In partnership with
Dear Readers,
Today, we are witnessing a moment in which artificial intelligence no longer just sorts data, but excels under competitive conditions. OpenAI and Google DeepMind won gold at the ICPC World Finals, one of the most demanding programming competitions, with OpenAI even achieving a perfect score. This shows that AI is rapidly moving toward creative problem solving, a field that was previously considered the domain of humans. What does it mean when machines not only calculate, but also develop strategies independently?
In this issue, we look at precisely this shift: from abstract benchmarks to real thinking power. We report on new records set by ARC-AGI, exciting advances in medical AI, and the unexpected role of accessibility as a killer feature in the world of work. We also take a look at the longevity scene, where Sam Altman's startup is testing a radical anti-aging pill. So there's plenty to read – and every story is a reminder of how close the future is coming.
In Today’s Issue:
🥇 AI just won gold at the programming Olympics
👩🦯 AI is making the workplace dramatically more accessible
🧠 Sam Altman's longevity startup is testing a pill to make your brain younger
⏰ This new multi-agent AI can build more accurate and reliable "aging clocks"
✨ And more AI goodness…
All the best,
NotebookLM's Flashcards & Quizzes Fully Launched
NotebookLM has officially completed the global rollout of its Flashcards and Quizzes features to all users, making it a comprehensive tool for learning and exam preparation.
Meta Unveils AI Display Glasses
Meta's Reality Labs has introduced Meta Ray-Ban Display, a breakthrough category of AI glasses designed to integrate advanced AI capabilities with a visual display.
Studio 3.0: AI Audio & Video Editor
Studio 3.0 has launched, offering the most advanced AI audio models in a single editor, now with video support, automatic captioning, speech correction, and multiplayer commenting for comprehensive audio-visual production.
Last week, our reasoning models took part in the 2025 International Collegiate Programming Contest (ICPC), the world’s premier university-level programming competition. Our system solved all 12 out of 12 problems, a performance that would have placed first in the world (the best
— #Jakub Pachocki (#@merettm)
5:17 PM • Sep 17, 2025
OpenAI's GPT-5 and an experimental reasoning model won gold at the ICPC (International Collegiate Programming Contest). After already winning gold at the IMO (International Math Olympiad), this is the next big step.
In this interview, Google DeepMind CEO Demis Hassabis shares his views on the development of AI models in medicine.
New SOTA on ARC-AGI
- V1: 79.6%, $8.42/task
- V2: 29.4%, $30.40/taskCustom submissions by @jerber888 and @_eric_pang_ are now the best known solutions to ARC-AGI
Both:
* Are open source
* Use Grok 4
* Implement program-synthesis outer loops with test-time adaptation— #ARC Prize (#@arcprize)
5:07 PM • Sep 16, 2025
There are new top scores on the demanding ARC-AGI benchmark: While version 1 achieves 79.6% at a cost of $8.42 per task, version 2 achieves 29.4% at $30.40 per task. The best known solutions currently come from @jerber888 and @eric_pang. Both approaches are open source, based on Grok 4, and rely on program synthesis outer loops with test-time adaptation to solve the tasks efficiently.
AI wins gold at programming Olympics
The Takeaway
👉 Gemini 2.5 Deep Think solves 10/12 problems in the ICPC under live competition conditions, achieving gold level.
👉 OpenAI achieves 12/12 – a perfect solution in the same competition – demonstrating peak performance.
👉 These successes show that abstract thinking + rapid implementation = AI can solve problems creatively, not just repeat data.
👉 Practical consequence: AI assistants could soon take on tasks that previously required highly experienced developers – or open up new areas where human capabilities are limited.
When AI solves as many programming tasks in five hours as top students do, you know that something fundamental has happened.
DeepMind's Gemini 2.5 Deep Think achieved gold medal level at the 2025 ICPC World Finals (International Collegiate Programming Contest), solving 10 out of 12 difficult algorithm problems under the same conditions as the human teams.
Even more impressive: OpenAI claims to have solved all 12 tasks in the same competition – a perfect solution.
Why is this so exciting right now? Because it's not just about arithmetic or code generators, but abstract thinking, quick planning, and creative problem solving—skills that for a long time were thought to be unique to humans. This achievement brings us one step closer to AI systems that can handle truly complex, new tasks.
What if we soon have AI coaches who participate in programming competitions with us? And AI systems that propose new mathematical conjectures instead of just analyzing existing ones? The path to AGI feels like a mountain that we are now visibly climbing.
Why it matters: It shows that AI is not only strong in standardized benchmarks, but also in real, variable competitions with time pressure and high complexity.
This makes it clearer how AI can soon become a real partner in research, teaching, and innovation—not just a tool.
Sources:
Your Shopify DTC Brand Can’t Afford Q4 Without Zipchat
BFCM traffic costs a fortune. If your Shopify brand isn’t converting at its possible best, you’re not just losing sales — you’re burning money and shrinking Q4 margins.
Zipchat.ai is the AI Agent built for DTC ecommerce. It doesn’t just chat — it sells.
Closes hesitant shoppers instantly with product answers and recommendations
Recovers abandoned carts automatically via web + WhatsApp
Automates support 24/7 so you scale without extra headcount
Boosts profit margins in Q4, when every order counts
That’s why brands like Police, TropicFeel, and Jackery — brands with 10k visitors/month to millions — trust Zipchat to handle their busiest quarter and fully embrace Agentic Commerce.
Setup takes less than 20 minutes with our success manager. And you’re fully covered with 37 days risk-free (7-day free trial + 30-day money-back guarantee).
On top, use the NEWSLETTER10 coupon for 10% off forever.
“I'm back at the top of ARC-AGI with my new program. I use @grok 4 and multi-agent collaboration with evolutionary test-time compute” (Jeremy Berman)
Why accessibility might be AI’s biggest breakthrough
A British government study on Microsoft's M365 Copilot shows that neurodivergent employees report significantly higher satisfaction and willingness to recommend than neurotypical colleagues. Those with ADHD and dyslexia benefit particularly; live transcripts helped those with hearing impairments. Productivity gains remained small and 22% reported hallucinations – but AI is already breaking down barriers. Conclusion: Accessibility could be the first real “killer use case” in the office.
Sam Altman’s longevity startup is testing a pill for a younger brain
Retro Biosciences, a longevity company backed by Sam Altman, will launch the first Phase 1 clinical trial in late 2025 with an experimental pill called RTR242, which is designed to reactivate impaired autophagy in the brain to eliminate Alzheimer's-like protein deposits. In addition to this approach, Retro is working on other programs such as blood stem cell therapies. If this works, it could significantly extend not only lifespan, but above all healthy lifespan.
Multi-agent AI builds more accurate aging clocks
A bioRxiv preprint presents a guided multi-agent approach that designs highly accurate, uncertainty-aware transcriptomic aging clocks. In addition to the age score, the system also provides uncertainties and marks sensitive zones (e.g., transitional and extreme ages). Potential: better patient stratification, monitoring of interventions, and more robust studies. Not yet peer-reviewed—validation pending.
Grow & Monetize Your Audience with ConvertKit.
Unlock the power of email to connect with your community and build your business with ConvertKit. Built for creators, writers, and builders.
Simple to start, powerful to scale.
