ai

Vibe – Local Audio & Video Transcription App

, , ,

Vibe is a free, open-source transcription app that runs entirely on your device, ensuring audio and video files never leave your machine for privacy. It supports nearly 100 languages using Whisper models, offers batch processing for folders, and exports transcripts in multiple formats such as SRT, VTT, PDF, and plain text. Users praise Vibe for its offline operation, speaker labeling, and ease of editing, making it a practical tool for private, automated transcription workflows.

https://thewh1teagle.github.io/vibe/

Codex Tool Reference

,

This reference details a comprehensive inventory of 232 callable tool interfaces and 44 skill definitions available in a Codex Work session as of August 31, 2026. Tools cover diverse categories such as runtime and file operations, browser control, data analytics, and automation, while skills package reusable instructions for tasks like document editing, image generation, data visualization, and plugin management. The repository enables integrated orchestration and execution of complex workflows through detailed TypeScript interface declarations and official usage descriptions.

https://codex-tool-reference.simonw.chatgpt.site/

Understanding ChatGPT Work

,

OpenAI's ChatGPT Work is a premium feature available to $20/month+ subscribers that offers advanced capabilities beyond regular ChatGPT, including model selection with GPT-5.6 variants, a code execution environment with internet access, a full headless Chrome browser for web interaction, persistent shared filesystems across sessions, and tools for building and deploying ChatGPT Sites. These features enable complex workflows such as automating tasks, running parallel sub-agents, and integrating real-time web data, making ChatGPT Work powerful but also raising questions about safety and clarity in OpenAI’s documentation.

https://simonwillison.net/2026/Aug/30/understanding-chatgpt-work/

LLM Counter-Defaults

, ,

LLM Counter-Defaults is a tool by AIxDESIGN that lets users customize their large language model's default behaviors by adjusting sliders across multiple dimensions such as confidence, agreeableness, and task focus. Users set their preferences once, generating tailored instructions that guide the LLM to follow these rules in all subsequent chats, effectively overriding factory settings to better fit individual needs and values. The project, created by Nadia Piet and built with Claude Code, offers an interface to fine-tune the LLM's style and output to promote more personalized interactions.

https://counterdefaults.aixdesign.co/

Keep AI Out of Your (Obsidian) Vault

, ,

The article argues against heavily integrating AI-generated content directly into your Obsidian vault, emphasizing that AI summaries often dilute personal insights and reduce clarity over time. Instead, it recommends maintaining a human-curated knowledge system where note connections reflect your own thinking, using AI selectively for tasks like advanced research or separate vaults for AI-generated material. This approach preserves the value of your “second brain” as a meaningful, personalized repository rather than a generic AI-driven database.

https://www.ssp.sh/brain/using-obsidian-with-ai/

Scroll to Top