its reasoning capabilities. It will hand control back to the user when it gets stuck or needs assistance, ensuring a collaborative experience.
• Personalized Workflows: Users can personalize their workflows by adding custom instructions for specific sites or for all sites. They can save prompts for quick access for repeated tasks.
• Multiple Task Management: Operator can run multiple tasks simultaneously by creating new conversations, similar to using multiple tabs on a browser.
• Computer-Using Agent (CUA) Model: Operator is powered by a new model called CUA, which combines GPT-4o's vision capabilities with advanced reasoning through reinforcement learning. The CUA model is trained to interact with graphical user interfaces (GUIs) like buttons, menus, and text fields.
• Safety measurement: Takeover mode for sensitive information (e.g. log-in credentials), user confirmation, task limitations (not accepting bank transactions).

Operator is currently available as a research preview for pro users in the US. Pro users in other countries will also soon have access to this, but for users in Europe, “it might take a while.”

OpenAI’s Operator marks a significant step towards a future where AI doesn’t just assist—it acts. While we’ve seen similar capabilities before, the power of OpenAI’s ecosystem and its massive user base might finally push AI agents into mainstream adoption. But are you ready for it?

Prompts and Tutorials

In 5 Steps: Build a Content Ideas Bank with ChatGPT Tasks

Keeping up with fresh, relevant content ideas for your blog and social media can be challenging—especially when you’re juggling leadership responsibilities. With ChatGPT’s Tasks feature, you can automate content brainstorming and ensure a steady flow of high-quality ideas every week. Follow these five simple steps to set it up and make it work for you.

‍→ Read more news at the end.

Need more AI Tools? Check out

The best AI tools for Recruiting, Marketing, Accounting, and HR.

​10 Tools Anyone Can Use (Writing, Powerpoints)

​10 AI Productivity Tools (Meeting Notetaking, Writing, Brainstorming.)

Our full AI For Work Top 100

Create better prompts with our GPT Prompt Generator

Every week, we pick must-read articles about the future of work so you can stay ahead.

‍Biggest Market Loss In History: Nvidia Stock Sheds Nearly $600 Billion As DeepSeek Shakes AI Darling

Nvidia’s stock plummeted nearly 17%, erasing approximately $600 billion in market value, as investors reacted to DeepSeek’s new AI model, which operates efficiently without relying on Nvidia’s high-end chips.

I Did 5 DeepSeek-R1 Experiments | Better Than OpenAI o1?

Explore the performance of DeepSeek-R1 through five experiments, comparing its capabilities to OpenAI’s models.

I Paid 200 $ for the first ChatGPT Agent. Does It Actually Work? (OpenAI Operator)

The AI Advantage shows how ChatGPT Operator can independently perform actions such as booking flights, responding to emails, and updating schedules.

In the Age of AI, Human is the New Luxury

In a world increasingly dominated by automation and AI, human connection and empathy are emerging as the ultimate premium values.

How to Make Your Team’s Work More Visible‍

Actionable strategies for leaders to increase the transparency of their team’s efforts, boosting accountability and engagement.