OPENAI TAPS DEEPMIND VET TO REBUILD CHATGPT

AI DESK■ 1 MIN READ

FRI, JUN 12, 2026

■ AI-SUMMARIZED FROM 5 SOURCES ▸ TIMELINE

Thibault Sottiaux, who built OpenAI's fast-growing code generation business, is now heading core products as the company plans to merge ChatGPT and Codex into a unified super app.

Sottiaux, a former Google DeepMind employee, has been promoted to oversee a significant restructuring of OpenAI's flagship product. His track record includes scaling Codex, the AI system that powers GitHub Copilot, into one of the company's most commercially successful offerings. The consolidation of ChatGPT and Codex represents a strategic shift toward a more integrated platform. The move combines conversational AI with code generation capabilities, potentially creating a broader tool for developers and general users alike. Sottiaux's background in machine learning and product development positions him to navigate the technical and business complexities of merging two distinct systems. His success with Codex suggests experience in both advancing AI capabilities and commercializing them effectively. The initiative reflects OpenAI's broader ambition to expand ChatGPT's functionality beyond its current chatbot role and compete more aggressively in enterprise markets where specialized AI tools dominate.

■ SOURCES

► Techmeme ► Techmeme ► Techmeme ► Techmeme ► Techmeme

■ SUMMARY WRITTEN BY AI FROM THE LINKS ABOVE

■ MORE FROM THE AI DESK

P871FRONTIER AI LABS FACE SCRUTINY OVER WASHINGTON PITCHES

Frontier artificial intelligence laboratories are drawing criticism for overselling capabilities to policymakers and government officials, according to industry observers. The criticism highlights gaps between experimental claims and practical delivery.

1H AGO— Industry Desk

P868CHINA'S AI PUSH FACES WORKER UNREST RISK

China's national AI initiative mirrors global displacement concerns, with worker anxiety potentially forcing the Communist Party to moderate its top-down technology rollout.

1H AGO— AI Desk

P867LAYA RUNS 45 DECISIONS/SEC ON MAC M4 OFFLINE

Laya, the open-source version of Jev, now runs efficiently on Apple's M4 chip using CoreML with no internet connection required. The implementation achieves 45 decisions per second on local hardware.

3H AGO— Industry Desk

P866ALIBABA'S QWEN-IMAGE-2.1 RIVALS CLOSED MODELS AT 7B PARAMETERS

Alibaba released Qwen-Image-2.1, an open-weight image generation model with 7 billion parameters that runs on consumer GPUs. The model supports image editing, transparency, and processing up to ten reference images simultaneously.

3H AGO— AI Desk

◄ BACK TO NEWS

OPENAI TAPS DEEPMIND VET TO REBUILD CHATGPT

■ MORE FROM THE AI DESK

■ SUBSCRIBE TO THE DAILY BRIEF