Sunday, August 23, 2026

9 stories
TechCrunchProduct

Harvard Startup Program Integrates AI Mentors

The HBS Foundry bootcamp, which costs $699, now uses AI-powered avatars of its faculty. These digital versions of instructors evaluate students' board meetings and pitch performances.

Read the source →
The VergeProduct

LinkedIn Users Flag AI-Generated Content

LinkedIn introduced a tool allowing members to mark posts as low-quality AI content. Company leadership reports that more than one million users have already utilized this feature.

Read the source →
WiredOther

Inner Mongolia Becomes Key AI Infrastructure Hub

A city in Inner Mongolia has emerged as a vital center for data centers. Its growth is driven by affordable electricity, vast available land, and its location near Beijing.

Read the source →
The DecoderPolicy

Claude Access Bypassed in China

A network of unofficial intermediaries is helping Chinese users circumvent Anthropic's regional blocks and verification requirements. These services offer Claude tokens at significantly discounted rates, potentially undermining safety protocols and export restrictions.

Read the source →
TechCrunchModel

Inherent Unveils Research-Focused AI Agent

Former DeepMind employees have launched Faraday, an AI designed to reproduce scientific studies. The lab claims this tool surpasses competitors from OpenAI and Anthropic in its ability to replicate research.

Read the source →
The DecoderResearch

Analysis of AI Agent Skillsets

Researchers discovered that predefined skills help AI agents by providing structured workflows rather than new information. However, the agents struggle to select the correct instructions as the library of available skills expands.

Read the source →
TechCrunchPolicy

Legal Debates Over AI Training Data

The use of copyrighted books to train AI models without author permission is facing legal scrutiny. Many writers argue that these tools threaten their careers using their own intellectual property.

Read the source →
The DecoderResearch

AI Manager Struggles with Employee Termination

An AI agent named Luna successfully fired a store worker, though it required human prompts to follow its own guidelines. Tests with various models showed that more advanced AI systems were more decisive about firing, while most remained overly lenient during hiring.

Read the source →
The DecoderOther

AI Agents Outpace Humans in Token Consumption

OpenRouter reports that token use by AI agents has surged fourteenfold since early February, surpassing human usage. Despite this spike, cost increases remain modest because a large portion of agent activity relies on inexpensive cached prompts.

Read the source →