What Actually Moves Between AI Tools
Trying to move a writing setup from Claude to ChatGPT showed the gap between portable memory, portable code, and portable workflow enforcement.
Trying to move a writing setup from Claude to ChatGPT showed the gap between portable memory, portable code, and portable workflow enforcement.
Heat turns availability into a physical dependency, from an overheating laptop in a London flat to data-centre cooling failures.
Clearing saved tabs after the AWS recert showed how quickly useful technical workarounds age into features, obsolete notes, or stale bets.
A heatwave fan problem showed how AI keeps sounding certain when it cannot check whether its answer works in the real world.
Machine-only web surfaces turn AI discovery into a trust problem when site owners cannot see what crawlers and agents actually receive.
AWS recertification keeps testing depth in one vendor while the architecture work that matters has become depth across the joins.
A disrupted London journey, three travel apps, and the difference between confident presentation and trustworthy source data.
Legacy systems and prompt-built systems share the same risk: the output can survive long after the reasoning behind it has disappeared.
Why deferring hardware refreshes no longer feels like a free hold when replacement costs are rising in the background.
Preparing AI training surfaced the difference between knowing a tool and carrying around assumptions that only survive until you have to explain them.