AI agents still need humans to teach them

AI agents need skills — specific procedural knowledge — to perform tasks well, but they can’t teach themselves, a new research suggests. The authors of the research have developed a new benchmark, SkillsBench, which evaluates agentic AI performance on 84 tasks across 11 domains including healthcare, manufacturing, cybersecurity and software engineering.  The researchers looked at…

Read More

AWS adds Design-first and Bugfix workflows to Kiro

AWS is recognizing that most developers don’t work the way Kiro, its Visual Studio Code–based agentic IDE, forces them too — so it’s adding two new software development workflows to Kiro that meet developers where they are: working on existing projects, fixing bugs. Kiro started out with a vision of helping developers through a process…

Read More

ThoughtSpot Pushes Upstream With Agentic Data Preparation

Agentic analytics platform ThoughtSpot launched its next generation of Analyst Studio, which it claims offers a “new suite of capabilities to revolutionize how data teams deliver AI-ready data with speed, flexibility and control.” The analytics market has spent the last decade obsessing over usability. They wanted faster dashboards and queries, while keeping the interface easy…

Read More

Python’s popularity slip: Here’s what we know

Python faces new challenges from old rivals, but is it a blip or something more? In other news, we have goodies like easy-install instances of PostgreSQL, a peek at the state of the art in visual development for generative AI projects in Python, and sneaky tricks to share Python tooling across local projects. Top picks…

Read More