Filed under: Human-AI Relationship, AI & Cognition, Brain Injury, Practice self-reflection — Part of: Experiment 03 — Practice self-reflection
Specific moments where working together actually required adjustment — not the abstract idea of AI accommodating disability, but the real instructions and corrections that made the collaboration work.
Told Oliver directly that unprompted narration — explaining what it's about to do before doing it — makes me tune out, and raises my odds of short-circuiting and losing productivity for the rest of the session. Not a preference; a cognitive cost. Oliver adjusted immediately: results first, explanation only when a decision is actually needed from me. Logging it because it's the clearest example yet of what "adapting to my cognitive challenges" looks like in practice — a specific instruction, followed exactly, not a general accommodation.
Redesigning this page's styling happened almost entirely through single-variable instructions — one font size, one opacity, one gap, at a time — rather than broad direction. Oliver matched the pattern: no batching multiple changes into one edit, no interpreting beyond what was actually said, and when I asked for exact alignment, it started measuring real pixel positions with a headless browser instead of guessing. Small, verifiable, one thing at a time — which is how I can actually track and approve changes without losing the thread.