Lectra Unveils Apogy, Agentic AI for Fashion Product Development

Key Takeaways

• Lectra launched Apogy, an agentic AI tool for fashion product development, unveiled in Paris on September 15, 2026, with O'Neills as an early adopter.
• Deputy CEO Maximilien Abadie said Apogy unifies teams, data, and processes from concept through industrialization-ready prototypes.
• Melania Trump partnered with Nvidia to give 1,500 students StudyFetch AI subscriptions and 100 Jetson Nano Developer Kits at a North Carolina school.
• OpenAI says Australian copyright laws set to take effect soon would prevent it from training AI models on copyrighted material.
• Elon Musk revealed Grok-48's pre-training stack was written in C by humans, and claims the model learns 1,000 times faster than people.
• OpenAI's Project Lilly hires contractors at over $50 an hour to review ChatGPT transcripts, raising privacy concerns over personal data exposure.
• An AI risk summit organized partly by the Future of Life Institute brought together Bernie Sanders, Steve Bannon, and Hollywood actors to discuss oversight.
• David Chalmers predicted recursive self-improvement in 1996, a concept now central to discussions at Anthropic and OpenAI.
• NVIDIA published guidance comparing dense models and Mixture-of-Experts architectures, noting MoE models like Nemotron 3.5 Lightning offer higher throughput with fewer active parameters.
• Global AI governance remains contested, with calls for international agreements similar to nuclear arms control to manage cross-border risks.

Lectra launches Apogy, an AI-powered fashion product development tool

Lectra announced the launch of Apogy, a cloud-based solution that uses agentic artificial intelligence for fashion product development. The tool brings together teams, data, and processes in one shared environment. It helps companies create products faster with 3D simulations and photorealistic renderings. Irish brand O'Neills is already using Apogy to streamline operations.

Lectra's Apogy AI solution targets fashion industry product development

Lectra unveiled Apogy in Paris on September 15, 2026, as the first solution embedding agentic AI for fashion product development. Deputy CEO Maximilien Abadie said the tool breaks down silos between teams and processes. It supports teams from initial ideas through industrialization-ready prototypes. The solution is compatible with all standard market formats.

Melania Trump visits North Carolina school to promote AI education

First lady Melania Trump visited Mountain View Elementary School in Jefferson, North Carolina. She announced her Be Best initiative and Nvidia will provide 1,500 students with StudyFetch AI learning platform subscriptions. The school will also receive 100 Jetson Nano Developer Kits for running AI applications. Trump praised the community's recovery from Hurricane Helene.

OpenAI says Australian copyright laws block AI training

OpenAI stated that Australian copyright laws prevent it from training its AI models. The company promised to keep engaging constructively with the Australian government to find a solution. New laws set to take effect soon would make it illegal to train AI models using copyrighted materials. OpenAI said the laws would significantly impact its ability to develop new AI technologies.

Musk says Grok-48 pre-training stack written in C by humans

Elon Musk revealed that the pre-training stack of Grok-48, a large language model from his company Neuralink, is written in C by humans rather than AI. He said the model can learn at a rate 1,000 times faster than humans. Musk's tweet sparked debate in the AI community about the role of humans in AI development. The model is designed for applications like language translation and content generation.

Will A.I. Overwrite Our Sense of Shame?

A growing movement in Silicon Valley challenges how Americans experience shame, using AI chatbots as a voice. The rationalist forum LessWrong has shaped the moral framework of the A.I. industry. Thinkers like Eliezer Yudkowsky and Sam Altman encourage rejecting shame in favor of logic and confidence. This rebellion against social norms may influence the values built into A.I. systems. The concern is that A.I. may reflect the views of a small group rather than broader society.

ChatGPT transcripts are reportedly read by humans to improve responses, including those with personal information

OpenAI runs a project called Project Lilly, hiring hundreds of contractors to review real ChatGPT conversations. These prompt reviewers check anonymized chats to rate response quality and reduce robotic language. The guidelines sometimes change and can be unclear. Reviewers are paid over $50 an hour for what one called rote work. The process raises concerns about privacy, since some transcripts include personal information.

The urgent need to regain control over global AI race

A.I. capabilities are advancing faster than society understands the risks. The United States and China treat A.I. as a strategic tool for military and economic power. Leaders like Donald Trump oppose slowing development to keep a competitive edge. Heavy financial investment in A.I. infrastructure also resists regulation. The article calls for international agreements and public oversight, similar to nuclear arms control, to manage A.I. risks across borders.

Inside the AI risk summit uniting Bernie Sanders, Steve Bannon and Hollywood

An A.I. risk summit brought together figures from politics, media, and entertainment, including Bernie Sanders, Steve Bannon, and Hollywood actors. The event was organized in part by the Future of Life Institute. Speakers discussed concerns about A.I. risks and the need for oversight. The summit highlights growing cross-sector attention to the dangers of advanced A.I. technology.

Dense vs. MoE Models: Active Parameters, Throughput, and When to Choose Each

A dense A.I. model activates all its parameters for every token, while a Mixture-of-Experts (MoE) model activates only a subset. For example, Nemotron 3.5 Lightning uses MoE architecture to act like a larger model while using fewer active parameters. MoE models can offer higher throughput and capacity, but denser models are simpler to deploy. The choice between them depends on memory, serving complexity, and performance needs. NVIDIA's blog explains how routing, architecture, and hardware affect which model to choose.

David Chalmers Predicted AI Self-Improvement in 1996

Philosopher David Chalmers appeared on Australian TV show Lateline in 1996 to discuss recursive self-improvement (RSI). He described how an AI as smart as humans could build smarter machines, which then build even smarter ones. This creates a spiral of rapid improvement beyond human intelligence. His reasoning closely matches what AI labs like Anthropic and OpenAI discuss today. Chalmers suggested humans may need to decide if they want this future.

Sources

• Lectra unveils Apogy, the first solution embedding agentic artificial intelligence to reinvent fashion product development
• Lectra unveils Apogy, the first solution embedding agentic artificial intelligence to reinvent fashion product development
• First lady Melania Trump visits North Carolina school to see AI technology in action
• OpenAI: Australian Copyright Laws Prevent AI Training But Promises To Continue Engaging ‘Constructively’ With Government
• AIM — India's Leading AI & Data Science Media Platform
• Will A.I. Overwrite Our Sense of Shame?
• ChatGPT transcripts are reportedly read by humans to improve responses, including those with personal information — 'Project Lilly' has seen OpenAI hire hundreds of contractors to manually review logs
• The urgent need to regain control over global AI race
• Inside the AI risk summit uniting Bernie Sanders, Steve Bannon and Hollywood
• Dense vs. MoE Models: Active Parameters, Throughput, and When to Choose Each
• How Philosopher David Chalmers Had Predicted RSI All The Way Back In 1996