The Mind Reader Interface

Ghost in the Lines

Ghost in the Lines is an ideation companion for creative practitioners. It builds on what a creator already does when mind-mapping: brainstorming in keywords, then combining and connecting them. As those lines take shape, the AI follows them and offers ghost suggestions from the context of the map and the strokes that grow from it. It supports open, serendipitous exploration while helping a design direction converge.

The pen stroke itself is the instruction. A line drawn out from a word asks for a leap, and its length sets how far that leap reaches. A loop around a cluster asks what those fragments fuse into. The drawing carries the intent. To preserve the agency of the creator, a word written by hand where a ghost would land takes priority over the suggestion. As the board fills, a quiet background agent reads the whole canvas and articulates one or more design directions.

Par Shinyoung Park

NOMAD

This project question our sense of orientation. It aims to go back to screenless navigation and reduce that habit of pulling out our phone when we’re lost. Can we navigate through a city without constantly looking at a screen?
 

Par Emilie Müller

DON'T

When you have body-focused repetitive behavior such as skin picking, nails biting or hair pulling, it often happens unconsciously. It starts with placing your hand on your cheek and ends with bleeding cuticles. 

This smartwatch anticipates an episode by emitting vibrations that help you become aware when your wrist is in a specific (predefined) position and small movements are detected. The vibrations intensify if the undesirable behaviour is detected continuously. Up to three positions can be registered as "unwanted".

Par Gwenaëlle Gustin

Reckoner

Reckoner is a note taking and comparison tool. It is a vocal-driven software working on mobile with barely any UI. The goal is to accompany or replace on-paper annotations and let the reader speak their mind as they're reading while minimising the digital presence and distractions. The software is built to detect the intent of any entry, whether it is a rhetorical question, a comparison with old notes, a quote, and so on. It works with a queuing system so that the tool is always available, even during processing. Reckoner removes the manual transcription step to build a digital collection of annotations and uses cross-referencing to suggest links between the existing notes upon processing. The project works with an open-source LLM run locally and edits markdown files so that it can be plugged into different note-taking software. In this version, Obsidian is used for its graph view and links visualisation.
 

Par Thomas Neyroud

Semantic

Text on the web is static, whereas reading preferences are dynamic. One person prefers to skim an article or want a gist, while another might be reading to improve their vocabulary or another is learning the language. Each brings a different intent with interacting with text on the web. With large language models, the economics or re-writing changes where the web can adapt to the personality, preference and need of the person using it. 

This is a new kind of reading surface where readers can mould the text itself, change it to their reading preferences. A passage can be tuned for detail, vocabulary complexity, and the voice. Respecting authorship and the source, the original is always available and easy to reinstate. This shifts the mental model from a document that is consumed as-is to something more fluid; a text that can be precisely dailed in, turning reading into a direct conversation with the material.
 

Par Rishab Sachidanand