ENFR
8news

Tech • IA • Crypto

TodayVideosVideo recapsArticlesTop articlesArchives

ChatGPT Ads CPC Shift, Google Cosmo Leak, Gemini Omni Revealed

AIMonday, May 4, 2026· 10 videos

Briefing

0:00 / 0:00

ChatGPT ads shift to CPC

OpenAI is expanding ads inside ChatGPT, moving from CPM to a CPC model ($3–$5 per click). The rollout spans the US, Canada, Australia, and New Zealand, with free users primarily affected. Targeting uses conversation context, memory, and location, raising privacy questions. The move aligns ChatGPT with Google Ads–style performance marketing and signals a major monetization push.

Google Cosmo app briefly leaks

Google accidentally published then pulled an experimental Cosmo assistant app on Android. It combines Gemini Nano (on-device) with cloud inference in a hybrid system. The app could read on-screen content via accessibility APIs, hinting at deep OS-level integration. Its unfinished state suggests an early test of next-gen assistants.

Gemini Omni unifies media models

Gemini Omni appears to merge video (Veo) and image systems into a single multimodal model. Early references suggest tighter integration across generation tools ahead of Google I/O 2026. This would simplify workflows and strengthen Google’s position against rivals in generative media. It also signals a shift toward unified AI stacks rather than fragmented models.

ChatGPT Workspace Agents automate tasks

OpenAI introduced ChatGPT Workspace Agents capable of executing multi-step workflows autonomously. Using MCP connectors, they integrate with tools like Gmail, Google Drive, and Calendar. Agents can retrieve data, make decisions, and act without constant user input. This positions ChatGPT as an operational layer inside business processes.

YouTube tests Gemini conversational search

YouTube is experimenting with an “Ask” feature powered by Gemini. Instead of browsing thumbnails, users interact via natural language queries to surface clips and videos. Recommendations become intent-driven, not just algorithmic history. This could significantly reshape discovery and visibility for creators.

Claude adds Higsfield media generation

Claude Co-work now integrates a Higsfield connector for native image and video generation. The update removes prior sandbox limits, enabling access to external creative tools within workflows. Users can manage local project folders and automate content pipelines. This expands Claude into a more complete production environment.

Hassabis pushes scientific AI vision

Demis Hassabis advocates for AI focused on irreversible scientific breakthroughs rather than fast-moving products. He contrasts this with the current race led by OpenAI and Anthropic, where models are quickly replicated. The argument highlights inefficiencies, including massive compute costs versus the brain’s ~20-watt efficiency. It reframes AI progress as a long-term scientific endeavor.

Robotics scale across China and Japan

China is piloting Agibot robots to direct traffic, embedding automation into urban systems. In Japan, robots are handling baggage at Tokyo Haneda Airport, addressing labor shortages. These deployments reflect demographic and infrastructure pressures driving adoption. The trend raises trade-offs between efficiency gains and workforce disruption.

Videos covered

Previous briefings · AI