Humans: Learn To Learn; AI Startup Growth Hacks; Reversing Disease
🎓 Deepmind CEO's Warning
Google DeepMind CEO Demis Hassabis told an audience in Athens that the most important skill for the AI era is “learning how to learn.”
“One thing we’ll know for sure is you’re going to have to continually learn ... throughout your career,” he said.
He predicted artificial general intelligence (AGI) could arrive within a decade, reshaping careers and requiring workers to continually adapt their skills. Lifelong upskilling, he argued, will no longer be optional but essential.
“It’s very hard to predict the future, like 10 years from now, in normal cases. It’s even harder today, given how fast AI is changing, even week by week,” Hassabis, a Nobel laureate, told the audience. “The only thing you can say for certain is that huge change is coming.”
Read more: The Rundown
💊 Harvard’s AI Helps Reverse Disease In Cells
Harvard Medical School introduced PDGrapher, a model that analyzes genes, proteins, and signals in combination.
It predicted drug targets across 19 cancer types with 35% better accuracy than rivals and delivered results 25x faster. Validated against known lung cancer therapies, it also suggested new approaches for Alzheimer’s and Parkinson’s. Researchers say the tool could slash billions wasted on failed drug trials.
Read more: The Rundown
🔐 Google Cracks The Privacy Paradox With VaultGemma
Google launched VaultGemma, a 1B-parameter model that delivers full differential privacy without crippling performance - a breakthrough long thought impossible.
By following new “DP scaling laws,” VaultGemma balances noise injection with training stability. Open-sourced on Hugging Face, it offers banks, hospitals, and regulators a production-ready privacy solution.
According to Google Research team:
"VaultGemma represents a significant step forward in the journey toward building AI that is both powerful and private by design. By developing and applying a new, robust understanding of the scaling laws for DP, we have successfully trained and released the largest open, DP-trained language model to date.
"While a utility gap still exists between DP-trained and non-DP–trained models, we believe this gap can be systematically narrowed with more research on mechanism design for DP training. We hope that VaultGemma and our accompanying research will empower the community to build the next generation of safe, responsible, and private AI for everyone."
Read more: AI Secret
📈 Gamma’s Growth Hack Beats AGI Dreams
Gamma, an AI presentation startup, revealed it scaled to $50M ARR in under two years by focusing on influencer marketing and aggressive user testing - not massive AI breakthroughs.
Its “Zero-Waste GTM” strategy leaned on AI to strip away 70% of busywork, letting tiny teams scale at startup speed. The result: profitability while peers burned cash on compute and ads.
Read more: AI Secret
🎶 AI Music And Soundscapes For Enterprises
Stability AI’s Stable Audio 2.5 (via Replicate) lets users create professional-grade tracks for ads and branded content at about $0.20 per piece. With parameter controls for tempo, genre and prompt detail, the tool offers 90-second outputs for commercial use. Tutorials recommend prompt engineering and remixing for maximum variety.
Read more: The Rundown
China Corner
🔥 Tencent Poaches OpenAI Researcher
Tencent recruited Yao Shunyu, a prominent researcher from OpenAI, highlighting the intensifying global talent war. The move comes as China pushes domestic chip and model innovation, pairing hardware breakthroughs like SpikingBrain with high-profile hires.
Read more: There’s An AI For That; The Rundown*
🧠 China’s Brain-Like AI Runs 100x Faster On Domestic Chips
Researchers in Beijing unveiled SpikingBrain 1.0, a neuromorphic AI that mimics the way human neurons fire. Running on China’s MetaX chips (not Nvidia), the system trained 7B and 76B models using just 2% of the data needed by conventional LLMs. Tests showed a 4M-token prompt processed 100x faster than standard systems, with stability lasting weeks. A free demo called Shunxi is online.
Read more: The Rundown
⚡ Qwen3-Next Pushes Efficient Model Design
Alibaba released Qwen3-Next, an 80B-parameter hybrid model tuned for efficiency. Instead of activating all parameters, it selectively engages subsets - slashing compute costs while maintaining top-tier accuracy. The design could mark a turning point toward sustainable scaling.
Read more: There’s An AI For That
❤️ ICYMI: Apple Watch Gets Blood Pressure Detection
Apple Watch (via watchOS 26) now offers FDA-cleared hypertension notifications on newer models, edging wearables deeper into real health territory (not just steps and vibes).
If adoption mirrors AFib alerts, expect a wave of earlier interventions - and lots of “my watch told me to call you” moments for clinicians.
Read more: There’s An AI For That
🛠️ Tools & Launches
- Guidde Magic Mic - Turns live narration and workflow clicks into polished AI-voiced guides. Ideal for onboarding or training.
- GraphBit Framework - Rust-powered speed with Python simplicity for agent development. Promises enterprise-grade performance.
- AI SEO Score - Benchmarks and tracks brand visibility inside AI systems, a new metric for the LLM search era.
- Scaloom Reddit Automation - Handles posting, replying, and account warming to automate community growth.
- Ito.ai - Dictate anywhere: pipes your speech into any text box for rewrites, edits or tone adjustments.
- Dia - Generates ultra-realistic dialogue in one pass from transcripts, with controllable tone and emotion.
- Kortix - A platform to build and train sophisticated AI agents across diverse use cases.
- Stable Audio 2.5 - Enterprise music generator with commercial rights and fine-grained controls.
Today’s Sources:
- The Internet
- AI Secret
- The Rundown AI
- There’s An AI For That