ThursdAI - The top AI news from the past week

Join AI researcher Jeremy Berman, who recently topped the ARC-AGI leaderboard, as he dives into his innovative approach to prompt and program generation. He shares insights on the impressive performance of GPT-5 during the ICPC competition, tackling all 12 questions flawlessly. Discover the latest advancements, including the powerful capabilities of the new GPT-5-Codex model and exciting announcements from Meta, including their new display glasses. Berman also discusses open-source contributions that are outpacing larger models in deep research and image understanding.

This week in AI news brings exciting developments! ByteDance's SeeDream 4 impresses with its ability to generate stunning 4K images and allows for editing through natural language. OpenAI's new Multi-Channel Protocols enhance ChatGPT's functionality. Mistral's rise as a Decacorn signals major shifts in European tech, while advancements in video generation, including Lucy, showcase speed and quality improvements. Plus, innovative tools and models like Descartes AI's Lucy 14B and Baidu's Ernie 4.5 are making waves in the market!

Roger Jin, a member of the post-training team at Nous Research, shares insights on the powerful Hermes 4 release and the intriguing PokerBots benchmark. Kwindla Kramer, founder of Daily, dives into OpenAI’s RealTime voice advancements. The duo discusses the resurgence of Codex and its exciting performance against competitors. They also touch on significant funding news, recent AI acquisitions, and the evolving landscape of AI models, highlighting challenges in multi-turn interactions and the importance of efficiency in voice technology.

This week, a wave of AI innovations took center stage. DeepSeek's hybrid V3.1 model boasts improved speed and efficiency. ByteDance introduced a long-context model with a unique thinking budget, while NVIDIA unveiled a faster model using open training data. The mind-bending image editor, Nano Banana, promised advanced scene edits. Additionally, OpenAI rolled out Agents.md to streamline development tools, and a new framework reportedly enhanced processing capabilities significantly. Exciting times are ahead in the AI landscape!

In a week of AI excitement, the hosts dive into their experiences with GPT-5, revealing surprising results and challenges. Open-source world models take center stage, showcasing real-time interactions and potential for gaming. Insights on a small model's web performance and advanced prompting techniques add depth. Discussions on emerging language models and their efficiency, plus innovations like interactive video game modeling, highlight the ever-evolving landscape of AI technologies. Get ready for a thrilling ride through the future of AI!

Derya Unutmaz, an immunology expert with a background in data analysis, joins the discussion on the groundbreaking release of GPT-5. They dive into the technical marvels of this new model, highlighting its enhanced reasoning and coding capabilities. The conversation also covers the implications of open-source AI, a shift that democratizes access to advanced technology. Derya shares insights on how GPT-5 can revolutionize industries like healthcare and finance, making it an exciting time for AI advancements.

A whirlwind of AI advancements steals the spotlight this week! New Qwen models from Alibaba are setting benchmarks, while Runway pushes video capabilities to new heights. The excitement around GPT-5 is palpable, with discussions about emerging multimodal technologies. Innovators are racing ahead, incorporating efficient cost models and local services. Plus, breakthroughs in 3D world generation are bridging gaming and AI creativity. It's a landscape rich with developments that promises to reshape our digital experience!

This week’s guests are Junyang Lin, a vital member of Alibaba's Qwen team, known for his work on large language models, and Joseph Nelson, co-founder and CEO of Roboflow, who recently engaged with the White House on its AI Action Plan. They discuss the exciting release of multiple new AI models from Alibaba, including the game-changing Qwen3-235B-A22B, along with insights into governmental regulations affecting AI. Plus, a live coding challenge creates a Tetris game, showcasing innovative AI integration and impressive problem-solving in competitive settings!

This week, a groundbreaking AI model with a trillion parameters was unveiled, shaking the foundations of open-source technology. The introduction of animated characters by Grok AI adds a cultural twist to the innovation landscape. Exciting developments include the launch of new AI agents and advancements in inference services. The podcast also highlights recent shifts in licensing affecting AI enterprises and the importance of community feedback for benchmarking AI models. Tune in for insights on the competitive dynamics within the AI sector!

Maxime Labonne, Head of Post Training at Liquid AI, and Elie Bakouch from Hugging Face, wade into the riveting world of AI advancements. They discuss the launch of Grok 4, a groundbreaking model that sets new benchmarks. Elie shares insights on SmolLM3, an open-source marvel that performs complex reasoning. The dynamic duo also highlights the rise of open-source tools, including Reka Flash, and the innovative Perplexity Comet browser. Join their conversation on the future of AI, its challenges, and how it's reshaping the tech landscape!