An orientation for curious humans
Start here, Earthling.
AI moves quickly. Understanding it should not require keeping up with every announcement.
We follow intelligent systems in digital and physical worlds. Choose a topic that matters to you, then look at the evidence behind the headline.
Choose your signal
AI & work
Jobs, skills, and what changes when software can take on more tasks. Begin with our guide to reading the evidence about graduate employment.
Is AI Taking Graduate Jobs? What the Evidence Actually Shows →Robotics & autonomy
Intelligence meets the physical world. Explore the difference between a demonstration, a supervised service, and a system operating on its own.
Robotaxis in London: What’s Really Driving Itself (and What Isn’t) →AI in practice
Applications, experiments, and the limits that can disappear from the headline. Start with how to interpret a comparison between AI and human experts.
Microsoft’s AI Beat Doctors at Diagnosis. Here’s the Fine Print. →A better way to read an AI claim
- What was actually tested? A controlled benchmark, a demonstration, or everyday use?
- What was the comparison? Check the tools, time, information, and support available to each side.
- What are the limits? Look for failures, missing cases, and uncertainty as well as successes.
- Where is the evidence? Follow the original study, documentation, or dataset when it is available.
These questions guide how we separate demonstrated capabilities from predictions and open questions.
Six terms worth knowing
- Machine learning
- Methods that learn patterns from examples rather than relying only on rules written by people.
- Model
- The trained system used to make predictions or produce outputs.
- Training
- The process of adjusting a model using data and feedback.
- Inference
- Using a trained model to produce an output from a new input.
- Benchmark
- A defined set of tests for comparing performance. A result describes those tests; it does not settle every question about real-world usefulness.
- Hallucination
- An AI-generated answer that presents incorrect or invented information as if it were factual.
Explore more terminology in the Stanford HAI AI Glossary and Google’s Machine Learning Glossary.
Keep exploring
- Stanford HAI AI Glossary →
A reference for unfamiliar AI terms. - Google’s Machine Learning Crash Course →
A more technical next step, with videos, visual explanations, and practice exercises. Check its prerequisites before starting. - Stanford AI Index →
Reports and data for putting AI developments into a wider context.
External learning resources checked 31 October 2026.
Receive the next transmission
One weekly briefing on what changed in AI, what the evidence shows, and why it matters.
Get the weekly briefing →We come in peace. We bring sources.