• Everyday AI
  • Posts
  • Can Claude’s AI Agent Simplify Your Work? A Live Test Drive

Can Claude’s AI Agent Simplify Your Work? A Live Test Drive

OpenAI unveils ChatGPT for U.S. Gov., Figure AI launches humanoid safety center, Hugging Face’s plans to challenge DeepSeek and more!

Author

Everyday AI
January 28, 2025

Outsmart The Future

Today in Everyday AI
7 minute read

🎙 Daily Podcast Episode: Today we’re showing you how to use Claude’s agentic AI with its Computer Use feature. Give it a listen.

🕵️‍♂️ Fresh Finds: Sam Altman’s thoughts on DeepSeek, Quartz secretly publishing AI articles and AI invents a new molecule. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: OpenAI unveils ChatGPT for U.S. Gov., Figure AI launches humanoid safety center and Hugging Face’s plans to challenge DeepSeek. For that and more, read on for Byte Sized News.

🚀 AI In 5: Mistral’s AI Agents have flown under the radar. How do they stack up against custom GPTs? We give you the pros and cons. See it here

🧠 Learn & Leveraging AI: We break down what’s new with Anthropic Claude’s Computer Use and how the everyday person can use it. Keep reading for that!

↩️ Don’t miss out: Did you miss our last newsletter? We talked about DeepSeek shocking AI stock market, DeepSeek dethroning ChatGPT in the App Store and DeepSeek releasing a new AI image model, Alibaba's new AI can control PCs. Check it here!

 Can Claude’s AI Agent Simplify Your Work? A Live Test Drive 🧑‍💻️

Wanna learn Claude's new agentic AI? 

Got 30ish minutes? This is the show for you.

In short -- Computer Use is an agentic AI system where you can control a virtual computer just by talking to Claude.

Join the conversation and ask Jordan questions on AI here.

Also on the pod today:

• Overview of Anthropic Claude 🤖
• Critiques of Anthropic's Tools 🤔
• Future of AI Agents 💭

It’ll be worth your 41 minutes:

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Octomind is an AI agent QA tool, Anything World lets you generate any 3D model with AI and Augment UI provides AI-powered UI generation.

OpenAI – Sam Altman has commented on DeepSeek’s recent rise, calling it impressive and great competition.

Although there have been rumblings of a fractured relationship between Microsoft and OpenAI, Altman took to twitter to reassure everyone.

AI Models – DeekSeek and OpenAI seem to answer sensitive questions about China differently.

Big Tech – Former Intel CEO Pat Gelsinger is already using DeepSeek instead of ChatGPT at his startup

AI in Media – Quartz has been quietly publishing AI news articles.

Trending in AI - Steve Cohen is expressing his bullish long-term outlook on AI.

AI in Science - AI has invented a new molecule that would have taken 500 million years to evolve in nature.

Read This – A new Vatican document examines the potential and risks of AI.

1. ChatGPT Gov Rolls Out for U.S. Government Agencies 🇺🇸

OpenAI has unveiled ChatGPT Gov, a version of its chatbot tailored for US government agencies, as reported by Emma Roth. This new tool allows agencies to securely use OpenAI's advanced models like GPT-4o through their own Microsoft Azure cloud instances, enhancing data privacy and security.

The launch aims to improve public services by integrating AI with democratic values, even as President Trump recently reversed AI safeguards introduced by Joe Biden. With over 90,000 users from 3,500 agencies already onboard, ChatGPT is proving pivotal in transforming how government handles sensitive data.

2. Figure AI Launches Safety Center for Humanoids 🦾

Figure AI has unveiled the Center for the Advancement of Humanoid Safety, spearheaded by ex-Amazon Robotics safety engineer Rob Gruendel. This initiative aims to establish industry safety standards as humanoids increasingly work alongside humans in warehouses and potentially homes.

The center plans to regularly release updates on safety testing and improvements, aiming to build trust with customers and regulatory bodies like OSHA.

3. Hugging Face Takes on DeepSeek in AI Showdown ⚔️

Hugging Face is on a mission to replicate DeepSeek's R1 AI model and open-source it for all to see. Their Open-R1 project aims to dismantle DeepSeek's "black box" approach by providing full transparency and accessibility to R1's architecture, which currently lacks open data sets and experiment details.

This initiative could shift the AI landscape by democratizing access to advanced reasoning models, allowing researchers and developers worldwide to innovate without the constraints of proprietary systems.

4. DeepSeek Sparks U.S. AI Concerns 🤔

U.S. officials are scrutinizing the national security risks posed by China's AI app, DeepSeek, with White House press secretary Karoline Leavitt describing it as a "wake-up call" for American AI firms. In an interview on Fox News, Trump's AI and crypto czar David Sacks suggested potential intellectual property theft via AI distillation techniques.

Amidst fears of U.S. tech dominance being undermined, President Trump encouraged American companies to focus on innovation, seeing the situation as a competitive spur. The National Security Council is actively reviewing these developments, highlighting the urgency for the U.S. to fortify its technological leadership

5. California Cracks Down on AI Misbehavior 🧑‍⚖️

California Attorney General Rob Bonta has fired a warning shot at the AI industry, highlighting potential legal pitfalls in its business practices. In a memo issued January 13th, Bonta outlined concerns about AI-driven deception, false advertising, and discriminatory impacts on protected classes.

The advisory urges companies to self-regulate, avoiding actions that could provoke legal consequences under California law.

6. Hugging Face Expands AI Model Hosting with New Inference Providers 🛠️

Hugging Face has teamed up with third-party cloud vendors like SambaNova to introduce Inference Providers, enhancing the flexibility for developers to run AI models using their preferred infrastructure. According to TechCrunch, this collaboration includes partners such as Fal, Replicate, and Together AI, allowing users to effortlessly deploy models like DeepSeek on various servers directly from a Hugging Face project page.

This shift highlights Hugging Face's evolving focus on collaboration and model distribution, moving away from solely using in-house solutions. With this new offering, developers can leverage serverless inference to scale AI models seamlessly, with costs aligning with standard provider API rates for now.

7. Former Google CEO’s Comments on DeepSeek AI 💬

Former Google CEO Eric Schmidt claims that DeepSeek's rise signals a pivotal moment in the global AI race, highlighting China's ability to compete with less. In a Washington Post op-ed, Schmidt urges the U.S. to boost open-source AI efforts and infrastructure investments like Stargate to maintain its edge.

He suggests sharing AI training methodologies to counter DeepSeek's advancements. Schmidt's comments come as his ventures, like White Stork and Holistic AI, stand poised to benefit from increased U.S. investment in AI.

Mistral AI Agents: What they are and how to build one or free

AI startup and LLM maker Mistral has a feature that allows you to make AI agents….for FREE!

There’s some good upsides to why you might want to use Mistral’s AI Agents over custom GPTs and Claude Projects.

We go over the pros and cons.

🦾How You Can Leverage:

Remember that ridiculously bad AI video of Will Smith eating Spaghetti?

Well, that’s what we have today with AI agents.

Technically usable.

Not very good.

(Yet.)

After a bunch of requests stemming from our Claude 3.5 Sonnet general update show last week, we came back for another round today with a more laser focus:

What is it?

It’s Anthropic’s new agentic AI tool that literally performs actions on a virtual computer and it only requires simple text commands. 

Today on our live show, an AI agent in the cloud literally did our work. 

Was it cool? 

Sure. 

Was it buggy? 

Absofrigginlutely. 

Should you still re-watch today’s show and follow along? 

YES!

Why? 

Even though Anthropic’s new ‘Computer Use’ mode is a mere preview into the future of Agnetic AI, orchestrating AI agents is a skillset we all must learn. 

(News flash — Salesforce JUST released its Agentforce agents.) 

Alright, after you’ve caught up with what’s new in Sonnet 3.5, then you’re ready to fully dive in. 

Here’s what ya need to know to make the most of Computer Use today. 

1 – Grab your pencils ✏️

Wanna follow along? 

Today’s show was more tutorial-based, so feel free to re-watch/re-listen as you need to. 

But, here’s the essentials for what you need to launch ‘Computer Use’ and get a very small taste of AI agents. 

Once you go through these simple steps, you’ll have your own AI agent to direct that runs in a virtual environment, and can execute tasks autonomously in a virtual environment. 

Try this: 

A. Claude API key — You can even do this on the free plan, but you’ll need to add a Credit Card and add money to your Claude Console, which is technically separate from a front-end Claude AI chatbot account. 

Once you get the API key, which we showed you at this point in the video, copy and paste it and get it ready. 

B. Download and Install the Docker Desktop program — This is a free program, and one of the recommended ways Anthropic recommends to use Computer Use. You can download it here. (Make sure to grab the right version according to your computer type.) 

C. This Snippet from Anthropic — On Anthropic’s Github repo, they give you this snippet. You’ll paste your API key in the area below. 

(Make sure not to modify anything else, or add/delete spaces.) 

docker run -e ANTHROPIC_API_KEY=$ANTHROPIC_API_KEY -v $HOME/.anthropic:/home/computeruse/.anthropic -p 5900:5900 -p 8501:8501 -p 6080:6080 -p 8080:8080 -it ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest

D. Launch the snippet in Docker’s Terminal 

After installing Docker, launch it and click the small ‘Terminal’ area at the very bottom. Then, you’ll copy/paste the updated snippet above with your API key, hit enter, and hope it works. 

(It’s normal for it to take an extra try or two) 

2 – Get practicing 🏃

Follow the steps above, and you should be off and running at the AI playground. 

In short, chat with Claude on the left hand side of your new local environment and the agentic ‘Computer Use’ tool will execute tasks on your behalf on the right side. 

Give it some simple commands that a human should be able to do on a computer, sit back, and watch it either succeed or fail. 

Also, you’ll really see the value of sharply worded prompting, as Tier 1 rate limits will likely push you into constant ‘Toke Timeout.’ 

For 99% of us, you’ll be dealing with a minute-by-minute token rate, which is pretty easy to hit each time you use it. 

Remember, you’re paying for usage here. For our short little demo, it costs us $1.42.

 

Try this: 

Be patient. No, Anthropic’s Computer Use isn’t ready to do your 9-to-5 just yet. Lolz. 

Buuuuuuuuut— Anthropic shipped. 

Which is a refreshing take, vs the normal marketing and wait lists we often get from other big players in the AI space. 

The downside of shipping an autonomous agent that is still VERY much a work in progress? 

You’ll see after a few run-throughs, that it’s super buggy. 

3 – Perspective is everything 🧠

So why the heck should you use this thing? 

It’s a bit buggy, it’s not a final/polished product, and there’s some tight guardrails in this initial offering. 

In short, Computer Use isn’t super futuristic and shiny. It’s buggy, yet kinda capable. 

But in a few months….

Sometimes, getting reps in now isn’t so you can reap benefits now. 

It’s about being prepared for what’s to come. 

But Anthropic has taken a front seat in the AI agents race. 

Try this: 

After you’ve crossed off #1 and #2 above, you’ve already got your feet wet in breaking new agentic tech. 

Congrats homies! 

Now, get caught up in everything AI agents. 

Cuz this space is moving CRAZY fast. 

Here’s a quick overview:

⌚

Numbers to watch

$24.6 Million

LinkedIn Founder Reid Hoffman’s Manas AI startup has raised $24.6M for AI drug discovery.

Now This …

Let us know your thoughts!

Vote to see live results

Do you attend our livestreams?

Every weekday, we bring you fresh AI insights, exclusive interviews, and breaking news with our Everyday AI livestream.

Login or Subscribe to participate in polls.

Reply

or to participate.