An independent look at AI

AI.
Without the
noise.

What is changing in AI, and how it works.
News with context and clear explanations —
for people who want to understand.

Grounded in sources. Free of hype.
HOW IT WORKS / 001INTERACTIVE
INPUTCONNECTIONSOUTPUT
From a single token
to a complete thought
↗
Events & context

In focus

Nine stories worth your attention · November 18, 2026

All stories
MEDIA / OPENAI
Media & tools·November 18, 2026

Le Figaro journalism is coming to ChatGPT

A publisher partnership puts attribution and original reporting in focus.

THE SHORT READ
OPENAI / GPT-6.2
Models·November 17, 2026

GPT-6.2: a new focus on getting work done

OpenAI introduces Astra and Sol for professional tasks and everyday work.

THE SHORT READ
GOOGLE / GEMINI TASKS
Agents·November 17, 2026

Gemini moves from conversation to errands

Google expands Tasks with calling, scheduling and other everyday workflows.

THE SHORT READ
OPENAI / SALES
Tools·November 17, 2026

ChatGPT gets a dedicated sales toolkit

A Sales plugin connects customer research, preparation and follow-up workflows.

THE SHORT READ
OPENAI / SCIENCE
Science·November 16, 2026

More researchers get access to ChatGPT

OpenAI expands its academic programme, with a new application round ahead.

THE SHORT READ
OPENAI / WALLET
Agents·November 13, 2026

ChatGPT Wallet adds a payment step to agent workflows

The launch separates shopping assistance from a user-approved purchase.

THE SHORT READ
GOOGLE / HYDROGEN
Models·November 12, 2026

Gemini 4 Hydrogen targets faster execution

Google positions Hydrogen for coding and everyday agent workflows.

THE SHORT READ
GOOGLE / DECISIONS
Developer tools·November 11, 2026

An AI API for choices, not chat replies

Gemini Decisions scores predefined options across multimodal inputs.

THE SHORT READ
GOOGLE / CARBON
Models·November 10, 2026

Gemini 4 Carbon opens up for longer, harder tasks

Google positions Carbon around sustained reasoning and extended verification.

THE SHORT READ
Go beyond knowing. Understand.

Big ideas.
Plain language.

From the first token to agents and local models.
Twelve explanations, practical examples,
and no advanced maths required.

Explore the learning hub
01How a neural network picks the next wordInteractive · No equations↗02Attention: connecting the wordsExplainer · No equations↗03How noise becomes an imageExplainer · No equations↗04Training and inference: learning, then answeringExplainer · No equations↗05Embeddings: meaning as a map of numbersExplainer · No equations↗06RAG: let the model look things upExplainer · No equations↗07AI agents: when a model can take a next stepExplainer · No equations↗08Reasoning models: spending more work on an answerExplainer · No equations↗09Fine-tuning, prompting or RAG?Explainer · No equations↗10Quantization: fitting a model into less memoryExplainer · No equations↗11Benchmarks: what does a higher score really tell you?Explainer · No equations↗12Context windows: room to read is not perfect memoryExplainer · No equations↗

Curiosity is a good place to start.

We connect the news to the ideas behind it, so you can see what actually changed behind a new model name.

The idea behind NNNews
One conceptToken

A small piece of data a model works with. Not necessarily a whole word.

↗