- The AI Product Report
- Posts
- DeepSeek Diving š³ā + The News You Missed
DeepSeek Diving š³ā + The News You Missed
Your AI Product - News Report - February 4th

For those that are new around here, this is an extension of my weekly newsletter where I highlight new and innovative AI products that are worth exploring.
Hey hey!
Happy not-Wednesday, hereās your news in newsletter!
Due to some technical errors This was supposed to land Friday, Jan 31st, and is now in your inboxes Feb 4th š
In this Issue:
The DeepSeek Splash š
Your Weekly Industry News Recap
The Raise Recap
What Iām Reading

The DeepSeek Splash
WHAT IS IT?
Coming out of stealth mode comes a new LLM Challenger: DeepSeek.
Boasting bold benchmark performance, this model was allegedly developed for a fraction of the cost of OpenAIās models on a per iteration basis.

This UI feel familiar? š Screenshot of DeepSeek, 2025
WHY IS ANOTHER LLM PLAYER SUCH A SHAKEUP?
A new entrant swings hard right out of the gates: New entrants arenāt something new to the AI scene - especially not for LLMs. But one that comes in this late in the game swinging this hard is refreshing, and frankly unexpected!
The Potential to Shrink Operating Costs vs competitors: Startups and those experimenting with LLMs in their environments are now looking at lower op costs due to DeepSeekās architecture barely requiring flashy GPUs. If this is flexible enough as a model, this can deliver a lot to small teams. On the other hand, can it do anything that the market leaders canāt? Not really. Does it make its major components open source Ć -la LLAMA? Yes it does. See for yourself: https://huggingface.co/models
DeepSeek Proves that Reinforcement Learning has a place in LLMs: that Itās the biggest proof that large models can improve their performance dramatically from pure reinforcement learning. Venture Beat covers the basics nicely of what is and why itās impressive. https://venturebeat.com/ai/deepseek-r1s-bold-bet-on-reinforcement-learning-how-it-outpaced-openai-at-3-of-the-cost/ Mind you that even when GPT4 released, OpenAI had tried adding a layer of reinforcement learning to tie bow on their modelās development, and had only noticed minor increases in performance.
After playing around with DeepSeek, Hereās what I thought of it.
This thing has range
I ran this LLM through what I could design as a gauntlet. Coding questions, philosophy, rationales behind decisions, detailed science questions, niche local law, etc.. and it all came back pretty clean. I steered clear of politics due to some obvious statements in the Privacy Policy. (Just so you know)
Diving deeper into some obscure Quebecās law structures was no problem at all for this model, itās got link, citation as well as recommendations for further sources. When asked a vague technical question about building a DIY air purifier, I turned on the R1 reasoning features, and I got a stream of consciousness response leading up to my āofficial promptā response. Even after further pressing the system by diving deeper into my lines of inquiry, the R1 model impresses in the thoroughness of its responses.
Consistency & Memory Retention were no issue at all.
It reliably remembered conversational context going even 15+ messages deep. Seeing the initial thread of conversation be ingested as part of the R1 āstream of consciousnessā was quite impressive.
This is versatile in tone.
My āwritten-as-spokenā style caused zero issues in getting results I wanted immediately. Conversations switching back and forth between casual tone and formalities was practically flawless.
DeepSeek Deployability - Inconclusive for all things API
The API creation and Docs were Out of Order when I got to it, so Iām a bit disappointed about not being able to test more granular behaviors. In the same vein, I wasnāt able to check if there were any unexpected limitations in the API (think stuff like rate limits and token caps at the API level.
Sidenote for my rapid prototyping folks: I noticed that Bubble.io has launched their 1st party integration API support for DeekSeek, so Iāve made a note to test that out once I can get my hands on some keys.
IN CONCLUSION
With all this being said, weāre now in a more honestly global market of LLMs than ever before, and contenders on all sides are (in their own ways) driving towards a more open, more accessible future for AI as a technology. I will think twice when choosing an LLM provider for my side-projects moving forward. This is a beast of a model.
Of Course itās fair to say weāve already seen the tip of the aftermath iceberg for the DeepSeek effectā¦
Weāve seen a short term dip in NVIDIA Stock like never before. https://www.investing.com/news/stock-market-news/nvidia-stock-bounces-back-after-deep-selloff-3835143
Some have (successfully) started poking at DeepSeekās Security https://www.wiz.io/blog/wiz-research-uncovers-exposed-deepseek-database-leak
Open AIās Sam Altman, admits on the record that theyāre on the wrong side of history when it comes to the organizationās approach to Open Sourcing their technology https://venturebeat.com/ai/sam-altman-admits-openai-was-on-the-wrong-side-of-history-in-open-source-debate/

Your Weekly Industry News Recap
Good Medical News! Curvebeam AI Expands with Stryker Partnership: AI-powered medical imaging firm Curvebeam expands to Australia and New Zealand in partnership with Stryker. š Read more: Link
The API legend veers into AI: Postmanās AI Agent Builder Launch: API development is getting smarter! Postman debuts an AI agent builder to automate API processes. šš” Read more: Link
LāOrĆ©al & IBMās AI for Cosmetics: AI is transforming beauty! LāOrĆ©al and IBM introduce generative AI for personalized skincare. šš¤ Read more: Link
CES 2025 Highlights AI Innovations: AI stole the spotlight at CES 2025, from smart home gadgets to self-driving tech. šš” Read more: Link
The AI Governance Balancing Act (World Bank): How should regulators balance AIās potential with risks? āļøš Read more: Link
Snowflakeās AI VC Insights: What are AI-focused VCs looking for in 2025? Snowflake shares insights on startup investment trends. šš” Read more: Link
Women Leading the AI Revolution: A new dataset identifies the next wave of women shaping AIās future. š©āš»š Read more: Link
AI-Powered Simulation Startups: The future of simulation is AI-drivenāsee how startups are building smarter virtual environments. šš¬ Read more: Link
AI & Product Security (Ericsson): Developers, take note! AI is reshaping securityāhereās what you need to know. ššØāš» Read more: Link
The Future of AI Product Management: AI isnāt just changing how we build productsāitās redefining how we manage them. š§š¤ Read more: Link


The Raise Recap šš°
Orchid Security Raises $36M: Orchid secures $36M to enhance AI-driven identity security, making cybersecurity smarter and stronger. šš¤ Read more: Link
Vasco Raises $11.5M CAD: Helping startups manage revenue, Vasco secures $11.5M CAD for AI-powered financial tools. šš° Read more: Link
MiniMax Unveils New AI Model: Chinese AI company MiniMax secures fresh funding, claiming its model rivals OpenAIās and Googleās. šš¤Æ Read more: Link
Blaize Becomes the First AI Chip IPO of 2025: AI chipmaker Blaize goes public, setting the stage for more AI hardware IPOs. šš¾ Read more: Link
MetaAI Secures $4M in Seed Funding: NVIDIA backs MetaAI with a $4M investment to expand its AI research initiatives. š¤š„ Read more: Link
South African AI Startup Taco Raises $1.6M: Taco secures $1.6M to scale its AI-driven sustainability solutions. š±š” Read more: Link
Former Google VP Raises $10M for AI Retail Startup: Ex-Google exec secures $10M to bring AI-powered solutions to retail businesses. šļøš” Read more: Link
Origami Agents Secures $2M Seed Round: AI-powered sales enablement startup Origami Agents raises $2M to enhance B2B automation. š¤š Read more: Link
Mitiga Secures $30M: AI-powered cybersecurity firm Mitiga lands $30M to enhance cloud and SaaS security incident response. šš° Read more: Link
Wyzr Secures $400K in AI SaaS Funding: Wyzr lands $400K following NatWestās fintech demo dayāAI-powered SaaS solutions are heating up! š„š Read more: Link
Waterlily AI Raises $7M for Predictive Healthcare: AI-driven long-term healthcare predictions just got a boost with Waterlilyās $7M seed funding. š„š” Read more: Link
Cantaloupe AI Wins Startup Slam: AI-powered automation startup Cantaloupe wins the 2025 Nspire Startup Slam, securing major recognition. šš¤ Read more: Link

What Iām Reading - A Small Dip Into Futurology
Our current trajectory is to make AI more human-like, but is that really a good thing? š¤ This article dives into the risks of anthropomorphizing AIāwhen we start treating chatbots and virtual assistants as if they have real emotions and intentions, we blur the lines between tool and entity. The problem? People may trust AI too much, overestimate its capabilities, or even form emotional bonds with systems that donāt actually āthinkā or ācare.ā Should AI really be designed to mimic us, or are we walking into a dangerous illusion? šØ

Did YOU Like This Issue? |

Stay well, and see you next week!
-āš½ Sam
Reply