
GEN-1.5 Learns From 12 Seconds—but Succeeds Only 59% on Average
Generalist says GEN-1.5 can attempt a robot task after one 3–12-second demonstration, but its company-run evaluation averaged 59% success across ten short tasks.

Generalist says GEN-1.5 can attempt a robot task after one 3–12-second demonstration, but its company-run evaluation averaged 59% success across ten short tasks.

The Aug. 11 GitSkills preprint counts 3.8 million skill-file occurrences in public GitHub repositories, with 50.5% identified as byte-for-byte copies.

AI agents make business sense when they work inside a real process, not as a clever chat demo. At Selleo, we treat an agent as part of the product and the workflow around it, not as a standalone feature built to look good in a meeting. That changes the whole conversation. A demo proves that a model can respond. A production-ready agent proves that it can use data, follow rules, handle approvals, and keep working after launch without creating chaos for the team.

The past week has been explosive for AI advancements, with major players dropping frontier-level updates focused on agentic capabilities, coding, video generation, and game development. Here's our Top-5 highlights of the week from the world of AI, curated by Quasa.

Artificial intelligence is transforming property search in Dubai's real estate market with AI tools, virtual tours, and predictive analytics.