Awareness Behind the Wheel
DEFRITEN
Campaign

The 24 Hours

Enter the series at the hour you are living now. Then explore the other twenty-three.

Chosen hour · 09:00

09:00 Coffee

Service area towards Aargau · Plateau · 57 s

Coffee masks sleepiness; it doesn't remove it. Pull over: a coffee, then a ten-minute nap. Drive on only when awake.

Go deeper into 09:00: larger video, text, sources and limitsGo deeper: 09:00 →

Start with episode 01 → Choose another episode

24 episodes, one for every hour of the day, in four languages. Each hour covers one aspect of sleepiness while driving: biology, the road, work, chemistry, technology, society. We make them in-house and publish them here, as an open workshop: you watch, you flag the mistakes, we fix them. The campaign runs from 24 October to 16 November 2026.

The full series is online. On our channels, one episode a day, from 24 October to 16 November 2026.

Tap an hour to hear it. Every hour has its note: low when the body asks for sleep, brighter as alertness rises.

The alertness curve

The inner circle follows the body's alertness over 24 hours: it widens towards the rim when we are awake and pulls back towards the centre when the body asks for sleep. The lowest point is around 4, the peak in mid-morning, a second dip after lunch.

The chosen hour

09:00

Coffee

Service area towards Aargau · Plateau · motorway

The substanceOnline · 2026-10-01

Coffee masks sleepiness; it doesn't remove it. Pull over: a coffee, then a ten-minute nap. Drive on only when awake.

Watch the episode →

Source
8 sources, the first: Philip P, Taillard J, Moore N, et al. The effects of coffee and napping on nighttime highway driving: a randomized trial. Ann Intern Med 2006;144:785-791.
Transferability
road
The declared limit
Randomised trial on a motorway at night. The benefit is measured right afterwards: the sleep debt remains and presents the bill again.
Alertness

The graphic signature

The Drive Awake mark is a clock. This series fills its dial. The graphic signature closes every episode: two chimes, the circle drawing itself, the orange hand reaching the threshold, night turning into wakefulness. In each episode the hand marks the hour of the chapter.

The 24 hours as a list

    Go deeper

    What the research says, where its limits lie and where it comes from: the four gates, the gesture that helps, and the sources for each hour.

    What carries the hour

    Twenty-four hours do not have twenty-four physiologies, and saying so beats pretending.

    The distinct physiological states in a day are few. Instead of filling empty boxes with an invented mechanism, each hour declares what carries it. When the door is the body, the card also says which state: this way the repetition shows, instead of hiding.

    The body14 hoursBut the distinct states are 10, not 14: the card says which.
    The road5 hoursWho is on the road at that hour, and how many.
    The substance3 hoursCaffeine, medication, alcohol.
    Technology2 hoursThe on-board sensor, the screen: what the machine measures.

    Three hours share the same post-lunch dip and two the same circadian minimum. It is not a flaw to cover: it is why the hours are not interchangeable, because what changes is who is on the road, not the body.

    The gesture

    Coffee, then a short sleep. In that order.

    The order looks wrong, and that is why it sticks: the caffeine arrives while you are asleep.

    1. You stop, in a lay-by or a rest area.
    2. You drink the coffee.
    3. Alarm at ten to twenty minutes, engine off.
    4. You sleep. Then you set off again.

    Philip P, Taillard J, Moore N et al. 2006, Annals of Internal Medicine 144(11):785-791. Two hundred kilometres of open motorway, night sessions between 02:00 and 03:30. Drove as in daytime: 9 drivers out of 12 after coffee, 8 out of 12 after thirty minutes of sleep, 3 out of 12 after placebo. DOI 10.7326/0003-4819-144-11-200606060-00004 DOI 10.1093/sleep/29.6.831
    Limit: twelve healthy men, mean age 21. The study tested coffee or sleep, separately: the combination was not measured. The ten-minute short sleep comes from Brooks and Lack 2006: it restores without the inertia of waking.

    The sources, all of them

    Twenty-four hours, the main source of each one, and where it was measured.

    Transferability says where the study was done: in the laboratory, in a simulator, on the open road, or whether it is a review, a national statistic or a regulation. A figure measured in a simulator is not a figure about real driving, and we do not write it as if it were. For hours in preparation the source is the planned one; for hours online it is the first of the episode's sources. Where a source is under review, the table says so instead of hiding it.

    HourEpisodeMain sourceWhere
    00:00 The edge of the dayonline Czeisler CA, Duffy JF, Shanahan TL, et al. Stability, precision, and near-24-hour period of the human circadian pacemaker. Science 1999;284:2177-2181.
    DOI 10.1126/science.284.5423.2177 +1    		 laboratory
    01:00 Sleep debtonline Van Dongen HPA, Maislin G, Mullington JM, Dinges DF. The cumulative cost of additional wakefulness: dose-response effects on neurobehavioral functions and sleep physiology from chronic sleep restriction and total sleep deprivation. Sleep 2003;26(2):117-126.
    DOI 10.1093/sleep/26.2.117    		 laboratory
    02:00 Approaching the low pointonline Akerstedt T, Kecklund G. Age, gender and early morning highway accidents. J Sleep Res 2001;10(2):105-110.
    DOI 10.1046/j.1365-2869.2001.00248.x    		 road
    03:00 The critical windowonline Wright KP, Hull JT, Czeisler CA. Relationship between alertness, performance, and body temperature in humans. Am J Physiol Regul Integr Comp Physiol 2002;283:R1370-R1377.
    DOI 10.1152/ajpregu.00205.2002 +2    		 laboratory
    04:00 The absolute minimumonline Wright KP, Hull JT, Czeisler CA. Relationship between alertness, performance, and body temperature in humans. Am J Physiol Regul Integr Comp Physiol 2002;283:R1370-R1377.
    DOI 10.1152/ajpregu.00205.2002 +2    		 laboratory
    05:00 The night shiftonline Åkerstedt T. Shift work and disturbed sleep/wakefulness. Occup Med (Lond) 2003;53:89-94.
    DOI 10.1093/occmed/kqg046 +1    		 review
    06:00 The way homeonline Lee ML, Howard ME, Horrey WJ, et al. High risk of near-crash driving events following night-shift work. Proc Natl Acad Sci USA 2016;113:176-181.
    DOI 10.1073/pnas.1510383112 +4    		 road
    07:00 School and the roadonline Fischer D, Lombardi DA, Marucci-Wellman H, Roenneberg T. Chronotypes in the US - influence of age and sex. PLoS ONE 2017;12:e0178782.
    DOI 10.1371/journal.pone.0178782 +4    		 laboratory
    08:00 The commuteronline Christian TJ. Trade-offs between commuting time and health-related activities. J Urban Health 2012;89:746-757.
    DOI 10.1007/s11524-012-9678-6 +3    		 observational
    09:00 Coffeeonline Philip P, Taillard J, Moore N, et al. The effects of coffee and napping on nighttime highway driving: a randomized trial. Ann Intern Med 2006;144:785-791.
    DOI 10.7326/0003-4819-144-11-200606060-00004 +7    		 road
    10:00 The morning windowonline Wright KP Jr, Hull JT, Czeisler CA. Relationship between alertness, performance, and body temperature in humans. Am J Physiol Regul Integr Comp Physiol 2002;283:R1370-R1377.
    DOI 10.1152/ajpregu.00205.2002 +5    		 laboratory
    11:00 The medicationonline Weiler JM, Bloomfield JR, Woodworth GG, et al. Effects of fexofenadine, diphenhydramine, and alcohol on driving performance. A randomized, placebo-controlled trial in the Iowa Driving Simulator. Ann Intern Med 2000;132:354-363.
    DOI 10.7326/0003-4819-132-5-200003070-00004    		 simulator
    12:00 Middayonline Reyner LA, Wells SJ, Mortlock V, Horne JA. 'Post-lunch' sleepiness during prolonged, monotonous driving - effects of meal size. Physiol Behav 2012;105(4):1088-1091.
    DOI 10.1016/j.physbeh.2011.11.025    		 simulator
    13:00 The diponline Monk TH. The post-lunch dip in performance. Clin Sports Med 2005;24(2):e15-e23.
    DOI 10.1016/j.csm.2004.12.002    		 review
    14:00 Ten minutesonline Brooks A, Lack L. A brief afternoon nap following nocturnal sleep restriction: which nap duration is most recuperative? Sleep 2006;29(6):831-840.
    DOI 10.1093/sleep/29.6.831    		 laboratory
    15:00 The second riskonline Pack AI, Pack AM, Rodgman E, Cucchiara A, Dinges DF, Schwab CW. Characteristics of crashes attributed to the driver having fallen asleep. Accid Anal Prev 1995;27(6):769-775.
    DOI 10.1016/0001-4575(95)00034-8    		 observational
    16:00 The technologyonline Regolamento delegato (UE) 2021/1341 della Commissione, del 23 aprile 2021, che integra il regolamento (UE) 2019/2144 (sistemi di avviso della sonnolenza e del calo di attenzione del conducente), CELEX 32021R1341. regulation
    17:00 Leaving workonline Philip P, Sagaspe P, Moore N, Taillard J, Charles A, Guilleminault C, Bioulac B. Fatigue, sleep restriction and driving performance. Accid Anal Prev 2005;37(3):473-478.
    DOI 10.1016/j.aap.2004.07.007    		 road
    18:00 The billonline bfu. Sinus 2024: Sicherheitsniveau und Unfallgeschehen im Strassenverkehr 2023. Bern: BFU; 2024, S. 63.
    DOI 10.13100/BFU.2.536.01.2024    		 national statistics
    19:00 Dinner and drinksonline Horne JA, Reyner LA, Barrett PR. Driving impairment due to sleepiness is exacerbated by low alcohol intake. Occup Environ Med 2003;60(9):689-692.
    DOI 10.1136/oem.60.9.689    		 simulator
    20:00 The familyonline Danner F, Phillips B. Adolescent sleep, school start times, and teen motor vehicle crashes. J Clin Sleep Med 2008;4(6):533-535. observational
    21:00 Melatoninonline Zeitzer JM, Dijk DJ, Kronauer R, Brown E, Czeisler C. Sensitivity of the human circadian pacemaker to nocturnal light: melatonin phase resetting and suppression. J Physiol 2000;526(3):695-702.
    DOI 10.1111/j.1469-7793.2000.00695.x    		 laboratory
    22:00 The screensonline Chang AM, Aeschbach D, Duffy JF, Czeisler CA. Evening use of light-emitting eReaders negatively affects sleep, circadian timing, and next-morning alertness. PNAS 2015;112(4):1232-1237.
    DOI 10.1073/pnas.1418490112    		 laboratory
    23:00 Night departureonline Tefft BC. Acute sleep deprivation and culpable motor vehicle crash involvement. Sleep 2018;41(10):zsy144.
    DOI 10.1093/sleep/zsy144    		 road

    Take part

    Drive Awake gives voice to research and to professionals, and opens the conversation to the public.

    Here we keep three things apart: what the research says, what people tell us, and what is up to decision-makers.

    A question about this hour · 01:00

    In weeks when you sleep less, do you notice any difference? If so, what?

    There is no right answer: we are interested in the experience of people who drive, including for work.

    The conversations happen on our channels. When there is one about this hour, we will link to it here.

    Reply on our channels: Instagram Facebook LinkedIn YouTube TikTok Telegram

    The questions, hour by hour 24
    1. 01:00In weeks when you sleep less, do you notice any difference? If so, what?
    2. 02:00If you ever drive between one and three in the morning, what decides when you set off?
    3. 03:00What do you do, or what have you seen others do, when sleepiness sets in on a night drive?
    4. 04:00Around four in the morning, have you ever had to choose between waiting and setting off? What influenced your decision?
    5. 05:00If you work nights, how do you organise your journey home after the shift?
    6. 06:00At work or in your family, who decides how people get home after a night shift?
    7. 07:00If you are a new driver, what shapes your decisions in the morning before you get behind the wheel?
    8. 08:00How much time do you spend travelling each day, and how does it fit with the rest of your day?
    9. 09:00When you take a break from driving, what do you usually do, and why?
    10. 10:00At what times of day is your attention at the wheel at its best, and how can you tell?
    11. 11:00If you have ever been prescribed or recommended a medicine, were you told anything about driving? If so, what, and by whom?
    12. 12:00When you know you will be driving in the afternoon, does anything change about your lunch? If so, what?
    13. 13:00At the wheel in the early afternoon, how do you feel compared with the morning?
    14. 14:00Have you ever stopped to sleep for a few minutes during a journey? What made it possible, or difficult?
    15. 15:00Are there times of day when you expect sleepiness at the wheel? If so, when, and where does that idea come from?
    16. 16:00What would you expect from an in-car system that warns of sleepiness, and what would you not expect from it?
    17. 17:00At the end of a working day, how do you decide whether you are fit to drive?
    18. 18:00In your view, what would it take to know how many crashes are linked to sleepiness?
    19. 19:00When an evening out runs late, how do you decide who drives home?
    20. 20:00In your household, who or what decides when people go to bed?
    21. 21:00What is the lighting like in your evenings, and what do you notice in the hours before sleep?
    22. 22:00What do you usually do in the last hour before sleep? And how do you generally feel in the morning?
    23. 23:00Before a long journey, how do you choose between leaving in the evening and leaving after a night's sleep?
    24. 00:00If you are planning to drive at midnight, what helps you decide whether to set off?

    Spotted a mistake?

    Reporting a mistake is separate from the questions: use the form below and, if your report is accepted, it will be recorded in the diary.

    Tell us what you noticed

    What happens next: every report is read. If it becomes a correction, the text changes here on the site and the correction goes into the episode's diary, with its date and its reason. The video plates are remade in the next production window: until then the video may still show the old line. Corrections already made →

    Details, if you like
    We keep nothing beyond what you write here.

    How we worked

    Method, diary, earlier versions and the current state of the text: all of it remains available.

    The method

    Made in-house, corrected together

    We produce the 24 Hours campaign ourselves, one hour at a time, and publish it as it takes shape. Images, voice and music are generated with artificial intelligence tools, chosen by us and declared; text, structure, editing, typography, signature and source checking are our own, with a DOI for every figure.

    The episodes are online. If we find a mistake, we correct it. The lines are written in Italian and then carried into French, German and English, and that is where mistakes creep in: a turn of phrase that doesn't sound right, a word out of place, a figure worth checking, a voice that runs too fast. If you notice one, write it here. We're getting this campaign right together.

    Spotted a mistake? →

    The diary below keeps track of what changes.

    The work
    The whole work

    The twenty-four minutes

    The episodes in a row make a complete circle, from dark to dark. The protagonist is not a person: it is the road. Twenty-four people pass the baton and none of them has a face. The trace never breaks, and whoever watches it all sees the day.

    Not yet edited

    The archives
    The layered archive

    Old versions are not deleted: they go down a level.

    The first nine hours online, released from 6 to 15 September, have been replaced by the new versions. Each replacement is in the diary, with the date and the reason.

    Level 0 · on the pageThe current version. It is the one you see, share and cite.
    Level 1 · previous versionsEvery replaced draft, with its date and a line saying what changed and why. Reached from the hour's card, not from the home page.
    Level 2 · the correctionsWhat was reported to us and what we did with it. It is the production diary, tied to the hour rather than to the date.

    It is not only honesty: it is the only proof that the method works. A campaign that shows where it corrected itself says something no statement of intent can say.

    Production timeline 24

    The order in which the hours were produced and put online on the platform. This differs from the viewing order.

    1. 23:00 Night departurefirst version online 03.10.2026 · current version 03.10.2026
    2. 22:00 The screensfirst version online 03.10.2026 · current version 03.10.2026
    3. 21:00 Melatoninfirst version online 03.10.2026 · current version 03.10.2026
    4. 20:00 The familyfirst version online 03.10.2026 · current version 03.10.2026
    5. 19:00 Dinner and drinksfirst version online 03.10.2026 · current version 03.10.2026
    6. 18:00 The billfirst version online 03.10.2026 · current version 03.10.2026
    7. 17:00 Leaving workfirst version online 03.10.2026 · current version 03.10.2026
    8. 16:00 The technologyfirst version online 03.10.2026 · current version 03.10.2026
    9. 02:00 Approaching the low pointfirst version online 03.10.2026 · current version 03.10.2026
    10. 01:00 Sleep debtfirst version online 03.10.2026 · current version 03.10.2026
    11. 00:00 The edge of the dayfirst version online 03.10.2026 · current version 03.10.2026
    12. 15:00 The second riskfirst version online 02.10.2026 · current version 02.10.2026
    13. 14:00 Ten minutesfirst version online 29.09.2026 · current version 29.09.2026
    14. 13:00 The dipfirst version online 25.09.2026 · current version 30.09.2026
    15. 12:00 Middayfirst version online 19.09.2026 · current version 30.09.2026
    16. 11:00 The medicationfirst version online 15.09.2026 · current version 30.09.2026
    17. 10:00 The morning windowfirst version online 09.09.2026 · current version 01.10.2026
    18. 09:00 Coffeefirst version online 08.09.2026 · current version 01.10.2026
    19. 08:00 The commuterfirst version online 08.09.2026 · current version 01.10.2026
    20. 07:00 School and the roadfirst version online 08.09.2026 · current version 02.10.2026
    21. 06:00 The way homefirst version online 08.09.2026 · current version 02.10.2026
    22. 05:00 The night shiftfirst version online 07.09.2026 · current version 02.10.2026
    23. 04:00 The absolute minimumfirst version online 06.09.2026 · current version 02.10.2026
    24. 03:00 The critical windowfirst version online 06.09.2026 · current version 02.10.2026
    Production diary 30 notes · last 2026-10-03

    The diary is a chronology: each entry describes the state on that day, not the current one.

    • Episode 02 online, in its final version: approaching the low point. 2 am at the entrance to the Gotthard. The source is Åkerstedt and Kecklund (2001): in Sweden, for drivers aged 18 to 24, the crash risk at night was about five times that of late morning, with alcohol-related crashes excluded. It is a relative risk by hour and age, not the risk of any one person.
    • Episode 01 online, in its final version: sleep debt. 1 am on the Leventina cantonal road. The source is Van Dongen and colleagues (2003): 48 healthy adults kept in the lab on four, six or eight hours of sleep a night for fourteen nights. On six hours, errors rose to the level of up to two nights without sleep, while reported sleepiness barely changed. The debt grows without being felt.
    • Episode 00 online, in its final version: the edge of the day. Midnight in the Muggio valley. The source is Czeisler and colleagues (1999): in the lab, without cues from outside, the body clock runs a cycle of 24 hours and 11 minutes on average. Morning light resets it every day (Khalsa and colleagues, 2003). At midnight, for the body, the night has already begun.
    • Episode 23 online, in its final version: night departure. 11 pm at the service station before the Chiasso crossing. The source is Tefft (2018): 6,845 drivers involved in crashes in the United States. Those who had slept less than four hours in the previous 24 hours had about 15 times the odds of being the driver responsible for the crash, compared with those who had slept 7 to 9 hours. The hours of sleep are as reported by the drivers.
    • Episode 22 online, in its final version: the screens. 10 pm, at home. The source is Chang and colleagues (2015): in the lab, people who read on a light-emitting screen before bed fell asleep later, had less melatonin and were less alert the next morning than those who read on paper.
    • Episode 21 online, in its final version: melatonin. 9 pm, on the descent through the Leventina. The source is Zeitzer and colleagues (2000): in the lab, ordinary room light of about 100 lux already produces half of the maximum delay of the body clock, the delay reached with bright light of about 9,000 lux. In the evening, the light we switch on at home counts too.
    • Episode 20 online, in its final version: the family. 8 pm in the courtyard of a village house in Surselva. The source is Danner and Phillips (2008): in a Kentucky county, the school district moved the start of high school classes one hour later. In the following two years, the crash rate of teen drivers fell by 16.5% compared with the two years before, while in the rest of the state it rose by 7.8%. It is an observational study, not an experiment.
    • Episode 19 online, in its final version: dinner and drinks. 7 pm in an inn in Uri. The source is Horne, Reyner and Barrett (2003): twelve young men in a driving simulator, after a short night and with blood alcohol at about half the UK limit of the time. Together, the two factors impair driving more than either alone, and the alcohol does not make them feel any sleepier.
    • Episode 18 online, in its final version: the bill. 6 pm on a bridge over an Oberland lake. The source is the bfu report Sinus 2024 (p. 63): on motorways, averaged over 2019 to 2023, the category «tiredness, microsleep» is the main cause in 7% of serious-injury crashes and 1% of fatal crashes. The figure comes from the people who recorded the crash: sleepiness cannot be measured at the scene.
    • Episode 17 online, in its final version: leaving work. 5 pm on the Simplon road. The source is Philip and colleagues (2005): twenty-two drivers, a thousand kilometres on a motorway, in real traffic. After a night of restricted sleep, 535 lane departures; after a normal night, 66. What predicted them was the sleepiness reported before driving, not feeling exhausted.
    • Episode 16 online, in its final version: the technology. 4 pm on a viaduct in the Rhône valley. The source is Delegated Regulation (EU) 2021/1341: since 7 July 2024 every new car registered in the EU must warn a driver who is nodding off, from level 8 of 9 on the sleepiness scale. The system warns; it does not stop you. The choice is made before setting off.
    • Episode 15 online, in its final version: the second risk. 3 pm in Geneva. The source is Pack and colleagues (1995): 4,333 crashes in North Carolina in which the driver was judged asleep, with two peaks across the 24 hours, at night and at 3 pm. It is the police's assessment, not a diagnosis: a minimum, not a total.
    • Episodes 07 to 03 move to their final version, in the same form as episodes 08 to 14. The images are those of the first release; the voice, the sound and the edit change. The night hours, from 03 to 05, also keep the colours they were first made with. In November, before launch, we review the 24 hours one last time.
    • Episodes 10, 09 and 08 move to their final version, in the same form as episodes 11 to 14. The images stay those of the first release; the voice, the sound and the edit change. Episodes 09 and 10 now have six boards, like all the others: in episode 10 the board on the three peaks is dropped, because it did not match the two peaks of episode 15. In November, before launch, we review the 24 hours one last time.
    • Episodes 13, 12 and 11 move to their final version, ahead of the hours after 2 pm: the round does not always follow clock order. The images stay those of the first release. The voice, the sound and the form change and now match episode 14, and the lines corrected in the meantime go into the videos. In episode 13 the closing line names an action: if you drive between one and three, plan a stop.
    • The final round begins. Episode 14, «Ten minutes», is the first in its final version: text, voice, images and sound remade with the final method. The other hours follow one at a time, in clock order, from 3 pm to 1 pm. Until their turn comes they stay drafts, and they say so.
    • Episode 14 online: ten minutes. The car park at the edge of a village on La Côte at two in the afternoon, with only one space in the shade. The sources are Brooks and Lack (2006), who compared naps of five, ten, twenty and thirty minutes of sleep in the laboratory, and Horne and Reyner (1996), in a driving simulator. Ten minutes work at once; twenty and thirty help, but later. These are measured minutes of sleep, not time spent at a stop: neither study tested a nap in a car on a real road.
    • Episode 13 online: the dip. Lavaux just past one, the road between the vines and the lake, a tractor and a car catching up with it. The source is Monk's review (2005): the early-afternoon dip appears even without lunch, and a carbohydrate-heavy meal deepens it but does not create it. It is a review, not a driving study: it shows that the dip exists and where it comes from, not how much the risk at the wheel rises.
    • Episode 12 online: midday. The plateau between the Broye and the Gros-de-Vaud, two plates on a table, the afternoon road. The source is the Loughborough driving simulator (Reyner 2012): after a lunch three times as large, on five hours of sleep, lane drifting increased. First version: the voices and the animations are under revision and will be replaced.
    • Episode 11 online: the medication. A village in the Jura at eleven, a pharmacy open, a box in the hand. The source is the Iowa driving simulator (Weiler 2000): an older antihistamine degraded driving more than alcohol, and those who drove worst did not notice.
    • Method statement corrected: in the eight episodes online the voice and the music are generated too, like the images, and now we say so. Text, structure, editing, typography, signature and source checking remain our own. No episode has changed.
    • Episode 10 online, The morning window: the lake road in Romandy, the high sun, the motorbike; the same curve as 04, seen from the peak; low sleep need and the body clock pushing towards the day; two to four hours to get going after waking (Jewett 1999); the three peaks of sleepiness-related crashes at two, six and four in the afternoon (Horne and Reyner 1995); early birds and night owls; without a night's sleep there is no peak (Van Dongen 2003). First episode on the positive side. Ring with the three peaks in orange and a blue hand. Landscape generated, declared.
    • Episode 09 online, Coffee: the moka pot, the café counter in Lugano, the lake; adenosine and the caffeine that takes its place; on a real road at night 75% drive as by day after a coffee, 66% after a nap (Philip 2006); every five hours half is left. First episode with eleven frames and three new forms: molecule, bars, half-life. Lugano generated, declared as an exception.
    • Episode 08 online, The commuter: the city at 8, tram and queue, milky sky; one more hour of commuting and almost a third of it taken from sleep (Christian 2012, Basner 2007). New one-hour share form; first city episode, generated and declared imagery.
    • Episode 07 online, School and the road: a school forecourt at 7, white winter light, an eighteen-year-old's midpoint of sleep around 4:30 (Fischer 2017, Roenneberg 2004). New two-curve form, adult and adolescent; generated and declared imagery, teenagers always seen from behind.
    • Episode 06 online, The way home: the mountain road at 6, light coming down the hairpins, one drive in three with a near-collision after the night shift (Lee 2016). First colour curve of dawn and new drive form; born from the template in one night, generated and declared imagery.
    • Episode 05 online, The night shift: a hospital forecourt at 5, a nurse in profile, daytime sleep 2 to 4 hours shorter (Åkerstedt 2003). Third episode, the first with the bar form; born entirely from the template, generated and declared imagery.
    • Episodes 03 and 04 realigned to the final template: text and mark outside the zones covered by social interfaces, the 24-hour curve also in 03, captions and thumbnails for every scene. From here on every episode is born with the same form.
    • Episode 04 online: the first with the infographic layer (circadian curve, core temperature, reaction distance, 24-hour ring), every figure with a DOI. The rest area is a still frame we animated ourselves: mist, light, leaves.
    • Episode 03 complete: pilot of the series, editing template, voice in four languages, graphic signature. The other twenty-three hours are declared ideas.
    The state of the text

    The state of the text, hour by hour and language by language

    The twenty-four hours are written in four languages, and every line goes through a blind test: the sentences are shown to independent engines, in the language of the text, without saying what they are. Into the deck we also slip certified sentences, written by native speakers and published by prevention bodies, to measure how severe the judge is. On 16 September the measurement showed that six of those eleven sentences were rejected: the verdict alone certifies nothing. We apply only what the reason pins to a mistake anyone can check. This grid says where the test has been done, where the text has already been corrected and what is still open. The empty boxes are work still missing, and we write it here rather than wait for it to show.

    1 / 24hours put through three engines
    0hour-language boxes out of 96 still to check
    215lines corrected and published, with the reason in the diary
    4lines reported and not yet resolved
    0 / 151Romansh: not one line read by a native speaker
    hour ITFRDEENRM
    00 never checked
    01 never checked
    02 never checked
    03 never checked
    04 never checked
    05 never checked
    06 never checked
    07 never checked
    08 never checked
    09 never checked
    10 never checked
    11 never checked
    12 never checked
    13 never checked
    14 never checked
    15 never checked
    16 never checked
    17 never checked
    18 never checked
    19 never checked
    20 never checked
    21 never checked
    22 never checked
    23 never checked

    three engines two engines one engine never checked lines corrected to resolve

    Did you spot a mistake in one of these boxes? Write it here: every report goes into the diary, with its date. Report a correction → Read every line, in all four languages →

    00 Written on 17.09.2026 and checked at once. Two blind rounds per language, with certified sentences mixed in unannounced, plus a third round on German and one open round over the whole episode for rhythm. No certified sentence was wrongly flagged. Nineteen plates out of twenty-four changed. One engine only, Google AI Mode: the box stays at one engine (P-CO-15). Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. English lines reviewed by a native speaker, George Zisa, on 28.09.2026: 5 of 6 now use his version. Where we adjusted them, two judges checked the line again, working blind. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 2 lines corrected.

    01 Written on 17.09.2026 from the source read on PubMed, and checked at once: one blind round per language on Google AI Mode, lines shuffled and certified sentences mixed in unannounced. One engine only: the box stays at one engine (P-CO-15). No certified sentence was flagged. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. English lines reviewed by a native speaker, George Zisa, on 28.09.2026: 6 of 6 now use his version. Where we adjusted them, two judges checked the line again, working blind. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 2 lines corrected.

    02 Written on 17.09.2026 from the source read on PubMed, and checked at once: one blind round per language on Google AI Mode, lines shuffled and certified sentences mixed in unannounced. One engine only: the box stays at one engine (P-CO-15). No certified sentence was flagged. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. English lines reviewed by a native speaker, George Zisa, on 28.09.2026: 5 of 6 now use his version. Where we adjusted them, two judges checked the line again, working blind. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 1 line corrected.

    03, 07 Italian: Google AI Mode, 16.09.2026. On the evening of 16.09 a second round with two engines and a calibration set in all four languages: partial coverage, and within that set six certified sentences out of eleven were rejected by both judges. The verdict alone certifies nothing, the reason is what counts: the box stays at one engine. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. English lines reviewed by a native speaker, George Zisa, on 28.09.2026. His versions go in when the video for this hour is remade; until then the page says what the video says. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind.

    04 Italian: Google AI Mode, 16.09.2026. On the evening of 16.09 a second round with two engines and a calibration set in all four languages: partial coverage, and within that set six certified sentences out of eleven were rejected by both judges. The verdict alone certifies nothing, the reason is what counts: the box stays at one engine. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. English lines reviewed by a native speaker, George Zisa, on 28.09.2026. His versions go in when the video for this hour is remade; until then the page says what the video says. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 1 line to correct when the video is remade.

    05, 09 Italian: Google AI Mode, 16.09.2026. On the evening of 16.09 a second round with two engines and a calibration set in all four languages: partial coverage, and within that set six certified sentences out of eleven were rejected by both judges. The verdict alone certifies nothing, the reason is what counts: the box stays at one engine. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. English lines reviewed by a native speaker, George Zisa, on 28.09.2026. His versions go in when the video for this hour is remade; until then the page says what the video says. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 5 lines to correct when the video is remade.

    06 Italian: Google AI Mode, 16.09.2026. On the evening of 16.09 a second round with two engines and a calibration set in all four languages: partial coverage, and within that set six certified sentences out of eleven were rejected by both judges. The verdict alone certifies nothing, the reason is what counts: the box stays at one engine. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. 27.09.2026: the note for scene 6 has been corrected; it credited Brooks and Lack 2006 with more than the study measures. The plate and the video are unchanged. English lines reviewed by a native speaker, George Zisa, on 28.09.2026. His versions go in when the video for this hour is remade; until then the page says what the video says. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind.

    08 Italian: Google AI Mode, 16.09.2026. On the evening of 16.09 a second round with two engines and a calibration set in all four languages: partial coverage, and within that set six certified sentences out of eleven were rejected by both judges. The verdict alone certifies nothing, the reason is what counts: the box stays at one engine. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. 27.09.2026: the note for scene 6 has been corrected; it credited Brooks and Lack 2006 with more than the study measures. The plate and the video are unchanged. English lines reviewed by a native speaker, George Zisa, on 28.09.2026. His versions go in when the video for this hour is remade; until then the page says what the video says. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 4 lines to correct when the video is remade.

    10 Italian: Google AI Mode, 16.09.2026. On the evening of 16.09 a second round with two engines and a calibration set in all four languages: partial coverage, and within that set six certified sentences out of eleven were rejected by both judges. The verdict alone certifies nothing, the reason is what counts: the box stays at one engine. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. English lines reviewed by a native speaker, George Zisa, on 28.09.2026. His versions go in when the video for this hour is remade; until then the page says what the video says. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 2 lines to correct when the video is remade.

    11 Italian: Google AI Mode, 16.09.2026. On the evening of 16.09 a second round with two engines and a calibration set in all four languages: partial coverage, and within that set six certified sentences out of eleven were rejected by both judges. The verdict alone certifies nothing, the reason is what counts: the box stays at one engine. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. English lines reviewed by a native speaker, George Zisa, on 28.09.2026. His versions go in when the video for this hour is remade; until then the page says what the video says. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 3 lines to correct when the video is remade.

    12 Full test: three independent engines, four languages, three rounds. Ten lines rewritten and remeasured. Fourth round on 17.09.2026 with Google AI Mode and a deck of certified sentences, before going on the site: nothing flagged on the six lines in Italian, French and German. One German line was rewritten so that it names the heavy lunch as the other languages do, and the new version passed that round. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 2 lines to correct when the video is remade.

    13 Text rewritten from scratch on 17.09.2026: the one checked with two engines on 16.09 no longer existed in any file. The old verdict does not apply to the new text. One blind round per language on Google AI Mode with a deck of certified sentences: one engine. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 4 lines to correct when the video is remade.

    14 Written on 17.09.2026 from the source read on PubMed, and checked at once: one blind round per language on Google AI Mode, lines shuffled and certified sentences mixed in unannounced. One engine only: the box stays at one engine (P-CO-15). No certified sentence was flagged. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. Rewritten on 25 and 27.09.2026 after two audits: location, closing line, plates and notes. Four blind rounds on French, German and English, with two judges given no context and control errors slipped in unannounced; every one was found. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 5 lines corrected.

    15 Written on 17.09.2026 from the source read on PubMed, and checked at once: one blind round per language on Google AI Mode, lines shuffled and certified sentences mixed in unannounced. One engine only: the box stays at one engine (P-CO-15). No certified sentence was flagged. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 6 lines corrected.

    16 Written on 17.09.2026 from the verified source and put at once through one blind round per language on Google AI Mode, with lines shuffled and certified sentences mixed in unannounced. One engine only: the box stays at one engine (P-CO-15). No certified sentence was flagged. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 2 lines corrected.

    17 Written on 17.09.2026 from Philip 2005 (DOI 10.1016/j.aap.2004.07.007), source read and not quoted from memory. One blind round per language at writing, a second round in the 17.09 audit inside a four-hour deck. One engine, two rounds: the box stays at one engine (P-CO-15). French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 4 lines corrected.

    18 Written on 17.09.2026. Editorial decision: the hour does not carry a blended percentage, it carries the fact that sleepiness is never counted. The bfu figure (about one crash in ten) is given as an estimate and never mixed with the EOC figure (up to 30%). Sources: the two bfu pages, no DOI but verifiable (P-SC-08). One blind round per language at writing, a second in the audit. One engine, two rounds. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 2 lines corrected.

    19 Written on 17.09.2026 from the verified source and put at once through one blind round per language on Google AI Mode, with lines shuffled and certified sentences mixed in unannounced. One engine only: the box stays at one engine (P-CO-15). No certified sentence was flagged. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. English lines reviewed by a native speaker, George Zisa, on 28.09.2026: 6 of 6 now use his version. Where we adjusted them, two judges checked the line again, working blind. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 5 lines corrected.

    20 Written on 17.09.2026 from the verified source and put at once through one blind round per language on Google AI Mode, with lines shuffled and certified sentences mixed in unannounced. One engine only: the box stays at one engine (P-CO-15). No certified sentence was flagged. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. English lines reviewed by a native speaker, George Zisa, on 28.09.2026: 6 of 6 now use his version. Where we adjusted them, two judges checked the line again, working blind. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind.

    21 Written on 17.09.2026 from the verified source and put at once through one blind round per language on Google AI Mode, with lines shuffled and certified sentences mixed in unannounced. One engine only: the box stays at one engine (P-CO-15). No certified sentence was flagged. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. English lines reviewed by a native speaker, George Zisa, on 28.09.2026: 6 of 6 now use his version. Where we adjusted them, two judges checked the line again, working blind. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 3 lines corrected.

    22 Written on 17.09.2026 from the verified source and put at once through one blind round per language on Google AI Mode, with lines shuffled and certified sentences mixed in unannounced. One engine only: the box stays at one engine (P-CO-15). No certified sentence was flagged. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. English lines reviewed by a native speaker, George Zisa, on 28.09.2026: 6 of 6 now use his version. Where we adjusted them, two judges checked the line again, working blind. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. 2 lines corrected.

    23 Written on 17.09.2026 from the source read on PubMed, and checked at once: one blind round per language on Google AI Mode, lines shuffled and certified sentences mixed in unannounced. One engine only: the box stays at one engine (P-CO-15). No certified sentence was flagged. Audit of 17.09.2026: a second blind round per language on a block deck, with three certified sentences mixed in. No false positive on the certified sentences across 28 rounds. The applied corrections are in the decisions glossary. English lines reviewed by a native speaker, George Zisa, on 28.09.2026: 5 of 6 now use his version. Where we adjusted them, two judges checked the line again, working blind. French and German: blind cross-check using two external engines (Google AI Mode and ChatGPT), 28.09.2026. A second round the same day, with the Italian alongside for meaning: Google AI Mode and two judges working blind. A third round the same day, on notes and page texts: Google AI Mode and two judges working blind. 6 lines corrected.

    RM Romansh is a working draft: the fourth national language arrives with Drive Awake 2.0, and no line will be published before a native reading.

    updated on 2026-09-28

    What changed in the text

    What changed in the text

    The series is online. If we find a mistake, we correct it: every correction is here, with its date, its reason and the lines it touched. What is still unresolved is written down too, further below. Entries follow the dates and describe the state on that day, not the current one.

    1. 16.09.2026Reported by RoadCross, resolved with George Zisa

      In English the hour is written with am and pm

      The English lines gave the hour bare: «At 4», «Six o'clock», «Between 2 and 6». An English reader cannot tell morning from afternoon, and in episodes about the lowest point of the night that ambiguity takes away the very thing being said. Every hour named now carries am or pm. Numbers that are not hours are unchanged.

      Lines touched 03:00 EN line 1 · 04:00 EN line 1, EN line 4 · 05:00 EN line 1 · 06:00 EN line 1 · 07:00 EN line 1, EN line 3 · 08:00 EN line 1 · 09:00 EN line 1, EN line 8 · 10:00 EN line 1, EN line 3, EN line 4, EN line 6, EN line 8

    2. 16.09.2026Reported by RoadCross, resolved with George Zisa

      «Protect your sleep» did not say what to do

      Thomas Ruckstuhl had not reported a sentence but a phrase: «protect your sleep» does not say what to do. It appeared in two English closing lines. In hour 07 George Zisa rewrote it naming the action: make sure you are rested before getting behind the wheel. In hour 08 the same closing line could not be repeated two episodes apart, so it now says not to shorten your sleep when you cannot shorten the drive. Searched through the whole text: none left.

      Lines touched 07:00 EN line 6 · 08:00 EN line 6

    3. 16.09.2026Two-engine blind test, with a calibration set

      Four loan translations, in Italian, German and French

      The lines of every hour online went past two judges with no context, mixed in with certified sentences written by native speakers. Only the four whose reason names a mistake anyone can check were applied: «piu tempo prende il viaggio» was a loan translation of «it takes time»; «non e poca volonta» is ungrammatical; the German talked around «Nachtschicht», a word the same episode already uses twice; in French «le total» is arithmetic and follows nobody to bed.

      Lines touched 05:00 IT line 5, DE line 6 · 08:00 IT line 2 · 09:00 FR line 9

    4. 17.09.2026Language audit by the direction

      «Newer» is not the name of anything

      The study cited compares two antihistamines: diphenhydramine, first generation, and fexofenadine, second. That is the name on the box and the one a doctor recognises. The plates said «an older» and «a newer» antihistamine: a comparison with no second term, and a word that is no use at the pharmacy counter. French already used the exact term in both lines; the other three languages now use it too. The note under the line already read «second-generation», two centimetres from the plate that said «newer».

      Lines touched 11:00 IT line 3, IT line 4, DE line 3, DE line 4, EN line 3, EN line 4

    5. 17.09.2026Cross-comparison test, two judges

      The night shift had five names

      In the 5 am episode the same thing went by five names: turno di notte, notte di lavoro, notte al lavoro, and in French nuit de travail where poste de nuit exists. German said Nachtschicht throughout and was the only consistent one. All four languages now use the name of the shift where the shift is what is meant. Anyone reading two episodes in a row does not have to wonder whether it is the same thing.

      Lines touched 05:00 IT line 4, IT line 6, FR line 1, FR line 4, FR line 6, EN line 6

    6. 17.09.2026Cross-comparison test, two judges

      French had lost one item from the list

      While fixing a loan translation on 16 September we had dropped «boissons energisantes» from the French list so the line would fit the plate. The other three languages named five things, French four. Both judges saw it. With different punctuation all five fit: «energy drinks», which is what French calls them.

      Lines touched 09:00 FR line 9

    7. 17.09.2026Native reading with a third engine (Google AI Mode)

      Two loan translations in German, and a missing referent

      A third engine read every line in its own language alone, without the others beside it: that is the test loan translations need, and the one cross-comparison cannot give. In German «besetzt seinen Platz» echoed the Italian «ne prende il posto» (German says «nimmt seinen Platz ein»), and «die Summe nimmst du mit ins Bett» was the same loan translation we had already fixed in French on 16 September. In hour 09 the coffee line did not say what it was about: German and English named the coffee, Italian and French did not. All four name it now. The other languages followed: in French «deux, six et seize heures» dropped the word «heures» from the first two, which Italian allows and French does not, and «comme en plein jour» places an action in time rather than describing how someone drives; in English «the body asks for sleep» echoed «chiede il sonno» and becomes «calls for». Italian had no loan translations at all: it is the language the lines are written in.

      Lines touched 03:00 EN line 1 · 09:00 DE line 4, DE line 9, IT line 8, FR line 8, FR line 5 · 10:00 FR line 4

    8. 17.09.2026Check of the plate measurements

      Six lines did not fit the plate

      Every line has a limit: three plate lines, four for the closing line. Six were over, and on screen the text runs out of frame or shrinks. In the closing line of hour 09 the words «if you're sleepy» were dropped: they contradicted the line before, which says that after a coffee you FEEL awake and the sleep is only covered. Waiting until you feel sleepy is exactly what does not work. «Pull over» on its own is truer, and all four languages fit.

      Lines touched 05:00 FR line 5, DE line 5 · 09:00 IT line 11, FR line 11, DE line 11, EN line 11

    9. 27.09.2026Director's audit and an independent reviewer

      2 pm: the closing line could not contradict its own chart

      The closing line said ten minutes of sleep do more than thirty. The chart in the same episode, built from the study's numbers, shows that thirty minutes help too, only later: ten minutes win because they work at once. The closing line now names the action: at the first signs of sleepiness, pull over, ten minutes of sleep, a fifteen-minute alarm. The location changed with it, a car park instead of a square where you cannot stop, and so did the other plates, in four languages. At that point the video had not been filmed yet.

      Lines touched 14:00 IT line 1, FR line 1, DE line 1, EN line 1, IT line 2, FR line 2, DE line 2, EN line 2, IT line 3, FR line 3, DE line 3, EN line 3, IT line 4, FR line 4, DE line 4, EN line 4, IT line 5, FR line 5, DE line 5, EN line 5, IT line 6, FR line 6, DE line 6, EN line 6

    10. 28.09.2026Native-speaker review by George Zisa

      The hours still to be filmed, reviewed in English by a native speaker

      George Zisa reviewed the English lines of the hours still to be filmed and of those already online. For the hours still to be filmed we took his versions wherever they fit on the plate and keep the meaning of the source: 45 lines out of 48, and three titles («Dinner and drinks», «Night departure», «Approaching the low point»). Where we adjusted them, two judges checked the line again, working blind. For the hours already online his lines go in when the video is remade; until then the page says what the video says.

      Lines touched 00:00 EN line 1, EN line 2, EN line 3, EN line 5, EN line 6 · 01:00 EN line 1, EN line 2, EN line 3, EN line 4, EN line 5, EN line 6 · 02:00 EN line 2, EN line 3, EN line 4, EN line 5, EN line 6 · 19:00 EN line 1, EN line 2, EN line 3, EN line 4, EN line 5, EN line 6 · 20:00 EN line 1, EN line 2, EN line 3, EN line 4, EN line 5, EN line 6 · 21:00 EN line 1, EN line 2, EN line 3, EN line 4, EN line 5, EN line 6 · 22:00 EN line 1, EN line 2, EN line 3, EN line 4, EN line 5, EN line 6 · 23:00 EN line 1, EN line 2, EN line 3, EN line 4, EN line 5

    11. 28.09.2026Cross-check using two external engines

      French and German, checked by two external engines

      All the French and German boards of the 24 hours went through a blind review by two external engines, Google AI Mode and ChatGPT. Each deck held sentences from prevention bodies and one control error: both engines found every control error, 24 out of 24. We correct where the two engines agree, or where the error can be shown. In the hours still to be filmed we corrected 13 lines, which two judges then reviewed blind. A second round, with the Italian alongside for meaning, corrected 28 lines, and the closing line of hour 14 in Italian as well. A third round reread the rest of the French and German copy: notes, texts, source limits, info boxes and the diary. It corrected 157 page passages (148 in French and German, 9 in Italian): 71 for accents, umlauts and spacing only, 86 for meaning or language, checked by two judges working blind. In hour 11, the notes and text in French, German and English had been written separately: they are now translated from the Italian. In the hours already online, the lines and the two info boxes needing correction will be fixed when the video is remade.

      Lines touched 00:00 DE line 4, FR line 4 · 01:00 DE line 3, DE line 4 · 02:00 DE line 2 · 14:00 FR line 4, DE line 5, FR line 5, DE line 6, FR line 6, IT line 6 · 15:00 FR line 1, FR line 2, DE line 3, FR line 3, DE line 5, FR line 6 · 16:00 FR line 2, DE line 3 · 17:00 FR line 2, FR line 3, DE line 4, FR line 4 · 18:00 DE line 3, DE line 5 · 19:00 FR line 2, DE line 3, FR line 3, DE line 5, FR line 5 · 21:00 FR line 2, DE line 3, FR line 3 · 22:00 DE line 3, FR line 3 · 23:00 DE line 2, FR line 2, DE line 3, FR line 4, FR line 5, FR line 6

    12. 30.09.2026New version of the 1 pm, noon and 11 am episodes

      1 pm, noon and 11 am: the corrected lines go into the videos

      Three hours already online go back into the edit, and the lines corrected after their first release go in with them. At 1 pm the review says the dip «can come» even without lunch, as the source puts it, and the closing line names an action: plan a stop between one and three. The page adds a source from real driving, Horne and Reyner 1995, with its limitation. At noon the two lunches get their numbers, 305 and 922 calories. At 11 am George Zisa's English lines go in, together with the French and German corrections that were still waiting. The new lines were reviewed blind by two judges per language.

      Lines touched 11:00 EN line 2, DE line 4, FR line 5, DE line 5, EN line 5, EN line 6 · 12:00 FR line 4, DE line 4 · 13:00 IT line 3, FR line 3, DE line 3, EN line 3, IT line 6, FR line 6, DE line 6, EN line 6

    13. 01.10.2026New version of the 10 am, 9 am and 8 am episodes

      10 am, 9 am and 8 am: six boards and the corrected lines in the videos

      Three hours already online go back into the edit, and with them come the lines corrected after the first release. Hour 08 gets two revised French lines, two revised German lines and three English lines by George. Hours 09 and 10 get George's English lines and, in hour 10, one French line. Episodes 09 and 10 go from eleven and eight boards respectively to six: we keep lines that were already there and write no new ones.

      Lines touched 08:00 EN line 3, FR line 4, DE line 4, FR line 5, DE line 5, EN line 5, EN line 6 · 09:00 EN line 2, EN line 3, EN line 4, EN line 5, EN line 6 · 10:00 EN line 2, FR line 4, EN line 4, EN line 5, EN line 6

    14. 02.10.2026New version of the 7 am, 6 am, 5 am, 4 am and 3 am episodes

      7 am, 6 am, 5 am, 4 am and 3 am: a new voice and the corrected lines in the videos

      Five hours already online go back into the edit, and with them come the lines corrected after the first release. All five now include George's English lines; hour 04 gets one French line, hour 05 one French line and two German ones. The videos use the same lines as the pages, some of which were revised after the first release. Each hour keeps six boards: no new lines.

      Lines touched 03:00 EN line 3, EN line 4, EN line 5 · 04:00 FR line 2, EN line 2, EN line 3, EN line 4, EN line 6 · 05:00 EN line 1, FR line 3, DE line 3, EN line 3, EN line 4, EN line 5, DE line 6 · 06:00 EN line 1, EN line 5 · 07:00 EN line 1, EN line 2, EN line 4, EN line 5

    15. 02.10.2026Source re-read, the director's decision

      6 pm: the Swiss figure comes from the bfu's Sinus 2024 report

      Slide 4 said «about one crash in ten», based on two bfu pages without a DOI. It now says 7%: the share of motorway crashes with serious injuries that the bfu attributes, as the main cause, to tiredness and falling asleep (Sinus 2024, p. 63, 2019-2023 average; 1% of fatal crashes). The scope is narrower, motorways only, and the source has a DOI. The line remains an estimate: it is the cause written down by whoever recorded the crash.

      Lines touched 18:00 IT line 4, FR line 4, DE line 4, EN line 4

    On the site the text is already the new one. The video plates are remade in the next production window: until then the video may still show the old line.

    Reported, not yet resolved

    • 04:00 · EN line 6 «The absolute minimum is not fought. It is waited out.» A double passive with no agent, in a closing line. Not a grammar mistake but a choice of style: the English copywriter decides, not the test. On 28.09.2026 George Zisa chose the active form: «You don't fight the body's lowest point. You wait it out.» It goes in when the 4 am video is remade.
    • 11:00 · FR line 6 Italian says «ask whoever prescribed them»; French, German and English say «your doctor». In Switzerland the pharmacist advises too: Italian is the more precise one, and it is rather the other three that should widen. A direction call, not a language one.
    • 05:00 · FR line 1 The third engine called «poste de nuit» a loan translation. Checked: that is wrong, «poste de nuit» is correct, attested French. But in Swiss official documents (SECO, SUVA) the term is «equipe de nuit». Not a mistake, a register: whether to speak like the Swiss employer or like the dictionary.
    • 04:00 · DE line 2 I had changed «Wachheit» to «Wachsamkeit» on one plate, thinking I was repairing an inconsistency. Checking every field, I found that «Wachheit» appears thirteen times in the German, always in the same sense: it is the chosen term for the curve, not a slip. Changing it on a single plate created the inconsistency rather than removing it, and I have put it back. The real question stands: Italian, French and English say vigilanza, vigilance, alertness; German says Wachheit, which is the state of being awake. If it changes, it changes in all thirteen places, and that is a decision of direction and of science, not of language.

    Every line, in all four languages →

    Credits and materials

    Concept, editorial direction and directing: Matteo Fieni. Production and development: Opening®.

    Drive Awake gives voice to research and to professionals, and opens the conversation to the public.

    Press office →   Public kit →