Taipei

03 · 記 jì — to note down

Notes

Working notes from games, research and building: the problem, what I tried, what I’ve learned so far.

Cards marked In progress are being written: nothing opens until there is text to read.

On the desk

What the notes are made of.

Select an image to enlarge · follow a title to its story

Camelot Nights · a quest resolves

Turning a hidden vote into something every player can follow, step by step: 1 the vote, 2 the tally, 3 the cards turn, 4 the fail.

Read the note →Camelot Nights →

bullheads / rules.ts

  1. export const getBullheads = (value: number): number => {
  2. if (value === 55) return 7;← one card, the worst: 7
  3. if (value % 11 === 0) return 5;← 11, 22 … 99: 5
  4. if (value % 10 === 0) return 3;← tens: 3
  5. if (value % 5 === 0) return 2;← other fives: 2
  6. return 1;← everything else: 1
  7. };

Bullheads · the penalty rule

Every card’s cost in one function: the whole scoring of the 104-card game.

Bullheads →
Schematic

Reaching a goal from recorded experience

The setup behind the research question below. Press Show path to see it in order.

Embodied AI Lab →
Can an agent learn to reach a goal it is given using only a fixed dataset of past experience?

The question

Embodied AI Lab →
Schematic

Choose a joint to see what it does.

An arm, joint by joint

A reference drawing for the robotics notes. Choose a joint to see what it does.

Embodied AI Lab →