{"course":{"slug":"ai-careers","title":"Working in AI","url":"https://addaly.com/learn/ai-careers","level":"beginner","track":"work"},"lesson":{"slug":"the-six-hour-week","title":"A year on six hours a week","url":"https://addaly.com/learn/ai-careers/the-six-hour-week","position":10,"of":85,"minutes":8,"answer":"Progress comes from a small weekly amount that survives bad weeks, because every gap longer than about ten days costs a session to re-enter.","keyIdea":"Progress comes from a small weekly amount that survives bad weeks, because every gap longer than about ten days costs a session to re-enter."},"body":{"format":"markdown","text":"## Plan for the time you actually have\n\nMost people reading this have a job, or a family, or both, and four to eight hours a week. A plan that assumes forty is not a plan; it is a reason to feel guilty. Here is what a year at six hours looks like when it works.\n\n## The year, in blocks\n\n**Weeks 1 to 8: the tool.** Three hours practising, three hours building something tiny with it. For analytics that is SQL against a real dataset. For application work it is one small service that runs.\n\n**Weeks 9 to 16: the mess.** One messy real dataset. Not a cleaned teaching set. Municipal budgets, transport timetables, exam results, your own bank exports. Clean it, and write down every decision you made about it as you go, because you will not remember in April what you decided in February.\n\n**Weeks 17 to 26: ship it and find one user.** Deploy it. Get one real person to use it weekly. One is enough; one changes what the project is.\n\n**Weeks 27 to 36: the second project, with a measurement.** In your domain. This time build a test set of thirty cases and report a number.\n\n**Weeks 37 to 44: the search apparatus.** CV, case study pages, the list of forty employers, twenty outreach messages. This is work, and it needs its own weeks rather than being squeezed around the edges.\n\n**Weeks 45 to 52: interviews, and keep building.** The building matters here because it is what you talk about, and because a search with nothing else in it is corrosive.\n\nThat is fifty-two weeks with roughly ten weeks of slippage built in, because you will lose weeks to illness, deadlines and family. The plan assumes it.\n\n## Three rules that do more than the plan\n\n**The same slot every week.** A fixed time defended against everything else beats a flexible plan that loses to whatever else appears. Tuesday 20:00 to 22:00 and Sunday 07:00 to 11:00 is a real schedule. \"When I get a chance\" is not.\n\n**End every session by writing the next action.** One line, in the file: `NEXT: the join is duplicating rows — check whether order_items has more than one row per order.` This is the single highest-return habit in the whole plan. Without it, the first twenty minutes of every session are spent rebuilding context.\n\n**Never start a session by reading news.** Feeds are engineered to hold you, your two hours will become forty minutes, and you will feel as though you studied.\n\n## The re-entry cost, which is the real reason bursts fail\n\nA three-week gap does not cost three weeks of progress. It costs a whole session, because you must re-read your own code and rediscover what you were doing before you can add one line. Two gaps a month and half your year is spent re-entering.\n\nThat is the mechanism. It is why two hours weekly beats twelve hours once a month, even though the monthly total looks similar. Memory of your own reasoning decays faster than memory of facts.\n\n> Figure: What a gap costs when you sit back down. Line chart of Minutes of a 2-hour session lost against Days since the last session: Re-entry cost, a solo project with a NEXT line in the file (3, 10) (7, 20) (10, 45) (14, 90) (21, 120) (30, 120). Past about ten days the whole of a two-hour session goes on re-reading your own code and rediscovering what you were doing. Two such gaps a month and half the year is spent re-entering, which is why two hours weekly beats twelve hours monthly.\n\n## If your only computer is a phone\n\nA large share of the people reading this are on a phone. Be honest about what that allows.\n\n**Genuinely possible on a phone:** reading; SQL practice on sites that run in a browser; writing your case study; drafting outreach messages; reviewing code on GitHub; Google Colab and Kaggle notebooks in a mobile browser, awkwardly.\n\n**Not realistically possible:** building and debugging anything of size. Not because you lack discipline, but because editing, running and reading errors in a 6-inch viewport costs you several times the keystrokes and most of your working memory.\n\nSo the plan needs a few keyboard hours a week from somewhere: a cybercafé, a public library, a friend's machine, your workplace computer at lunch, a college lab. Two focused keyboard hours plus four phone hours is a working week. Pretending a phone is enough is how a year disappears.\n\nThe free tools that make those keyboard hours count, because they need no installation and no payment:\n\n- **Google Colab**: free Python with a browser, sessions time out after a period of inactivity and the disk is wiped, so push to GitHub before you close it.\n- **Kaggle notebooks**: free, roughly thirty GPU hours a week, and datasets already attached.\n- **GitHub Codespaces**: a real editor and terminal in a browser, on a free monthly allowance of core-hours.\n- **DB Fiddle, SQLime, or DuckDB in the browser**: SQL with nothing installed.\n\nThe paid alternatives (a laptop, a cloud VM, a subscription notebook service) are better and are not required to start.\n\n## Measure finished things, not hours\n\nHours studied is a comforting and useless number. Count things finished: a query that answers a real question, a script that runs on a schedule, a page written, a message sent, a bug found. A month with zero finished things did not happen, no matter how many hours went into it.\n\n## The limitation\n\nThis plan produces a hireable beginner in a market that is hiring beginners. It does not defeat a market with no junior openings. If twelve months of honest work produces nothing, the problem is more likely the market or the employer categories you are targeting than your effort: go back to the employer map before you conclude anything about yourself."},"check":{"question":"Tomás has four hours a week. He does nothing for three weeks while a work deadline runs, then puts in twelve hours over one weekend to catch up. After four months he feels he has learned almost nothing, though his total hours match a steady schedule. What is the main mechanism?","options":[{"text":"Twelve-hour sessions exceed his concentration limit, so most of that weekend was low-quality study.","correct":false,"tempting":"Blames fatigue within the session, which is real but secondary to what the gaps themselves cost."},{"text":"Each long gap forces him to spend most of the next session rebuilding context (re-reading his own code and rediscovering his reasoning) so a large share of his total hours is spent returning to where he already was.","correct":true},{"text":"He is studying the wrong material, and a better curriculum would make the schedule irrelevant.","correct":false,"tempting":"Reaches for content quality, the explanation that leads to buying another course rather than changing the pattern that is actually costing the time."},{"text":"Four hours a week is simply too little, and there is no schedule that works at that level.","correct":false,"tempting":"Concludes the amount is the problem when the distribution is; steady four-hour weeks do produce progress."}]},"prev":{"slug":"the-ladder-and-the-levels","title":"Levels, titles, and why \"senior\" means nothing on its own","url":"https://addaly.com/learn/ai-careers/the-ladder-and-the-levels"},"next":{"slug":"data-analyst-in-depth","title":"The data analyst","url":"https://addaly.com/learn/ai-careers/data-analyst-in-depth"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'A year on six hours a week', in Working in AI. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-careers/the-six-hour-week (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"A year on six hours a week","courseTitle":"Working in AI","module":{"index":1,"slug":"the-terrain","title":"The terrain: who hires, for what, and where you fit"},"answer":null,"keyIdea":"Progress comes from a small weekly amount that survives bad weeks, because every gap longer than about ten days costs a session to re-enter.","hero":{"kind":"figure","figure":{"t":"figure","label":"Figure","figure":{"type":"line","title":"What a gap costs when you sit back down","series":[{"label":"Re-entry cost, a solo project with a NEXT line in the file","points":[[3,10],[7,20],[10,45],[14,90],[21,120],[30,120]]}],"xLabel":"Days since the last session","yLabel":"Minutes of a 2-hour session lost","caption":"Past about ten days the whole of a two-hour session goes on re-reading your own code and rediscovering what you were doing. Two such gaps a month and half the year is spent re-entering, which is why two hours weekly beats twelve hours monthly."},"alt":"Figure: What a gap costs when you sit back down. Line chart of Minutes of a 2-hour session lost against Days since the last session: Re-entry cost, a solo project with a NEXT line in the file (3, 10) (7, 20) (10, 45) (14, 90) (21, 120) (30, 120). Past about ten days the whole of a two-hour session goes on re-reading your own code and rediscovering what you were doing. Two such gaps a month and half the year is spent re-entering, which is why two hours weekly beats twelve hours monthly."}},"blocks":[{"t":"h2","id":"plan-for-the-time-you-actually-have","runs":[{"text":"Plan for the time you actually have"}]},{"t":"p","runs":[{"text":"Most people reading this have a job, or a family, or both, and four to eight hours a week. A plan that assumes forty is not a plan; it is a reason to feel guilty. Here is what a year at six hours looks like when it works."}]},{"t":"h2","id":"the-year-in-blocks","runs":[{"text":"The year, in blocks"}]},{"t":"p","runs":[{"text":"Weeks 1 to 8: the tool.","b":true},{"text":" Three hours practising, three hours building something tiny with it. For analytics that is SQL against a real dataset. For application work it is one small service that runs."}]},{"t":"p","runs":[{"text":"Weeks 9 to 16: the mess.","b":true},{"text":" One messy real dataset. Not a cleaned teaching set. Municipal budgets, transport timetables, exam results, your own bank exports. Clean it, and write down every decision you made about it as you go, because you will not remember in April what you decided in February."}]},{"t":"p","runs":[{"text":"Weeks 17 to 26: ship it and find one user.","b":true},{"text":" Deploy it. Get one real person to use it weekly. One is enough; one changes what the project is."}]},{"t":"p","runs":[{"text":"Weeks 27 to 36: the second project, with a measurement.","b":true},{"text":" In your domain. This time build a test set of thirty cases and report a number."}]},{"t":"p","runs":[{"text":"Weeks 37 to 44: the search apparatus.","b":true},{"text":" CV, case study pages, the list of forty employers, twenty outreach messages. This is work, and it needs its own weeks rather than being squeezed around the edges."}]},{"t":"p","runs":[{"text":"Weeks 45 to 52: interviews, and keep building.","b":true},{"text":" The building matters here because it is what you talk about, and because a search with nothing else in it is corrosive."}]},{"t":"p","runs":[{"text":"That is fifty-two weeks with roughly ten weeks of slippage built in, because you will lose weeks to illness, deadlines and family. The plan assumes it."}]},{"t":"h2","id":"three-rules-that-do-more-than-the-plan","runs":[{"text":"Three rules that do more than the plan"}]},{"t":"p","runs":[{"text":"The same slot every week.","b":true},{"text":" A fixed time defended against everything else beats a flexible plan that loses to whatever else appears. Tuesday 20:00 to 22:00 and Sunday 07:00 to 11:00 is a real schedule. “When I get a chance” is not."}]},{"t":"p","runs":[{"text":"End every session by writing the next action.","b":true},{"text":" One line, in the file: "},{"text":"NEXT: the join is duplicating rows — check whether order_items has more than one row per order.","code":true},{"text":" This is the single highest-return habit in the whole plan. Without it, the first twenty minutes of every session are spent rebuilding context."}]},{"t":"p","runs":[{"text":"Never start a session by reading news.","b":true},{"text":" Feeds are engineered to hold you, your two hours will become forty minutes, and you will feel as though you studied."}]},{"t":"h2","id":"the-re-entry-cost-which-is-the-real-reason-bursts-fail","runs":[{"text":"The re-entry cost, which is the real reason bursts fail"}]},{"t":"p","runs":[{"text":"A three-week gap does not cost three weeks of progress. It costs a whole session, because you must re-read your own code and rediscover what you were doing before you can add one line. Two gaps a month and half your year is spent re-entering."}]},{"t":"p","runs":[{"text":"That is the mechanism. It is why two hours weekly beats twelve hours once a month, even though the monthly total looks similar. Memory of your own reasoning decays faster than memory of facts."}]},{"t":"h2","id":"if-your-only-computer-is-a-phone","runs":[{"text":"If your only computer is a phone"}]},{"t":"p","runs":[{"text":"A large share of the people reading this are on a phone. Be honest about what that allows."}]},{"t":"p","runs":[{"text":"Genuinely possible on a phone:","b":true},{"text":" reading; SQL practice on sites that run in a browser; writing your case study; drafting outreach messages; reviewing code on GitHub; Google Colab and Kaggle notebooks in a mobile browser, awkwardly."}]},{"t":"p","runs":[{"text":"Not realistically possible:","b":true},{"text":" building and debugging anything of size. Not because you lack discipline, but because editing, running and reading errors in a 6-inch viewport costs you several times the keystrokes and most of your working memory."}]},{"t":"p","runs":[{"text":"So the plan needs a few keyboard hours a week from somewhere: a cybercafé, a public library, a friend’s machine, your workplace computer at lunch, a college lab. Two focused keyboard hours plus four phone hours is a working week. Pretending a phone is enough is how a year disappears."}]},{"t":"p","runs":[{"text":"The free tools that make those keyboard hours count, because they need no installation and no payment:"}]},{"t":"ul","items":[[{"text":"Google Colab","b":true},{"text":": free Python with a browser, sessions time out after a period of inactivity and the disk is wiped, so push to GitHub before you close it."}],[{"text":"Kaggle notebooks","b":true},{"text":": free, roughly thirty GPU hours a week, and datasets already attached."}],[{"text":"GitHub Codespaces","b":true},{"text":": a real editor and terminal in a browser, on a free monthly allowance of core-hours."}],[{"text":"DB Fiddle, SQLime, or DuckDB in the browser","b":true},{"text":": SQL with nothing installed."}]]},{"t":"p","runs":[{"text":"The paid alternatives (a laptop, a cloud VM, a subscription notebook service) are better and are not required to start."}]},{"t":"h2","id":"measure-finished-things-not-hours","runs":[{"text":"Measure finished things, not hours"}]},{"t":"p","runs":[{"text":"Hours studied is a comforting and useless number. Count things finished: a query that answers a real question, a script that runs on a schedule, a page written, a message sent, a bug found. A month with zero finished things did not happen, no matter how many hours went into it."}]},{"t":"h2","id":"the-limitation","runs":[{"text":"The limitation"}],"role":"limit"},{"t":"p","runs":[{"text":"This plan produces a hireable beginner in a market that is hiring beginners. It does not defeat a market with no junior openings. If twelve months of honest work produces nothing, the problem is more likely the market or the employer categories you are targeting than your effort: go back to the employer map before you conclude anything about yourself."}]}],"check":{"question":[{"text":"Tomás has four hours a week. He does nothing for three weeks while a work deadline runs, then puts in twelve hours over one weekend to catch up. After four months he feels he has learned almost nothing, though his total hours match a steady schedule. What is the main mechanism?"}],"options":[{"text":[{"text":"Twelve-hour sessions exceed his concentration limit, so most of that weekend was low-quality study."}],"correct":false,"tempting":[{"text":"Blames fatigue within the session, which is real but secondary to what the gaps themselves cost."}]},{"text":[{"text":"Each long gap forces him to spend most of the next session rebuilding context (re-reading his own code and rediscovering his reasoning) so a large share of his total hours is spent returning to where he already was."}],"correct":true,"tempting":null},{"text":[{"text":"He is studying the wrong material, and a better curriculum would make the schedule irrelevant."}],"correct":false,"tempting":[{"text":"Reaches for content quality, the explanation that leads to buying another course rather than changing the pattern that is actually costing the time."}]},{"text":[{"text":"Four hours a week is simply too little, and there is no schedule that works at that level."}],"correct":false,"tempting":[{"text":"Concludes the amount is the problem when the distribution is; steady four-hour weeks do produce progress."}]}],"order":[3,0,2,1],"why":null},"prev":{"slug":"the-ladder-and-the-levels","title":"Levels, titles, and why “senior” means nothing on its own","minutes":8},"next":{"slug":"data-analyst-in-depth","title":"The data analyst","minutes":9}}}