Neura News

AI News

News reporting focused on AI and machine learning, covering the companies behind these technologies, their real-world applications, and the ethical concerns they raise. This includes areas like generative AI (large language models, text-to-image and video), speech tech, and predictive analytics.

Latest News

4 articles

Developer

MCP Drops Sessions, Adds Mandatory Headers, and Reignites the REST Debate

The Model Context Protocol (MCP) released its 2026-07-28 specification, removing protocol sessions and adding mandatory Mcp-Method and Mcp-Name headers, making the protocol stateless. This allows gateways to route and throttle agent traffic without parsing JSON bodies. The update has sparked developer debate over whether MCP has essentially become just another REST API, with supporters citing operational benefits and critics questioning its novelty.

Aug 126 minNeura News

Developer

condense-json 1.0 Shrinks Repeated JSON Data With Reference Syntax

Simon Willison released version 1.0 of condense-json, a Python library that reduces JSON size by replacing repeated substrings with a compact reference syntax. The library, about 18 months old, includes non-disruptive fixes and is designed for storing logs and caches with duplicated data. It integrates with Willison's LLM tool to save space in SQLite logs.

Aug 33 minNeura News

AI Models

Simulating a Woodland Economy with a 3B Model

A developer built Thousand Token Wood for the Build Small Hackathon, a multi-agent simulation where five woodland creatures trade goods, gossip, and hoard using a Qwen2.5-3B model. The article details how engineered scarcity, smarter prompting, and historical market legends brought the tiny economy to life. Key results include 100% valid JSON output, sustained trades, price crashes, and a widening wealth gap.

Jun 55 minNeura News

AI Models

Microsoft VibeVoice Speech-to-Text Model Tested

Microsoft's VibeVoice offers speech-to-text transcription with speaker diarization under an MIT license. Developer Simon Willison ran it on a Mac using a quantized MLX version, processing an hour of podcast audio in under nine minutes. The model outputs detailed JSON with timestamps and speaker labels, though it limits input to one hour.

Apr 283 minNeura News