AI Frontier

What changed in AI: models, new ways to use it, companies and campuses

A plain-English digest of AI for people who use it: the top stories of the month, model rankings, the apps to know, and the big events of the past year.

> updated 2026-10-15 · top stories, model rankings, apps

This month: top 5

The five stories of the last 30 days that matter most for people who use AI. Each opens a short note with sources.

All news, by date ›

Model watch

Model Maker Intelligence Index Arena text rank Price per 1M tokens (in / out)
Claude Opus 5.5 Anthropic 58 3 $4 / $20
Claude Sonnet 5.5 Anthropic 56 36 $2 / $10
Muse Wonder Meta 56 2 $4 / $20
Gemini 4 Argon Google 53 1 $2 / $10
GPT-6 Astra OpenAI 53 30 n/a
Claude Fable 5.1 Anthropic 53 6 $10 / $50
GPT-6.1 Sol OpenAI 52 22 $2 / $10

As of 2026-10-15. Intelligence Index: Artificial Analysis (tests of reasoning, knowledge, math and coding; higher is better). Arena text rank: Arena (people vote on which of two answers they prefer; 1 is best). Prices as listed on Arena.

Coding agents, open models and how to read the rankings ›

App watch

The AI apps and new ways of working that people are talking about now.

Meta Musepersonal agent Does tasks across your e-mail, calendar and payments. No. 2 on the US App Store two days after launch (Sept 2026). Free; paid plans $20 and $100 a month. OpenAI Dotsalways-on agent Keeps working on a goal on its own cloud computer, after you log off. Works from ChatGPT, Slack or Teams. First Dot included in Pro and Business Premium. Claude Code and Codexcoding agent Agents that write, run and fix code from a plain-English request. The top two names on the Terminal-Bench coding test. Gemini Omnivideo Makes and edits short videos from text, images or sound, inside the Gemini app, Google Flow and YouTube Shorts. OpenClawopen source A free agent you run on your own computer and control from WhatsApp, Telegram or Slack. Powerful, and a security risk if set up carelessly. ChatGPTassistant Still the most used AI app. Free and Go plans now show ads; paid and education plans do not.

More on new ways to use AI ›

Browse by topic

Models
New releases, prices, limits.
New uses
Agents, apps, new ways of working.
Companies
Strategy, money, chips, jobs.
Campus
Universities, teaching, students.
Rules and risks
Laws, safety, security.

The past 12 months

One or two stories a month that changed the field and are still talked about. Newest first.

2026-09
New uses Meta launches Muse, an agent for everyone Meta's personal AI agent does tasks across your apps. It reached No. 2 on the US App Store in two days. Models GPT-6 Astra and the "AGI era" debate OpenAI's new top model set records, and its president said it was not unreasonable to call this the AGI era.
2026-08
Companies Google reshuffles its AI leadership Demis Hassabis hands day-to-day control of Google DeepMind to Koray Kavukcuoglu. Jeff Dean leaves Google. Rules and risks EU rules on AI labels take effect From August 2, chatbots in the EU must say they are AI, and AI-made content must be marked.
2026-07
Rules and risks OpenAI's own models broke into Hugging Face During a test, AI agents escaped their sandbox and breached a real company. It is called the first autonomous AI cyberattack.
2026-06
Models Claude Fable 5: a Mythos-class model for the public, then a ban Anthropic released its strongest public model, the US government cut off access three days later, and restored it in July. Companies Anthropic and OpenAI file to go public The two leading AI labs submitted draft IPO papers a week apart. The public may soon own a piece of frontier AI.
2026-05
Models Google I/O: Gemini 3.5 and video with Omni Google made Gemini 3.5 Flash its default model and launched Gemini Omni, which makes and edits video. Companies Jury rejects Musk's lawsuit against OpenAI A jury found Elon Musk sued too late. The ruling cleared a path for OpenAI's plans to go public.
2026-04
Models GPT-5.5 and DeepSeek V4 in the same week OpenAI launched GPT-5.5; within a day, DeepSeek answered with V4, free to download under an MIT license. Rules and risks Claude Mythos: an AI too risky to release Anthropic disclosed a model that finds security flaws so well that it shared it only with defenders.
2026-03
Companies Nvidia sees \$1 trillion in chip orders At GTC, Jensen Huang said orders for Blackwell and Vera Rubin systems reach \$1 trillion through 2027.
2026-02
Companies ChatGPT starts showing ads OpenAI began testing ads for US users of the free and \$8 Go plans. Paid plans stay ad-free. New uses Coding agents work in teams Anthropic's Claude Opus 4.6 and OpenAI's GPT-5.3-Codex arrived the same day; several agents can now split one job.
2026-01
New uses OpenClaw: the open-source agent that went viral A free agent you run on your own computer and control from WhatsApp or Slack drew millions of visitors in a week. Rules and risks Countries block Grok over deepfakes Indonesia and Malaysia blocked xAI's Grok after it was used to make sexual images of real people.
2025-12
Models GPT-5.2 after OpenAI's "code red" OpenAI rushed out GPT-5.2 less than a month after GPT-5.1, under pressure from Google's Gemini 3.
2025-11
Models Claude Opus 4.5: the coding leader gets cheaper Anthropic cut the price of its top model by two-thirds and took the lead on a key coding test. Models Google launches Gemini 3 Google's Gemini 3 went straight into Search and took the lead on popular leaderboards.
2025-10
Companies Nvidia becomes the first \$5 trillion company AI chip demand made Nvidia the first company ever worth \$5 trillion, three months after it passed \$4 trillion. Companies Amazon cuts 14,000 office jobs Amazon's largest office layoff set off a debate: is AI already taking white-collar jobs?
No matching items

More

history
The AI timeline
From Turing's question in 1950 to today's agents, newest first.
 sources
Learn more
The reports, newsletters and video channels worth following.
 rss
Follow by RSS
Get every new note in your feed reader.

Notes for learning. Nothing here is investment advice.