Urgent.News

What's breaking now, across thousands of outlets.

AI

Harness Engineer: How a centralized knowledge base (RAG) can positively impact your company

Imagina um novo colaborador quando chega na empresa. A barreira mais comum entre todas é o ramp-up de negócio. Regras, fronteiras, o que o código cobra e não está no diff. Quem chegou agora não carrega isso na cabeça. A regra só aparece quando alguém lembra, ou quando o erro volta. Aí ele abre só o livro que o PR deixou na mesa. Lê com atenção. Página certa, capítulo certo, nota de rodapé ok.…

Translated from Portuguese Read in Portuguese

A company has implemented a system to reduce the "ramp-up" time for new employees to understand its business rules. The system, which uses Git, involves creating a repository of "cards" or "fichas" that document specific business rules and are linked to specific files in the company's code. When a new employee reviews a piece of code, they can access the relevant card and learn about any business rules that may not be immediately apparent from the code itself.

This approach, called RAG, has helped the company reduce errors and improve code reviews. The system has been shown to be effective in preventing repeated mistakes, with 194 cards already available for code review and 116 of them related to sensitive areas such as money, access, or personal data.

Written by urgent.news from Dev.to's report — not a translation of it. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

Julia 1: A 144M-Parameter Decision Model Trained for $104

A small AI lab in Brazil just released a model with an unusual claim. Not "beats GPT." Not "agentic superintelligence." The claim is this: a 144.3-million-parameter model that picks the right answer…

  • Julia 1 is a 144.3-million-parameter decision model
  • Trained for $104 on CPU, tablet, or low-end chip
  • Designed to classify answers from a list of options

How We Made an LLM Actually Use Recalled Memory

Making Recalled Memory Actually Influence LLM Recommendations When we integrated Hindsight into PayEcho, retrieving a customer's history was not the difficult part.

  • PayEcho struggled to make LLM use recalled memory for recommendations.
  • Model could see recalled information but produced generic recommendations.
  • Connecting retrieval to reasoning made memory part of the reasoning process.

More from Sunday 27 September →