👋 Leaving Meta and PyTorch
A major milestone for the AI world — Soumith’s departure marks the end of an era for open-source AI leadership. His confidence in PyTorch’s future shows it has matured into a robust, self-sustaining platform for global AI innovation.
Soumith Chintala announced his departure from Meta and the PyTorch project after over a decade with the company and nearly eight years leading PyTorch’s growth. In his farewell post, he reflected on PyTorch’s evolution from a research framework to a cornerstone of global AI infrastructure, praised its community and leadership, and expressed confidence in its continued success. He also shared gratitude to collaborators across Meta and the open-source ecosystem and revealed plans to explore new, smaller projects driven by curiosity.
🔗 Read more 🔗
🧪 Results from Testing Six AI Models on Advanced Security Exploits
🔗 Read more 🔗
⚙️ AI’s 70% Problem
Osmani’s ‘70% problem’ insight captures the balance between AI’s speed and the enduring need for human craftsmanship. His message underscores a shift toward responsible, human-centered AI engineering.
In a talk hosted by Zed, Google’s Addy Osmani explored what he calls the “70% problem” in AI-assisted coding. He noted that while AI can generate around 70% of code efficiently, the remaining 30% — involving production readiness, security, and complex edge cases — still requires deep human involvement. Despite increasing adoption of AI coding tools, developer trust continues to waver. Osmani stressed that engineers must stay responsible for design, reviews, and quality, using AI as a helper, not a replacement.
🔗 Read more 🔗