Beyond the DeepSeek Hype—What It Really Means for You
The AI world just had a moment.
If you followed the DeepSeek saga last week, you know what I’m talking about.
Wall Street panicked.
Tech influencers flooded your feed with hot takes.
Trump called it a wake-up call for the US tech industry.
All because of DeepSeek R1—the most talked-about AI release since ChatGPT.
Why?
Because DeepSeek did what seemed impossible.
They built a world-class reasoning model for just $6 million.
For context, OpenAI spent hundreds of millions to achieve the same performance.
Now, the headlines scream:
- Is China overtaking the US in AI?
- Is NVIDIA overpriced?
- What does this mean for the future of AI?
But here’s the truth:
None of that matters.
The debates, the drama—it’s all noise.
Nobody truly knows where AI is heading. Not OpenAI, not DeepSeek, not even the experts making predictions.
What does matter is this:
- You now have access to a free, powerful reasoning model with internet search. (Just weeks ago, this was ridiculously expensive.)
- You can use it via API—at a fraction of OpenAI’s cost.
- It’s fully open-source, meaning anyone can fine-tune it or push its capabilities further.
Ultimately, DeepSeek has dramatically accelerated the "AI Revolution" by making artificial intelligence more accessible and providing better building blocks for future innovation.
If you’re learning how to use AI, this is great news.
If you’re still on the sidelines, this is your wake-up call.
The cost of intelligence is dropping fast.
You need to start building on this new “intelligence baseline”—we have no choice.
What Happened Last Week
Before we dive deeper and explore what this means for all of us, let's review the key facts about this saga to ensure everyone's on the same page.
Here's what happened:
On January 20, 2025, DeepSeek (a Chinese AI company) announced and released DeepSeek-R1, their first reasoning model (R1 for short; try it here using the "deepthink" option). R1 is comparable to OpenAI o1 (OpenAI's top-tier premium model), which launched on December 5, 2024.
DeepSeek R1 is free (compared to $200 for unlimited OpenAI o1 access). R1 includes internet search capabilities—a feature not yet available in the more expensive OpenAI o1. R1's cost per request is approximately 1/27th of o1's cost, which sparked intense reactions last week, particularly on Wall Street.
Performance-wise, R1 and o1 are nearly identical across evaluations, with the only notable difference being a modest 4-percentage point gap on GPQA Diamond (71.5% vs 75.7%). For context, here's how o1 compares to GPT-4o (OpenAI's best non-reasoning free model) on three benchmarks, with Anthropic's Claude 3.5 and Google DeepMind's Gemini 2.0 performing similarly to GPT-4o:
For those interested in the technical details, DeepSeek's research paper is a tremendous source of information (you can simply upload it to NotebookLM and listen to it like a podcast episode for free).
This letter isn't meant to be another analysis of last week's events or a detailed comparison of different models. You can find plenty of those on the internet.
Instead, let's focus on what really matters: how this affects YOU. I'll show you how to leverage this breakthrough to your advantage and make the most of it—rather than getting lost in the news cycle drama.
If you've read my past letters, you know I've highlighted tasks that are particularly suited for reasoning models, which provide excellent results (far better than the free ChatGPT 4o models). Because these models were so expensive ($20 for dozens of requests a day or $200 for unlimited access), we usually had to find workarounds or alternatives to get comparable results without breaking the bank.
We even had to carefully ration our usage of these models in our work routines.
Now, having such a powerful model available for free completely changes the game. In this letter, I'll show you several use cases where reasoning models excel—tasks that were prohibitively expensive just days ago but are now freely available to you, without limits.
Access to DeepSeek R1 will enhance every single task where you previously used AI. Better model means better outputs.
But here are the key areas where, I believe, this shift will drastically change the game:
1. Programming For Anyone
This isn't just for programmers.
DeepSeek R1 ranks in the 96th percentile for competitive programming problems. It can solve complex programming challenges that would test even the most skilled developers!
If you read my letter about "programming with words" a few weeks ago, you'll recall that coding without any programming knowledge was only possible using OpenAI O1's capabilities—which was prohibitively expensive at that time (just two weeks ago!).
Now these capabilities are free!
Anyone, anywhere, regardless of technical knowledge, can now develop professional-grade apps without writing code. Simply open DeepSeek R1 or connect it to a tool like Cursor (see my letter), and you can build applications just by describing what you want in plain English—even debugging is as simple as copy/past the errors you encounter.
This is a powerful advantage available to anyone with internet access. Remember: if you don't seize this opportunity now, catching up later might be much more challenging.
2. Complex Problem Solving
Remember this letter?
Let's say you're a business (or farm) owner who wants to optimize your operations—whether it's spending on leads, energy, supplies, or anything else.
In most cases, you already have access to data, which isn't the challenging part nowadays. What's significant is that DeepSeek R1 ranks in the 97th percentile for complex math and problem-solving—equivalent to having an excellent, well-trained engineer at your disposal, 24/7, never tired, and for free.
In this letter, we demonstrated this using a simple example with free ChatGPT (back when o1, which has capabilities equivalent to DeepSeek R1, was expensive) to optimize an agency's commercial strategy. You simply explain your problem to the AI, and it generates a very powerful and brilliant optimization solution, writes the necessary code, and either executes it or guides you through the execution—all within minutes.
Just a few years ago, this would have seemed magical. Now you don't need to hire staff, pay expensive freelancers, or spend time explaining requirements and coordinating tasks. This is truly a game-changer (if you know how to use it).
3. Education
Education is another big shift.
Like AI coding, this isn't entirely new—existing models were already capable teachers. However, reasoning models take education to another level. The content is richer, the insights are sharper and more nuanced, and most importantly, these models excel at teaching even highly advanced topics.
Anyone, anywhere, can now learn advanced skills with just an internet connection. Never before in history has knowledge been so easily accessible at this scale worldwide.
I wrote a whole letter on this subject in my last letter. The implications are clear: Universities are evolving into certification and networking hubs for those seeking corporate careers, while also providing environments for cutting-edge research (like PhDs). Traditional "basic learning" is becoming less central to their mission.
DeepSeek R1's ability to reason through problems makes it an ideal tutor for subjects like math, science, and programming.
4. Legal Help (Without The Lawyer Fees)
Before DeepSeek R1, I relied on O1 for legal questions.
Reasoning models are tremendously helpful for addressing questions about taxes, legal matters for my company, reviewing and adapting contracts, and generating or refining terms of use.
Previously, accessing such curated, cross-checked information to resolve legal roadblocks was expensive.
DeepSeek R1 (and O1) excels at reviewing contracts and ensuring numerical consistency across long documents—something GPT-4o sometimes misses.
Now, DeepSeek R1 has made this level of intelligence accessible to everyone, everywhere, for free (which is incredible).
Let me be clear: I'm not saying you should trust everything AI tells you about legal matters—that could be costly.
However, it's undeniable that many legal and complex questions that used to halt operations can now be addressed to maintain momentum. What makes DeepSeek R1 even more powerful is its ability to search the internet simultaneously and provide original references for deeper investigation.
As an entrepreneur (navigating the complexities of French administration), this has changed my life.
5. Turning Vague Ideas Into Concrete Plans
Here’s a real-world example:
A fertility clinic wants to digitize patient care with AI.
During a call, I jot down rough notes:
- Start with a questionnaire to assess patients
- Track their treatment journey in an app
- Provide automated next steps
I feed this into DeepSeek R1:
Prompt:
“Turn this into a detailed product description for a software developer.[put here my notes]”
In seconds, I get:
- A full product spec
- UX/UI design recommendations
- A proposed tech stack
- Compliance & security notes
Before the meeting is over, I already have a draft proposal and a prototype.
Here’s how you can create a quick prototype:
Go to Claude for free (because Claude can develop and show the UI faster), and just write this prompt:
Prompt:
"Develop an app based on these UX/UI design specifications:
[paste UX/UI design specifications here from R1]"
Within minutes, I generated this mockup based on the client's description:
While not perfect, it serves as an excellent starting point for discussions (created in 1 minute).
This is just a basic approach. For a more sophisticated implementation, you can use the specifications with an AI-powered IDE like Cursor (cf. this previous post).
6. Diagram Generation in Seconds
Want to visualize a complex process?
Just describe it to DeepSeek R1.
Example:
“Generate a flowchart for how a hotel booking system works.”
It spits out Mermaid syntax, which I paste into Excalidraw—and instantly, I have a clear, structured diagram.
No need for expensive design tools.
No need to struggle with clunky software.
If you think in visuals, this is a game-changer.
- Step1: Ask R1 to generate Mermaid syntax
- Step2: Past the syntax in Excalidraw
- Step3: Export your diagram or adjust before exporting manually if you want
Cool Stuff People Are Already Doing with Reasoning Models
I've shown you some of my favorite ways to use these reasoning models, but that's just scratching the surface! You wouldn't believe some of the amazing things people are creating with tools like O1 and DeepSeek R1.
Let me share some cool projects I've come across that might spark your creativity:
The Bottom Line
DeepSeek R1 just leveled the playing field.
The best AI tools are no longer locked behind expensive paywalls.
They are now free, open-source, and accessible to anyone.
That means two things:
- Opportunities just exploded. You can build, automate, learn, and optimize like never before.
- Excuses just disappeared. If you’re not using AI to improve your life, you’re falling behind.
This is a major shift.
The question is: Are you changing with it?
See you next week.
— Charafeddine