// a daily digest from the AI world

AI news
August 22, 2026

New models, tools and announcements — every day, always with a source link. This digest is assembled by an AI agent following my rules, and it's the same briefing I use myself to keep up with AI without burning time.

announcementThe Verge

Google Discover gets an AI chatbot-tuned feed ↗

Google will soon let you customize your Discover feed by describing what you want to see. The feature rolls out in the Google app in the coming days — AI automatically tweaks the feed based on your description and remembers your preferences for future visits. You will find it in the three-dot menu.

announcementTechCrunch

OpenAI is gaining on Anthropic among business users ↗

New data indicates OpenAI is closing the gap with Anthropic among business users. Companies are willing to flop back and forth depending on which lab has just shipped a new model — a volatility that shows how far from sticky enterprise AI spending still is.

toolSimon Willison

llm-openrouter 0.7 adds reasoning traces and server-side tools ↗

The llm-openrouter 0.7 plugin is compatible with LLM 0.32 and can display reasoning traces for models available through OpenRouter. Models now use OpenRouter's implementation of the Responses API, and three new server-side tools arrive — Shell, WebFetch and WebSearch, which you enable with options like -T WebSearch.

announcementThe Verge

Over a million people have clicked LinkedIn's AI slop button ↗

Late in July LinkedIn added a “Seems like AI slop” button, and the company says over a million people have already clicked it. You will find it in the three-dot menu on a post. It shows how much users care about feeds being flooded with generated content.

announcementSimon Willison

ChatGPT search now uses the site: operator at scale ↗

ChatGPT search has started using the site: operator at scale. Around it, the GEO space (Generative Engine Optimization) is growing — a chatbot version of SEO, where tools like Promptwatch automatically track how brands show up in replies across chat products.

researchTechCrunch

Nvidia shows the harness, not the model, is the real hero ↗

Nvidia research shows that AI agents can perform well, and not go off the deep end, through fine-tuning — even when the underlying model is not that great at the task. The center of gravity shifts from the model to the harness around it.

← All editions · RSS

I build exactly this kind of automation

This section writes, assembles and publishes itself every morning — from source selection to deployment. If you'd like a quiet helper like this in your company, let's start with an initial consultation — the fee is credited toward the project if we work together.