A guide to afraid2feel.

Interactive experiments about AI and human skills, created by claudiu.space. Explore a world, inspect a result, or follow a source.

Walk — LATENT

Turn a conversation into a walkable 3D world of words and connections. LATENT is an interactive artwork, not a view inside a model.

Open Walk — LATENT

Learn

Explore a map of sources about artificial intelligence, from how models work to privacy, safety, ethics and energy.

Open Learn

Fear

Explore fear-related Wikipedia page views and GDELT news coverage. These charts measure attention, not how frightened people feel.

Open Fear

Keep

Explore research on memory, judgement, thinking and collaboration through an interactive bookshelf. Open a finding and follow its source.

Open Keep

Race

Compare human and AI results study by study. Explore timelines, evaluation methods, baselines and limitations alongside original sources.

Open Race

Play — Overprint

Play a browser pinball game whose ball trails become an artwork. Export the result as a print.

Open Play — Overprint

Watch — London in Motion

Explore London through transport, weather and city data in an interactive 3D map. Check the live, modelled and unavailable labels for each layer.

Open Watch — London in Motion

INHERIT

Inspect an illustration of research on AI trait transmission through training data. Provenance labels distinguish published findings from illustrative content.

Open INHERIT

Read the evidence alongside the visuals.

Race and Keep link findings to their sources. A benchmark result describes a particular test; it is not a general measure of intelligence or a prediction about jobs.

Fear uses Wikimedia page-view data and GDELT news data. Its date is the latest shared day available, which may lag behind today. The live edit counter observes Wikimedia events in your browser and can miss events during a disconnection.

INHERIT refers to Subliminal Learning, Cloud, Le et al. (2025). It labels published findings and illustrative content separately. The site does not train a model, and displaying numbers does not reproduce the reported fine-tuning effect.

Questions or corrections: info@claudiu.space.