General AI News

LM Studio v0.3.17 introduces Model Context Protocol support for connecting local and remote MCP servers.

Users can add MCP servers by editing the mcp.json file or via the new “Add to LM Studio” deeplink button.

Tool call confirmations let users review, edit, and approve tool invocation arguments with one-time or always-allow options.

Iroh provides a simple API to connect peers by public key and maintains the fastest route automatically

It uses QUIC (via the Quinn library) for encrypted, multiplexed, datagram-capable connections

Iroh attempts NAT hole-punching first and falls back to public relay servers when needed

Claude now supports building, hosting, and sharing interactive AI-powered apps directly within the Claude app.

When users access your app, they authenticate with their own Claude account and consume their own subscription—creators incur no usage costs.

Claude generates real, modifiable code to orchestrate AI functionality with no need to manage API keys or deployment.

A Columbia Business School study found Uber’s upfront pricing algorithm raised rider fares and cut driver pay on billions of US trips.

The US study showed Uber’s take rate climbed from about 32% to over 42% by the end of 2024.

A University of Oxford study in the UK found dynamic pricing increased the median take rate from 25% to 29%, with some trips exceeding 50%.

Pursue problems you can genuinely solve and contribute to, regardless of perceived prestige.

Start with simpler or ‘humble’ problems to gain success, satisfaction, and help others.

Feynman regrets assigning a grand problem rather than letting the student choose his own.

You can speed up audio (2x or 3x) with ffmpeg before sending it to OpenAI’s transcription API to reduce duration, tokens, cost, and wait time with minimal quality loss.

A simple workflow: extract audio with yt-dlp, apply atempo filter in ffmpeg, then send the faster audio to gpt-4o-transcribe and summarize with an LLM.

At 2x speed a 40-minute file becomes ~20 minutes, cutting input token cost roughly in half; at 3x (~13 minutes) it cuts cost by ~33%.

Scream to Unlock is a Chrome extension that blocks social media until the user shouts an embarrassing phrase (“I’m a loser”), granting more access time for louder screams.

The tool is open-source, processes audio locally, and claims no data collection or recordings.

Commenters raised concerns about negative self-talk and suggested positive or neutral phrases, alternative deterrents like math problems, staring challenges, or disturbing images.

83% of U.S. adults did not pay for news in the past year; 17% did by subscribing, donating, or membership.

74% of Americans encounter paywalls at least sometimes, including 38% who do so often.

Upon hitting a paywall, 53% look for the information elsewhere, 32% give up, and only 1% pay for access.

Gemini CLI is an open-source (Apache 2.0) AI agent that brings Gemini 2.5 Pro capabilities directly into the terminal.

Free tier offers access to Gemini 2.5 Pro with a 1 million-token context window, 60 requests/minute, and 1,000 requests/day at no charge.

Supports coding tasks (writing, debugging) and non-coding workflows like content generation, research, and task automation via natural language.

Modular design with 12 replaceable parts for easy repairs and upgrades

5-year warranty and 8-year Android software support

Durable build featuring Gorilla Glass 7i, IP55 rainproofing, and military-grade drop testing