Generative AI Tools & Use

Offrun, an integrated workspace tool for running and monitoring multiple coding agents in parallel and checking account balances, has been released.

• •It allows running Claude Code, Codex, AGY, and Grok Build in parallel.
• •Users can check which agent is working and whether their intervention is needed.
• •It features a function to check the remaining balance of each account.

It was announced that the developer conference DevCuro will be held in Tokyo on November 25.

GitHub Copilot has reportedly added a "computer use" feature, enabling the operation of desktop applications.

Meta open-sourced code to allow users to create custom gadgets featuring its new AI agent, Muse.

• •Meta open-sourced the code to enable users to create their own Muse gadgets equipped with the company's new AI agent, Muse.
• •Meta has proposed project examples, such as loading Muse onto a color E-Ink display to show reminders or integrating it into an HDMI stick for large-screen display.

An open-source tool that uses ChatGPT or Claude to generate LDraw source code for Lego models and convert it into CAD models.

• •ChatGPT or Claude generates source code describing Lego assembly instructions in the LDraw language.
• •LDraw is a low-level language for describing Lego part placement, executable in tools like LDView or LeoCAD.
• •The generated code is converted into an interactive Lego CAD model that can be manipulated and modified.

An explainer covered ds4, a C-language inference engine by Redis founder antirez, including documentation and benchmarks for running models like DeepSeek V4 locally.

• •ds4 is a C-language inference engine supporting DeepSeek V4/V4.1, Qwen3.8 Flash Next, and GLM 5.x.
• •It supports operation on different hardware environments such as Metal, CUDA, and ROCm.
• •Documentation, benchmark results, and setup instructions for ds4 have been published.

A report on using GLM 5.3 Flash for coding for one month detailed 2 billion tokens of usage and factors affecting model selection and costs.

• •There was a total of 2 billion tokens of AI usage during one month of coding.
• •The report analyzed elements influencing model selection criteria and cost structures.

An article explains how developers should strengthen AI instruction, critical review of outputs, and retention of technical judgment as AI agent adoption grows.

• •The importance of the ability to effectively instruct AI agents is highlighted.
• •Skills for critically reviewing and verifying AI-generated outputs are required.
• •Placing technical judgment at the center of workflows is emphasized.

Matthew Schwartz discussed changing his problem-solving approach with Claude to develop BootLoops, a rigorous calculation toolkit for quantitative science.

• •He shifted from opposing Claude to finding 'Claude-shaped' problems that Claude excels at.
• •He built a toolkit called BootLoops for rigorous calculations in quantitative science.
• •The toolkit is being applied across multiple scientific fields in collaboration with experts.

Earendil released version 1.0 of its AI agent harness, Pi, adding features to save and resume tasks.

• •Earendil released version 1.0 of the AI agent harness, Pi, on October 1, 2026.
• •Pi Durable, a feature enabling task saving and resumption, was also released simultaneously.