The best ways to use Deepseek AI 🐳
Learn where the AI model everyone talks about excels for Product work.
Thank you for reading The Product Courier.
We’re now 2,400, which is 155 more than last month! 🙏🏻
If you're new here, welcome!
I’m Toni, I help B2B companies build scalable growth engines to generate more revenue.
👉🏻 Let’s connect on Linkedin.
With Lucas, we share actionable tips to help you deliver better work and accelerate your career by leveraging A.I. and other tools.
In this 41st edition, I want to share with you:
Deepseek is taking on the world 🐳
How you can make the best of Deepseek’s R1. 🤓
Happy reading!
News of the week. 📰
The Chinese whale is crushing Tech giants 🐳
If you live in a cave, DeepSeek, the Chinese AI startup we talked about 3 weeks ago, has dropped a bombshell with its R1 model, shaking up the whole AI ecosystem. Here is why:
DeepSeek released R1, its deep reasoning model, outperforming models like OpenAI’s o1 or Gemini 2.0. But here’s the kicker:
Cost-Efficiency: DeepSeek spent under $6M on computing resources to train its model - peanuts compared to the 100s of millions spent by OpenAI and others.
Pricing: The web version and app offer free access to R1, and Deepseek’s API is 30 times cheaper than OpenAI’s.
Public adoption: DeepSeek became the most downloaded in the U.S. within a week of launch.
Accessibility: the model is Open Source.
Market Impact: Nvidia shares plunged 17%, wiping out $600 billion in market cap. Nasdaq took a $1 trillion hit, with ripple effects across Microsoft, Meta, and others.
Privacy & Training: Deepseek also raised concerns because of its Terms of Use and Privacy concerns as we had already mentioned.
And it engaged very political stakes between China and the USA.US Tech giants also suspect that they used OpenAI models to train their data.
Why It Matters:
AI innovation can be cost-efficient.
DeepSeek proves that cutting-edge AI doesn’t have to cost billions.
Companies no longer need an "OpenAI-sized" budget to compete.
You don’t necessarily need 10,000 GPUs to build a powerful model, smaller teams can innovate like tech giants.It reinforces competition.
DeepSeek's rise signals that global competition is heating up and will drive costs down.
Their efficiency and performance also challenge the narrative that U.S. tech firms have an unshakable edge.AI adoption is speeding up.
With more efficient and cost-effective training, “AI models” will become commodities to streamline workflows, improve customer experiences, and scale operations efficiently.
Early adopters are gaining an advantage and those who delay risk falling behind faster-moving competitors.
What’s in it for you and Product Teams:
Increased AI development opportunities
More affordable, advanced AI tools will let product teams deliver value faster and on a leaner budget.
Less reliance on one supplier.
Most professionals use one single AI.
With more players entering the space, you could implement AI features with fallback systems to avoid over-relying on one unique supplier (i.e. when ChatGPT breaks for a few hours) and ensure the availability of your AI apps and workflows.Accelerated learning at a marginal cost:
You can now test multimodal and reasoning capabilities for free and invent use cases to work smarter ( and apply
’s weekly tips! )
DeepSeek’s sudden fame is a reminder that innovation doesn’t always come from where you expect it, and success is unpredictable.
Is this “AI’s Sputnik moment” Marc Andreessen mentions? WDYT?
You don’t want to be left behind, and need help to train your Product team to use AI?
Check out our AI training offer for Product teams.
Actionable modules and practical exercises tailored for Product teams’ tasks.
Leverage AI and see instant results to increase productivity.
Book a call to define your tailored training plan and empower your team to master AI.
Our Community Poll. 📊
Your weekly tutorial. 🧰
The best way to use Deepseek AI 🐳
“The almost cosmic nature of the whale is beyond human understanding."
Herman Melville
✅ Traditional AI models (ChatGPT, Claude) still cannot combine advanced reasoning and web search in one query.
✅ Deepseek R1 combines deep thinking with real-time browsing, making research faster and more comprehensive.
✅ It’s useful to prioritize features, refine pricing strategies, and craft data-driven strategies with up-to-date data.
✅ Deepseek is free online and it’s as powerful as ChatGPT or Claude.
You can use Deepseek V3 as a day-to-day assistant for most tasks too.
DeepSeek R1 is the talk of the town.
But how does it compare to ChatGPT, Claude, or Perplexity?
One of the main benefits of Deepseek is its pricing:
Free on the web (including R1, the deep thinking model - similar to OpenAI o1)
30x cheaper than ChatGPT for the API.
It ranks as high as ChatGPT or Claude for most tasks.
But it has a feature that makes it perfect for specific use cases.
AI models & real-time search
Like Deepseek R1, other leading AI Players provide “advanced reasoning” models (OpenAI o1, Gemini Deep Research).
These models can handle complex problems as they break down queries and proceed with a Chain of Thought approach to offer better and more thoughtful outputs.
For example, o1 will go deeper into a competitive analysis and identify gaps or patterns more accurately, than ChatGPT 4o.
However, most AI models have (for now) a fixed knowledge cut-off date:
For example, if you run market research on Agentic AI in January 2025, you won’t get the full picture.
Now, ChatGPT 4o lets you run a web search to gather recent information from the web, and Perplexity gathers the latest information easily from numerous sources.
But it wasn’t possible to couple Advanced reasoning with online search.
Until now.
Deepseek R1: The AI that can think & browse
This is Deepseek’s key differentiator:
➡️ You can combine a web search + advanced reasoning in a single query.
This means you get deep analysis with the most recent, up-to-date information.
Let’s see how it performs compared to ChatGPT-4o + Web Search. 👇
ChatGPT vs DeepSeek
For our examples, we’ll act as a Lead PM for Buffer (social content tool).
We’ll detail:
feature prioritization
pricing strategy
objection handling
1) Feature prioritization
We’ll start with a review of a product backlog prioritization:
Prompt:
Apply the RICE framework (Reach, Impact, Confidence, Effort) to prioritize the following backlog features for Buffer.com’s social media management platform.
For each feature, validate its market impact by analyzing the latest 2024 user reviews from Capterra and G2, competitor benchmarks, and emerging social media trends.
Provide a detailed scoring breakdown and actionable recommendations for prioritization.
Backlog Features:
Tag or mention profiles in a LinkedIn post
Bulk schedule posts
Dark mode
Publish to Bluesky
Include insights on how each feature aligns with Buffer’s core user base (SMBs, agencies, and solo marketers) and its potential to drive retention, engagement, or competitive differentiation.”
Results 📝
👉🏻 For the full output, check the link below.
Deepseek reviewed 29 sources.
It detailed and calculated RICE scores for each feature
It provided a detailed analysis, linking its recommendations to the sources.
ChatGPT 4o reviewed 19 sources.
It gave an output much faster than Deepseek.
It provided vague RICE scores for each feature (i.e. Low Reach, Medium Impact, Medium Confidence, Medium Effort.)
The recommendations were generic and not detailed.
ChatGPT 4o’s conclusion is pretty generic and not explained clearly.
