- Projects of AI
- Posts
- Samsung tiny AI model beats DeepSeek and Gemini
Samsung tiny AI model beats DeepSeek and Gemini
Plus: Nanobot cancer fighters,
Amandeep Singh
October 09, 2025
Together with
Welcome Humans🤖,
Here is what we have Today:
🧠 Samsung's micro-model beats DeepSeek and Gemini.
🤖 CodeMender auto-patches 72 vulnerabilities.
🧬 Duke's TuNa-AI transforms drug delivery.
🤖 Google's AI navigates UIs like a human.
💼 New Job Opportunities
AI & Medicine
🧬 Duke’s Robot Chemist Reinvents Drug Design
Smart drugs, robotic labs, and AI teamwork — Duke’s TuNa-AI is rewriting the rules of cancer treatment.
Something revolutionary is brewing in Duke University’s labs. It’s called TuNa-AI, and it’s changing how we design nanoparticles — the tiny delivery vehicles that help drugs reach the right spot in the body.
The twist? TuNa-AI blends robotics and machine learning to do what human researchers can’t: test thousands of formulations fast, accurately, and safely.
Image source: Ideogram
⚙️ What TuNa-AI Did
🤖 Tested 1,275 drug formulations using automated lab robots
📈 Boosted success rate by 43% over traditional methods
💉 Improved a hard-to-deliver leukemia drug — dissolving faster and killing more cancer cells
☠️ Cut a toxic ingredient by 75%, while keeping treatments just as effective
⚗️ Optimized both materials + mixing ratios — something old methods couldn’t do
💥 Why It Matters
Most drugs fail not because they’re weak —
but because they can’t reach their targets.
TuNa-AI changes that.
It could revive shelved drugs, design safer therapies,
and accelerate cures for the world’s toughest diseases.
🚀 Big Picture
AI isn’t just analyzing data anymore —
it’s creating better medicine.
Welcome to the future of AI-powered drug design.
Full Story - Duke’s Robot Chemist That Designs Cancer Drugs.
Together with AdQuick
Modernize Out Of Home with AdQuick
AdQuick unlocks the benefits of Out Of Home (OOH) advertising in a way no one else has. Approaching the problem with eyes to performance, created for marketers and creatives with the engineering excellence you’ve come to expect for the internet.
You can learn more at www.AdQuick.com
AI Research
🧠 Samsung’s 7M AI Outsmarts the Titans
A model smaller than a smartphone app just out-reasoned some of AI’s biggest names.
Meet Samsung’s Tiny Recursion Model (TRM) — just 7 million parameters — yet it beats DeepSeek R1 and Gemini 2.5 Pro on complex reasoning puzzles.
How? By thinking like a human.
Instead of spitting out an answer word by word, TRM drafts, rethinks, and refines — looping through up to 16 cycles of internal reasoning before finalizing its answer.
Image Source: Alexia Jolicoeur-Martineau
🔁 How It Thinks
Instead of generating answers token by token...
TRM goes through up to 16 internal reasoning cycles, each time:
🧩 Drafting a solution
🧠 Critiquing its logic (6 times per cycle!)
📝 Updating its “scratchpad” until it gets it right
A true self-improvement loop in action.
📊 The Results
💥 45% on the notoriously tough ARC-AGI-1
🔍 8% on ARC-AGI-2
👉 Beating models thousands of times larger
Sure — TRM’s specialty is puzzle-style reasoning for now...
But it’s a glimpse into how small models can punch way above their weight.
💡 Why It Matters
In an AI world chasing billions of parameters and mountains of compute, TRM proves something radical —
👉 Brains > Brawn.
With smart architecture, even tiny models can reason, self-correct, and rival AI giants.
This could reshape how smaller labs — with limited budgets — build the next generation of AI systems.
🧩 Quick Takeaway
“Efficiency is the new intelligence.”
— TRM shows that the future of AI might not be bigger… just smarter.
Presented by Typeless
AI voice dictation that's actually intelligent
Typeless turns your raw, unfiltered voice into beautifully polished writing - in real time.
It works like magic, feels like cheating, and allows your thoughts to flow more freely than ever before.
Your voice is your strength. Typeless turns it into a superpower.
Trending Today
SF Weekly Pulse - New technology for robotics to walk on water
DeepMind introduced an autonomous agent CodeMender, powered by Gemini Deep Think models that identifies and patches code vulnerabilities using static and dynamic analysis, fuzzing, symbolic reasoning, and an internal “LLM judge” to validate fixes. In testing, it automatically repaired 72 open-source vulnerabilities, including hardening the libwebp image library that was once exploited in iOS zero-click attacks, hinting at a new era of self-healing software.
Google quietly removed the num=100 search parameter, the option that let users and crawlers view 100 results at once instead of the default 10. It sounds minor, but it’s not: AI systems from OpenAI, Anthropic, and Perplexity depend on Google’s index for retrieval and training. With this change, those systems can now see only a tenth of what they used to.
Google DeepMind has launched Gemini 2.5 Computer Use, a powerful AI model that directly operates web and mobile UIs visually (clicking, scrolling, typing, filling forms, navigating dropdowns, and even functioning behind logins). This makes it possible for agents to complete real multi-step workflows, like transferring CRM data, organizing notes, and automatically testing user interfaces, all through a built-in computer-use tool available in the Gemini API (Google AI Studio) and Vertex AI.
Learn how to make AI work for you
AI won’t take your job, but a person using AI might. That’s why 1,000,000+ professionals read The Rundown AI – the free newsletter that keeps you updated on the latest AI news and teaches you how to use it in just 5 minutes a day.
Job Opportunities
AI News
Click on the image to know full story 👇️
Source: Ideogram
Ideas? Comments? Complaints?
We read your emails, comments and poll replies daily.
Get the most important AI, tech, and science news in a free daily email.
New Here? Subscribe!
Sponsorship Slots Open for October and Reach over 800+ active readers. (Now 40% off) 🤯
What`d you think of today`s edition? |
Until next time, Stay Informed!
Reply