Adobe Firefly + Luma AI: Firefly gets a cinematic boost with the Ray3 HDR video model, delivering true HDR, advanced physics, and seamless Creative Cloud exports.

• Adobe Firefly x Luma AI: Adobe brings cinematic-grade video generation into Firefly with Ray3 HDR video model, offering creators true HDR, advanced physics, and seamless Creative Cloud workflows.

• AI Outperforms Humans in Coding: At the ICPC World Finals, OpenAI’s GPT-5 achieved a perfect score, while Google’s Gemini 2.5 solved problems no human team could, marking a historic milestone in programming contests.

• Meta’s Next-Gen Smart Glasses: From Ray-Ban Meta Gen 2 to Oakley Meta Vanguard and the groundbreaking Ray-Ban Display with private in-lens overlays, Meta is merging AI, fashion, and immersive tech like never before.

From cinematic video to coding milestones and AI-powered eyewear, the future of creativity, productivity, and style is here.Adobe Firefly Debuts Luma AI's Cinematic HDR Video Gen Model

Luma AI’s Ray3 video model is now live in the Adobe Firefly app, offering paid users exclusive early access until October 1, 2025, for unlimited cinematic video generation up to 10 seconds with native 16-bit HDR, advanced physics, motion blur, scene consistency, and multimodal reasoning. Ray3 supports text-to-video and image-to-video, prototyping b-roll, dynamic transitions, and high-fidelity shot exploration in Firefly Boards, all exportable to Adobe Creative Cloud apps for full editing workflows. All Firefly Ray3 content is tagged with Content Credentials for full transparency, and nothing created in Firefly is trained back into generative models, ensuring privacy and creative control. This early-access integration, featuring powerful Draft and Keyframe Modes, lets creators preview and master productions in true HDR, positioning Firefly as a leading cinematic-grade AI video tool.

GPT and Gemini Surpass Humans at Global Coding Contest

At the International Collegiate Programming Contest (ICPC) World Finals, OpenAI’s GPT-5 and Google’s Gemini 2.5 Deep Think both achieved gold-medal performances, surpassing all human competitors. GPT-5 delivered a perfect score, solving all 12 complex algorithmic problems under the same official contest conditions as human teams. Gemini solved 10 out of 12 problems, including one that no human team could crack, ranking second overall among university teams. These AI models demonstrated advanced reasoning, multi-step problem-solving, and code execution capabilities within the contest’s five-hour time limit, marking a historic milestone in AI’s ability to outperform humans in high-level programming competitions. Neither model received special training for the contest, underscoring their general reasoning prowess.

Meta Unveiled Ray-Ban and Oakley Smart Glasses

At the event, Meta unveiled its latest launches in the United Arab Emirates, showcasing a diverse lineup of AI-powered eyewear and immersive devices. The Ray-Ban Meta (Gen 2) blends iconic style with smart AI features, supporting prescriptions, Transitions® lenses, and multiple color choices, while the Oakley Meta HSTN delivers a bold, sporty design with hands-free connectivity. The Oakley Meta Vanguard pushes performance further with PRIZM™ lenses, open-ear audio, adaptive volume, and extended battery life, designed for active lifestyles. Breaking new ground, the Ray-Ban Display introduces the first in-lens private display, offering real-time information overlays, built-in cameras, and full Meta AI integration. Alongside eyewear, Meta also introduced the latest Meta Quest VR headsets, accessories, and apps, providing ultra-immersive entertainment, gaming, and productivity experiences. Together, the launches highlight Meta’s vision of merging style, performance, and cutting-edge AI technology into everyday life in the UAE.

Hand Picked Video

OpenAI just released their most comprehensive study ever, analyzing over 1 million conversations from 700 million users worldwide. The findings reveal surprising shifts in how we're actually using AI

Top AI Products from this week

• Aiyouneed - Aiyouneed makes it easy to discover the best AI tools in one place. Whatever your field, find the right tool, save time, and boost productivity. From coding to content creation, design and more—get the latest AI solutions at your fingertips.

• Will AI Take My Job? – Get personalized insights into how AI might impact your career. Analyze your job's automation risk with task-by-task breakdowns and future-ready advice.

• BrandGPT – BrandGPT is your AI co-pilot for branding — build strategy, names, messaging, visuals, and a launch plan in hours, not weeks.

• Pixshop – No layers, no complex tools — just tell Pixshop what you want. Pixshop makes photo editing as simple as chatting with a friend. Just type: “Remove the background”, “Make the sky blue”, or “Smooth my skin” — Pixshop understands and delivers in seconds.

• kit-mcp - Supercharge your AI coding assistant with real-time precise context, deep documentation research, and semantic search. Open source, local, easy to use, From Cased.

• Worktagg – Receive AI-curated business case studies in your inbox daily. Learn from real companies, analyze strategies, and sharpen your business acumen with Worktagg.

This week in AI

• Fluid Benchmarking - Fluid Benchmarking by Ai2 improves LLM evaluation via adaptive item selection using 2-parameter IRT and Fisher information, boosting accuracy, stability, and efficiency

• Elevanlabs Studio 3.0 - ElevenLabs Studio 3.0 is an all-in-one AI editor for creators, offering realistic voiceovers, music, sound effects, noise removal, and video support in one seamless platform.

• AI Agents Milestone - A 2025 Nature paper details AI agents' evolution into autonomous systems with reasoning, memory, tool use, and collaboration, shaping the future of AI interaction.

• World Labs 3D AI - World Labs builds AI that generates interactive 3D worlds from single images, enabling rich spatial intelligence and new creative workflows for games, films, and AR/VR.

• Grok 5 AGI Edge - Elon Musk says Grok 5 might reach AGI, following Grok 4’s top scores on AGI benchmarks like ARC-AGI. Grok leads in technical reasoning, coding, and real-time data use.

Paper of The Day

VCBench is the first benchmark designed to predict startup founder success using anonymized data. It includes 9,000 founder profiles with 92% anonymization to protect identities. Evaluations show several state-of-the-art large language models surpass human investor benchmarks in precision and predictive accuracy. The benchmark uses detailed data cleaning, anonymization, and testing to ensure predictive validity while minimizing re-identification risk. VCBench supports reproducible research with a public leaderboard and aims to evolve with richer features and simulation modes for realistic venture capital decision-making.

To read the whole paper 👉️ here.