# A machine that doesn't get sick (slim) > Vlad Cealicu on human non-determinism and sick days as a cost of office work, and what changes when admin runs on something almost as good as average human judgement that works every hour of the year. The numbers behind both, the machine's own version of a sick day, and why the second is the one to worry about. - Full: https://www.anchorterminal.com/blog/a-machine-that-doesnt-get-sick.md (~4,450 tokens) · this version ~1,380 tokens · JSON https://www.anchorterminal.com/blog/a-machine-that-doesnt-get-sick.json · canonical https://www.anchorterminal.com/blog/a-machine-that-doesnt-get-sick - Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-11-03 - Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data) By Vlad Cealicu, 3 November 2026. 1,771 words, about 8 min to read. - 148.8 million: working days lost to sickness in the UK in 2025, 4.4 a worker - 55%: the median gap between two underwriters pricing the same risk, when the executives expected 10% - 1,533: hours a UK worker works in a year, out of 8,760 What a human seat costs, before anyone makes a mistake. This morning I wrote about [the models giving us more work than we can check](/blog/a-machine-that-gives-us-work), and that post was about the jobs that stay. ## The sick day The Office for National Statistics counted 148.8 million working days lost to sickness or injury in the UK in 2025, 4.4 days a worker, 2.0% of all working hours, and that's a rate that hasn't moved since 2024 and is still above 2019 [[1]](#ref-1). The days off are the smaller cost. ## The noise The second cost is that human judgement doesn't repeat. It gets worse with the hour. We don't count this as a cost because there's no line for it. ## The machine seat Now put a machine in the seat. The hours are the easy part of the arithmetic. - ~1,500: hours of work a year from a human seat, after sickness - 8,760: hours from a machine seat, if the machine is up - 5.8×: the hours, before quality comes into it One seat, two ways. So on the office and admin work the Bureau counts, the 17.8 million jobs I used this morning [[12]](#ref-12), the scenario is a seat that never takes the 4.4 days, never comes in with a cold, never prices the risk at 55% off its colleague, and does it at four in the morning. ## The machine's sick day What I don't want to do is pretend the machine doesn't have bad days, because it has two kinds and they're both worse than ours in one way. The first is that it isn't deterministic either. The second is the outage, and this is the one that matters. The comparison turns on one word for me. ## What I think happens So the office I'm describing has fewer human seats and the ones left are doing what I wrote about this morning, checking, and the checking is the thing that doesn't scale to 8,760 hours, because the people doing it still sleep. I'm not sure about the timing. If you run an office and you're reading this, the number I'd want from you is the one I couldn't find, which is how far apart two of your people are when they handle the same case. --- More in [Opinion](https://www.anchorterminal.com/blog/opinion/index.md) - Older: [A magic machine that gives us work](https://www.anchorterminal.com/blog/a-machine-that-gives-us-work.md) - All posts: https://www.anchorterminal.com/blog/index.md ## References Read on 3 November 2026. Quotes are as published. [1] Office for National Statistics, 1 May 2026, "Sickness absence in the UK labour market, 2025". [2] HEC Paris, news on Olivier Sibony's work with Daniel Kahneman and Cass Sunstein on their 2021 book "Noise, A Flaw in Human Judgment". [3] OECD, the annual hours worked per worker dataset (ANHRS), as tabulated on Wikipedia's list of countries by average annual labour hours, read on 3 November 2026. Source for 1,533 hours a year in the United Kingdom and 1,800 in the United States in 2025. The 8,760 hours in a year and the seats arithmetic are mine. [4] US Bureau of Labor Statistics, Current Population Survey, Table 47, "Absences from work of employed full-time wage and salary workers by occupation and industry", 2025 annual averages. [5] Deloitte, 17 May 2024, "Poor mental health costs UK employers £51 billion a year for employees". [6] Christopher P. [7] Luis C. [8] TestMu AI, "OpenAI GDPval Benchmark, Tasks, Grading, Scores and Limits", a write-up of the GDPval paper (Tejal Patwardhan and colleagues, OpenAI, 5 October 2025), read on 3 November 2026. Source for the benchmark's 1,320 tasks across 44 occupations, grading by occupation experts blind to the source, and two experts agreeing with each other 70.8% of the time on the paper's measure, with OpenAI's automated grader agreeing with experts 65.7% of the time. [9] Nikhil Agarwal, Alex Moehring, Pranav Rajpurkar and Tobias Salz, 17 November 2025, "Combining Human Expertise with Artificial Intelligence, Experimental Evidence from Radiology". [10] TechRadar, 29 September 2025, "Claude just beat GPT-5, Gemini, and Grok in real-world job tasks, according to OpenAI's own study". [11] Rajarshi Haldar and Julia Hockenmaier, November 2025, "Rating Roulette, Self-Inconsistency in LLM-As-A-Judge Frameworks", Findings of EMNLP 2025. Source for models used as judges giving different scores to the same input across three runs with settings unchanged, Qwen 3 giving the same score on all three runs in 61.3% of SummEval cases, and majority voting across runs improving balanced accuracy against human labels. [12] US Bureau of Labor Statistics, 15 May 2026, "Occupational Employment and Wages, May 2025", Table 1. Source for 17,753,430 jobs in office and administrative support occupations. [13] Simon Willison, 11 September 2025, a link post on Thinking Machines' "Defeating Nondeterminism in LLM Inference". [14] IncidentHub, 20 July 2026, updated 7 October 2026, "H1 2026 Cloud and SaaS Reliability Report".