Reading

Text A

When I began as a museum conservator, I assumed the job would be largely solitary: cotton gloves, quiet rooms and the slow, reverent repair of objects that had outlived their makers. In reality, the work is relentlessly collaborative. Before a single fibre is reattached, we convene with curators, scientists and—more awkwardly—donors, whose enthusiasm can be inversely proportional to their patience. The most delicate negotiations are not about chemistry but about narrative: whether a crack should be disguised to satisfy a sponsor, or left visible because it is part of the object’s biography. I have, on occasion, refused funding rather than varnish over evidence of use. Ironically, the more scanning and imaging technology improves, the less room there is for bluff; every intervention is traceable, and future colleagues will know exactly what you did and why.

Text B

I run a small company that trains ‘prompt engineers’—a title that still makes my mother laugh. The work isn’t about typing clever questions into a chatbot; it’s about designing repeatable workflows so that a model produces reliable outputs under pressure. What surprises newcomers is how much of the job is linguistic: you have to anticipate ambiguity, constrain tone, and build in checks that catch hallucinated facts before they reach a client. We spend as long writing evaluation rubrics as we do writing prompts. The biggest misconception is that the role is a shortcut for people who can’t code. In practice, the best trainees are those who can translate messy human intentions into testable instructions, and who are willing to document failures as meticulously as successes—because the next iteration depends on it.

Text C

As a mountain guide, I’m often congratulated on my ‘bravery’, as though the job were a string of heroic improvisations. If I’m doing it properly, nothing is improvised. The real skill lies in deciding not to go: turning clients around when the snowpack is unstable, even if the summit is visible and the group has paid handsomely. That decision is rarely popular in the moment, yet it is precisely what clients later thank you for—once they’ve watched the evening forecast and realised how close they were to trouble. I keep a log of near-misses, not to dramatise them but to prevent them from becoming stories with funerals attached. Experience helps, but only if you treat it as evidence rather than as entitlement.

Text D

I translate legal contracts for a living, and the hardest part is not vocabulary but accountability. Clients sometimes ask for ‘a quick version’, as if precision were an optional extra. Yet a single modal verb can shift liability by millions, and a misplaced comma may be argued over for years. I’ve learned to push back politely but firmly: if a clause is vague in the source text, I flag it rather than smoothing it into something that merely sounds authoritative. The job also involves managing expectations across cultures. A British firm may prefer understatement, while a partner abroad expects explicitness; if you mirror the wrong style, you can appear evasive or aggressive. My best days are those when nobody notices my work—because that usually means nothing has gone wrong.

Text E

I used to design sound for theatre, but I now specialise in audio for virtual reality training simulations—fire drills, medical emergencies, industrial accidents. People assume it’s all about making things ‘realistic’, yet realism is often counterproductive. If the siren is too authentic, trainees freeze; if the background noise is too dense, they miss the instructor’s cues. So I build soundscapes that are psychologically plausible rather than acoustically perfect, nudging attention where it needs to go. The ethical questions arrive unexpectedly: should we include distressing screams if they improve performance, or does that cross a line? I’ve argued for restraint, even when the client wanted maximum impact. The irony is that the more immersive the technology becomes, the more carefully we have to ration intensity.