Urgent.News

What's breaking now, across thousands of outlets.

AI

Fan Uses AI to Fix Oblivion Remastered PC Performance

The Elder Scrolls IV: Oblivion Remastered was excellent when it came out, bringing the beloved RPG to the modern era, but just like many other games running on Unreal 5, it was a technical mess. However, a fan has used Claude to create a performance mod that improves the game's performance.

Fan Uses AI to Fix Oblivion Remastered PC Performance

An Elder Scrolls IV: Oblivion Remastered fan has utilized Claude to craft a performance-enhancing mod, addressing the technical issues plaguing the game. Released in April 2025, the game was already notorious for its performance woes, which persisted even after patch 1.2 in June 2025. Despite improvements, significant frame rate drops and stuttering remained a persistent issue.

This led to a community effort to address the problem. Reddit user LTSharp showcased a side-by-side comparison of the unpatched and patched versions, demonstrating a marked reduction in hitches. The root of the problem lay in the game's continuous reconstruction of lighting for each frame, placing an undue strain on the CPU. To tackle this, LTSharp employed Claude Opus 5.5, instructing it to pinpoint the issues.

Following two weeks of meticulous work, multiple patches were developed. The most substantial improvement came from distributing processing tasks across more CPU cores, resulting in a noticeable rise in CPU utilization – from a meager 35-40% to a more robust 50-60%. With these enhancements, the game's performance has seen an overall boost, offering a smoother experience to players.

This endeavor showcases the potential of AI in software development, even for those without formal technical backgrounds.

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

Read the original at ign.com →

More in AI

Five Parallel AI Builders, One Contract, 25 Review Fixes

On September 26 I had five builder agents work from one markdown contract that said which files were theirs. When P0, the first local version, was done, it ran at 127.0.0.1 with 224 passing tests and…

  • Five AI builders worked on one markdown contract for a project
  • First local version passed 224 tests with no outgoing requests
  • 25 fixes identified during review process

LazyDay AI: Less Overthinking, More Tiny Wins — A Slow & Steady Approach to Productivity

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady What I Built 🦥 LazyDay AI — Do less. Stress less. Move forward, one tiny step at a time.

  • LazyDay AI simplifies productivity with small, manageable steps
  • Features include brain dump, energy-based planning, and tiny tasks
  • Future plans involve integrating open-source AI for personalization

Fork, Fix, Delete: Safe Database Migrations with GitHub Copilot and Tiger Cloud

What I Built A safe production workflow for fixing slow time-series queries without ever touching prod. The problem: Direct fixes on Postgres are risky.

  • GitHub Copilot uses FORK step to clone live service database into test environment
  • FIX step involves reproducing slow query, converting table to hypertable, and adding indexes
  • DELETE step removes fork to keep production database untouched and ready for use

More from Tuesday 20 October →