- AI Business Asia
- Posts
- ChatGPT Gets a Brain Boost
ChatGPT Gets a Brain Boost
And DeepSeek just got called a “threat to national security.

Hello AI Innovators.
The smartest person in the room isn’t the one with all the answers—it’s the one who knows what to ask (and which model to ask it to).
Because in 2025, the question isn’t “Can AI do this?”
It’s “Which AI does it better, cheaper, and without hallucinating halfway through?”
This week’s lineup is full of updates that blur the lines between intelligence, intuition, and surveillance.
Let’s dive in! 🚀
In today’s newsletter ⬇️
🧰 4 AI tools to upgrade your workflow—fast.
🔍 ChatGPT now personalizes your web searches.
🚨 Google’s AI safety report is raising eyebrows.
🌏 China opens its doors to global AI markets.
⚠️ DeepSeek flagged as a national security threat.
🧠 OpenAI’s o3 & o4-mini are breaking benchmarks.
📚 Expert picks + quick links to stay ahead this week.

🚀 Top Tools of the Week!
Monday.com: Streamline tasks, track progress, and collaborate effectively to enhance productivity in any project
Soundraw: AI Music Generator that allows you to create custom tracks tailored to your projects.
TabNine: AI tool that can help you write your code faster.
FeedHive: Social media using AI & Automation.

The News: East Meets West
News from The WEST 👈:
🧠 ChatGPT’s Memory Powers Web Searches
OpenAI is rolling out memory-based personalization for ChatGPT’s web browsing. The tool now recalls your past preferences to tailor search results in real-time.
💡 Why it matters: Search just got personal. If AI remembers what you care about, it can serve better—and sell better. This is ChatGPT quietly becoming your new homepage.
📉 OpenAI’s o3 Didn’t Score as High as We Thought
Turns out OpenAI’s o3 model didn’t perform quite as well on the MMMU benchmark as initially suggested. It scored 54.9%—not the top-tier result many assumed.
💡 Why it matters: It’s a reminder: AI hype ≠ AI truth. As models race for dominance, transparency will be the real trust builder.
⚠️ Experts Slam Google’s AI Report for Missing Safety Details
Google’s latest Gemini AI transparency report is getting flak for dodging hard questions around safety and misuse. Experts say it’s “not even close” to industry standards.
💡 Why it matters: In 2025, trust isn’t optional—it’s currency. And vague safety talk from Big Tech? That’s a red flag for regulators and users alike.
News from The EAST 👉:
🌏 China Unveils Plan to Open Up Services Sector
China’s State Council released a plan to expand foreign access in industries like finance, education, and healthcare—signaling a broader shift in market openness.
💡 Why it matters:
China’s AI ecosystem isn’t just software—it’s being shaped by how fast markets open to global players. This could reshape where the next wave of AI startups land.🚫 US Chip Ban May Be Supercharging Huawei’s AI Rise
U.S. sanctions meant to cripple Huawei may be having the opposite effect. A new report shows Huawei gaining serious ground in the global AI hardware race.
💡 Why it matters: Pressure makes diamonds—and geopolitics may be turning Huawei into China’s Nvidia. The chip war just flipped.
🔒 US House Panel Calls DeepSeek a “Profound Threat”
A U.S. House panel labeled China’s DeepSeek AI as a potential national security threat, citing risks of intellectual espionage and model misuse.
💡 Why it matters: The AI cold war is heating up. As DeepSeek scales globally, scrutiny from Washington might soon turn into real bans or restrictions.

🧠 OpenAI’s New Models Are Quietly Breaking the Game

o3 and o4-mini aren’t upgrades. They’re replacements.
Key Intel:
o3 handles 600 tool calls in one go
o4-mini just beat most humans at coding
Both now see, think, and remember
OpenAI slashed prices by 95% since GPT-4
But here’s what nobody’s telling you: these models are already showing up in real-world apps—and outperforming the bigger names. They’re not just chatbots. They’re plug-and-play reasoning engines that can plan, solve, and personalize—all in one go.
Why It Matters:
Whether you're coding, analyzing, or automating—this duo changes how fast and how well you can do it. And yes, o4-mini is already replacing GPT-4 for a growing number of devs.
⚔️ Who wins in your stack? |


How would you rate today’s newsletter? |
Whether it’s improving your SEO, launching on platforms like Product Hunt, managing newsletters, or building a strong LinkedIn presence, we’re here to help. Curious to see how we can work together?
Until next time!
Leo & Lex

Hello AI Innovators.
Is AI peaking—or just getting started?
Some days it feels like we’re sprinting toward AGI. Other days, we’re watching AI play Quake II... and crash into a wall.
This week, Meta dropped Llama 4. Microsoft showed off an AI-generated game demo (with a few bugs). DeepSeek flexed new reasoning upgrades. Meanwhile, Samsung’s AI chip sales stumbled, and AI copyright lawsuits are heating up fast.
But beyond the headlines? AI apps are making real money, changing how we work, and proving that smart, useful products always win.
Let’s dive in! 🚀
In today’s newsletter ⬇️
🧰 4 AI tools that’ll 10x your workflow this week.
🐑 Meta’s Llama 4 is here—and it’s not playing
🎮 Microsoft drops an AI-made Quake II demo (but it’s… rough)
📉 Samsung’s profits take a 21% AI chip hit
🧠 DeepSeek’s new method makes AI way smarter
📚 Quick links + must-read resources to level up 👀

🚀 Top Tools of the Week!
Dream Studio Beta - Mind-blowing visuals with Stable Diffusion 3.5.
Fiverr Logo Maker - Create a stunning logo in seconds.
Lexica - Snap your imagination into epic AI art.
AI Dungeon - Be the hero of your own adventure.

The News: East Meets West
News from The WEST 👈:
🚨 Meta Drops Llama 4 — And It’s a Beast
Meta just launched Llama 4, its newest open-weight AI model lineup—and it’s gunning straight for GPT-4’s throne. The models are already available to researchers, with Llama 4b (smaller, faster) landing first, and the full-scale Llama 4 model coming soon.💡 Why it matters: Meta’s Llama family is no longer playing catch-up—this release could seriously shake up the open-source AI scene.
🚨 Microsoft’s AI-Generated Quake II Demo Drops—With a Catch
Microsoft just dropped a Quake II level entirely generated by AI, showing off its latest Copilot Runtime tools. The twist? They admit it’s clunky and needs major refinement. It's cool, but not quite ready for primetime.💡 Why it matters: AI in game dev is real—but still rough around the edges. This demo shows what’s possible (and what’s not... yet).
🚨 OpenAI & Microsoft Hit with Merged Copyright Lawsuits
A U.S. court just combined several lawsuits from authors and news publishers accusing OpenAI and Microsoft of copyright infringement. The consolidated case will play out in New York and could set major legal precedent.💡 Why it matters: This could reshape how AI companies use training data—and who gets paid when AI learns from your content.
News from The EAST 👉:
🚨 Samsung Forecasts 21% Profit Drop—AI Chips to Blame
Samsung expects Q1 profits to slump 21%, citing weak AI chip demand and ongoing losses in its foundry business. Not exactly the AI gold rush they hoped for.💡 Why it matters: Even the giants are feeling the pressure—AI chip wars aren’t always a win, especially when demand cools.
🚨 Japan’s Rapidus in Talks to Mass-Produce Chips with U.S. Tech Giants
Japan's homegrown chipmaker Rapidus is deep in talks with American tech companies to start mass production of next-gen semiconductors. Their goal? Be the global alternative to TSMC by 2027.💡 Why it matters: The U.S.-Japan chip alliance could be key to countering China's dominance—and reshaping the global AI supply chain.
🚨 DeepSeek Reveals New AI Reasoning Method
China’s DeepSeek just dropped a new technique aimed at improving LLM reasoning performance. Early tests show it boosts accuracy significantly without bloating compute costs.💡 Why it matters: DeepSeek’s momentum isn’t slowing. This breakthrough could make it a serious rival to GPT and Claude in reasoning-heavy tasks.

The Best AI Apps of 2025—and What’s Fueling Their Success 🚀👀

AI-powered subscription apps aren’t just trending—they’re outperforming nearly every other category in the store.
From ChatGPT and Jasper to Perplexity and Remini, the best AI apps of 2025 are:
Launching faster
Monetizing smarter
Earning 400x more than the average app.
And here’s the kicker—most don’t even offer free trials anymore. They’re that confident.
📈 Key Intel:
AI apps now make $0.63 per install (2x the industry average).
General AI assistants pulled in $726M+ revenue in 2024 alone.
Only 14% offer trials—yet trial start rates are 96%.
The best apps use smarter pricing, better onboarding & vertical focus.
💡 Why It Matters:
If you're building or investing in apps and not thinking about AI—you're already behind. The winners in 2025 are the ones blending fast dev cycles with smart monetization, personalization, and real value.
Which AI app feature matters most to you? |


Whether it’s improving your SEO, launching on platforms like Product Hunt, managing newsletters, or building a strong LinkedIn presence, we’re here to help. Curious to see how we can work together?
Until next time!
Leo & Lex
Reply