Urgent.News

What's breaking now, across thousands of outlets.

AI

What Changed When the Same Kata Needed a UI

The next question needed a visible result I began this exploration in From an Empty Repository to a Java Kata, One Module at a Time , following one Codex run from an almost empty repository to a Java project and a String Calculator implementation. Seed4J CLI gave the agent modules it could inspect and combine into the project foundation. I then compared six Codex runs . Their paths through the…

When the same game Kata needed a user interface, the results varied among three different implementations using the same coding agent. All three models satisfied the shared functional requirements, but the path taken and the outcomes differed significantly.

Luna composed a Java application with Spring Boot and Thymeleaf, resulting in a layout and flow distinct from the other two implementations. Terra and Sol both chose TypeScript, React, and Vite, yet they produced different user interface designs and behaviors.

Despite the visual differences, the shared behavior was the baseline for all three implementations. The Seed4J CLI provided a common way for the agents to discover, validate, and apply various module combinations while working towards the same functional target. This experiment demonstrated that even with a single, focused prompt, the agent could effectively navigate the Seed4J ecosystem to produce diverse, functional products.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

The mathematicians rebel against AI

Here is the statement, signed by Terry Tao among many other math notables, most of you probably have read it by now. I do not accept the most cynical interpretations of this proclamation.

  • Terry Tao and other mathematicians sign statement rejecting AI cynicism
  • Tao worries AI may reduce need for human economic thought, but sees benefits
  • Tao expects math funding and interest to rise as mathematicians adapt to AI

Prompt Caching: Why cache_control Writes But Never Reads

I turned on prompt caching for an agent loop that resends a 12K-token system prompt on every turn. Obvious win, right? Input tokens are the whole bill in a tool loop. The bill went up. Not a little.

  • Cache caching mechanism writes but never reads
  • Cache control breakpoint placed after changing content
  • Cache key invalidated by any change at breakpoint

More from Saturday 12 September →