- The Shift
- Posts
- AI Sees What Doctors Can’t
AI Sees What Doctors Can’t
Plus, 🧠 AI System Gauss Cracks Long-Standing Math Challenge, Thousands of Dollars Worth of AI Training now for free, and more!
Hello Readers👀
Curious about the biggest AI moves today? You’re in the right place. In today’s edition, we have:
🧬 Harvard’s PDGrapher AI Could Transform Drug Discovery
🧠 Thousands of Dollars Worth of AI Training, Now Free
🔢 AI System Gauss Cracks Long-Standing Math Challenge
🔨Tools and Shifts you cannot miss
🧬 Harvard’s PDGrapher AI Could Transform Drug Discovery
Harvard Medical School has unveiled PDGrapher, a free AI model that pinpoints gene and drug combinations capable of restoring diseased cells to healthy states. By analyzing complex cellular interactions, it goes far beyond traditional single-target drug discovery.
The Shift:
1️⃣ Multi-Target Mapping - PDGrapher examines entire gene–protein networks instead of one drug target at a time, uncovering multiple cellular “pressure points” that can reverse disease more effectively than conventional methods.
2️⃣ Superior Accuracy and Speed - When tested across 19 cancer types, PDGrapher outperformed competing AI systems by 35% in accuracy and delivered answers up to 25 times faster, dramatically shortening drug discovery timelines.
3️⃣ Validated with Real-World Data - The AI accurately predicted known lung cancer treatments and revealed promising new targets such as KDR (VEGFR2) and TOP2A, aligning with existing and emerging clinical evidence.
4️⃣ Broad Disease Applications - Researchers are already applying PDGrapher to explore treatments for complex brain disorders like Parkinson’s and Alzheimer’s, with ongoing collaborations at Massachusetts General Hospital.
PDGrapher’s ability to identify multiple therapeutic targets simultaneously offers a faster, more precise path to breakthrough treatments for diseases that have defied traditional approaches, potentially saving billions in failed trials and accelerating the arrival of personalized medicine.
🧠 Thousands of Dollars Worth of AI Training, Now Free
The AI giants just gave away their playbooks. Google, Anthropic, and OpenAI released world-class guides on prompt engineering, agent building, and enterprise integration—no fluff, just gold.
Here’s what’s inside:
Google’s Prompting Guide – Practical tips for getting the best from Gemini
→ Prompting GuideAnthropic’s Claude Guides – Agent building + Claude prompt tactics
→ Effective Agents
→ Prompt EngineeringOpenAI’s Enterprise Guides – From use case discovery to live AI agents
→ Agent Building Guide
→ Scaling Use Cases
→ AI in the Enterprise
If you’re building with AI, start here. These aren’t theory decks—they’re senior-level execution manuals.
🔢 AI System Gauss Cracks Long-Standing Math Challenge
Math Inc.’s new AI agent, Gauss, has solved the Strong Prime Number Theorem in only three weeks, a task that world-class mathematicians Terence Tao and Alex Kontorovich wrestled with for 18 months.
The Shift:
1. Breakthrough in Record Time - Gauss tackled the Strong Prime Number Theorem formalization challenge, originally set in 2024, autonomously, generating over 25,000 lines of verified Lean code and 1,000+ interconnected proofs and definitions, compressing years of work into weeks.
2. Beyond Human Progress - While top mathematicians managed only intermediate progress and a “Medium” version of the theorem in 18 months, Gauss leapfrogged ahead, formalizing complex analysis results that had stymied human experts.
3. Infrastructure at Scale - The project leveraged Trinity environments and Morph Labs’ Infinibranch infrastructure to run thousands of concurrent Lean agents using multiple terabytes of RAM, proving that AI can handle unprecedented formalization workloads.
4. Path to Verified Superintelligence - Math Inc. plans to expand Gauss’s mathematical code output by 100–1000× within a year to train future “machine polymaths” capable of verified superintelligence, paving the way for autonomous discovery in advanced mathematics.
Complex mathematics is a foundation of scientific progress, and Gauss’s rapid formalization shows AI’s growing ability to generate new, verifiable knowledge. This breakthrough could dramatically shorten timelines for proving deep theorems and create massive datasets for next-generation AI systems capable of creative, rigorous reasoning.
🔨AI Tools for the Shift
💻 LeetPilot – An invisible AI assistant that supports you through coding interviews seamlessly.
🎧 ASMR.so – Transform your ideas into relaxing AI-generated ASMR videos in seconds with the advanced VEO3 generator.
🧸 Daisy | Labubufy – Turn any outfit photo into a magical Labubu character with AI-powered artistry.
🔁 RepurposingBot – Transform one article into dozens of posts for every platform with ease.
💬 FluxyAI – A 24/7 AI chatbot that delivers consistent and reliable customer support.
🚀Quick Shifts
📰 Rolling Stone’s parent, Penske Media, claims Google’s AI summaries divert readers from source links, slashing affiliate revenue by one-third. The lawsuit argues that Google illegally profits from journalists' content without proper compensation.
⚡ Business Insider reports xAI abruptly laid off over 500 general AI tutors who trained Grok, pivoting instead to expand a specialized AI tutor team tenfold for focused, high-skill training.
🎉 Mets Team Up with OpenAI for ChatGPT Fun at Citi Field. Fans can snap AI-generated portraits, grab a custom Mets pin powered by ChatGPT, and explore an AI-written Citi Field guide, bringing a playful tech twist to game day.
💡 OpenAI board chair Bret Taylor agrees with CEO Sam Altman that we’re in an AI bubble, likening it to the late-’90s dot-com era: massive potential, inevitable shakeouts, but lasting transformation.
📖 AI chatbots are increasingly shaping spiritual exploration, with apps like Bible Chat surpassing 30 million downloads and Hallow topping Apple’s App Store. Experts caution that these bots reflect user opinions rather than true spiritual discernment.
Reply