Urgent.News

What's breaking now, across thousands of outlets.

Saturday, 3 October 2026 Reading Room

AI

Haven: A Private, Voice-First AI Companion for a Friend in Need

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend What I Built I built Haven — an empathetic, voice-first AI companion designed for a close friend who often navigates heavy personal and family matters late at night. When dealing with sensitive issues, people often hesitate to reach out to acquaintances because they fear being judged, feeling like an emotional…

dev.to ·

Haven is a voice-first AI companion designed for a close friend navigating heavy personal and family issues late at night. It provides an empathetic, non-judgmental presence through an intimate phone call-like experience. Users can simply start a call and Haven will respond with a comforting "Hey, I'm right here. How are you feeling right now?"

The AI can either listen or help think through solutions together. It uses natural language mirroring in multiple languages, including Hindi and Hinglish. The AI is built with privacy as its top priority, running completely offline on local hardware using Google Gemma 2 (2B) via Ollama. It also includes a fallback to connect with remote OpenAI/Hugging Face endpoints if needed. The open-source code is available on GitHub.

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

Fridge Oracle: I Built My Friend a Recipe Helper That Never Leaves the Laptop

This is my entry for the Hacktoberfest Weekend Challenge: Build for a Friend. Demo This is what it looks like start to finish, typed into the real app on my own laptop.

  • Fridge Oracle app helps users cook with fridge contents while respecting dietary restrictions.
  • Uses Google's Gemma 3 model locally via Ollama to ensure privacy and offline use.

·dev.to

Jev as a Tool Router: Cutting Agent Cost Without Killing the Investigation

By now, you have probably heard about Jev, a System 1 model that has been getting a lot of attention lately. In simple terms, a System 1 model is built for fast, cheap, bounded decisions, while a…

  • Jev + Kimi maintained golden pass scores across all catalog sizes
  • Jev + Kimi setup cost significantly lower than Astra at all sizes
  • Jev required fewer hops than Astra full-menu option for tool selection

·dev.to

Building an Open-Source, Multi-Agent Study Companion for My Sister

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend What I Built I built Bud AI (Buddy AI) —an open-source, multi-agent educational assistant and interactive study…

  • Bud AI is an open-source, multi-agent educational assistant for college students and their sister.
  • Addresses context window overhead, emotional tracking, and on-demand tool discovery.
  • Live app accessible at bud-ai-rho.vercel.app with code repository at jouzia/Bud-AI.

·dev.to

Why a Successful Agent Transaction Can Still Fail Authorization Checks

An AI agent submits a transaction. It executes without reverting, and the dashboard displays “Success.” That label leaves an important question unanswered: did the agent execute the call the user…

  • Transaction appears successful on dashboard but may not be authorized by user.
  • Evidence validity, execution outcome, and authorization compliance must be reviewed separately.

·dev.to

More from Saturday 3 October →