{"course":{"slug":"choosing-your-tools","title":"Choosing and Using the Tools","url":"https://addaly.com/learn/choosing-your-tools","level":"beginner","track":"work"},"lesson":{"slug":"deciding-and-writing-it-down","title":"Making the decision, and recording it","url":"https://addaly.com/learn/choosing-your-tools/deciding-and-writing-it-down","position":19,"of":76,"minutes":7,"answer":"Write down the choice, the exact model strings, the date, and the specific conditions that would change your mind, because a decision you have not recorded gets silently re-litigated every time somebody publishes a benchmark.","keyIdea":"Write down the choice, the exact model strings, the date, and the specific conditions that would change your mind, because a decision you have not recorded gets silently re-litigated every time somebody publishes a benchmark."},"body":{"format":"markdown","text":"## Deciding is a separate act from testing\n\nPeople run a comparison, form an impression, and then keep half-deciding for the next three months: trying the loser again, reading a thread about it, wondering. The test was fine; the decision never happened.\n\nA decision has three parts, and all three are written down:\n\n1. **What you chose**, with the exact model strings and the date.\n2. **What you chose it for.** Rarely one tool for everything. Usually a default, a second opinion, and something local for confidential material.\n3. **What would change your mind.** Named in advance. This is the part almost everybody skips and it is the one that saves the most time.\n\n## The record\n\n```\nDECISION 2026-03-14\ndefault:    toolB paid ($20/mo) — best on long documents, my main daily work\nsecond:     toolA free tier — factual second opinion, and when B is capped\nlocal:      qwen3-8b via Ollama — anything with a client name in it\nnot chosen: toolC — 14/24, weak in Hindi, refuses ordinary clinical questions\n\nwould change my mind:\n  - toolA reaches parity on long documents (test prompts 4 and 5)\n  - toolB's price rises above $30, or Hindi quality drops\n  - a local model at 8b passes prompts 7-9\n  - my work shifts to mostly code\n\nrevisit: 2026-09-14 (in calendar)\n```\n\nEight lines. Its whole job is to stop you re-running the argument in your head, and it does that because you can look at it.\n\n## Why \"what would change my mind\" matters\n\nWithout it, every announcement is potentially decision-relevant, so you read all of them and act on some of them at random. With it, almost everything is filtered instantly: a new model is interesting, but does it bear on prompts 4 and 5? Usually no, and you get on with your work.\n\nIt also protects against the opposite failure, which is quieter and more common. Having paid for something, people defend the choice and stop noticing that the free tier caught up eighteen months ago. Writing the condition down while you are still neutral is the only reliable way to notice when it is met.\n\n## Reviewing the decision without redoing it\n\nBetween the six-month checks, most new information should hit the \"would change my mind\" list and bounce off. Two things deserve an immediate look rather than waiting: a price change in your currency, and a change in the data terms for the tool you paste work material into. Both are decisions somebody else made about your arrangement, and neither waits for your calendar.\n\nEverything else (a launch, a benchmark, a thread saying the model has become lazy) goes on the list to check in September. If it is real, it will still be real then, and you will have your twelve prompts to measure it with instead of an argument.\n\n## Deciding not to decide\n\nSometimes the honest outcome is that two tools are equivalent for your work. That is a real finding, and the correct response is to pick one on a tiebreak that has nothing to do with quality:\n\n- Which is easier to pay for from your country, in your currency?\n- Which has better terms for the data you handle?\n- Which works acceptably on a poor connection or an old phone?\n- Which does your team, your school or your client already use?\n- Which one can you leave most easily if you need to?\n\nNone of these are quality questions and all of them will matter more, over two years, than a two-point difference on twelve prompts.\n\n## The three-line version\n\nMost people will not write eight lines. Write three:\n\n```\n2026-03-14  using: toolB paid + toolA free.  toolC weak in Hindi.\n            switch if toolA handles long docs, or if B goes over $30.\n            recheck: September.\n```\n\nAnything is enormously better than nothing, because in September you will not remember which model you tested, what score it got, or why you ruled out the one you nearly chose. Nobody does. That is not a memory failing; it is what six months does to a detail that seemed obvious at the time.\n\nPut the file somewhere you will find it: next to your prompt file, in whatever folder you actually open. Then close it and go back to work. The point of doing this properly once is that you get to stop thinking about it."},"check":{"question":"Why does naming in advance the conditions that would change your mind matter more than recording which tool won?","options":[{"text":"It filters future announcements against a fixed test, and it is how you notice a paid choice going stale.","correct":true},{"text":"Because providers change models frequently, so the winner is likely to be obsolete before the conditions are met.","correct":false,"tempting":"Assumes the record expires quickly, when the model strings and scores are exactly what makes a later change detectable."},{"text":"Because it produces a rigorous experiment, converting an informal comparison into something with a stated hypothesis.","correct":false,"tempting":"Borrows the language of experiments for a decision aid; twelve prompts scored by one person is not made rigorous by writing a condition down."},{"text":"Because it lets you justify the choice to colleagues who prefer a different tool.","correct":false,"tempting":"Casts the record as an argument to win rather than a trigger to watch for, which is the use that actually saves time."}]},"prev":{"slug":"testing-refusals","title":"Testing the edges: refusals and house manners","url":"https://addaly.com/learn/choosing-your-tools/testing-refusals"},"next":{"slug":"web-app-phone-app-desktop-app","title":"The web app, the phone app and the desktop app","url":"https://addaly.com/learn/choosing-your-tools/web-app-phone-app-desktop-app"},"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) 'Making the decision, and recording it', in Choosing and Using the Tools. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/choosing-your-tools/deciding-and-writing-it-down (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":"Making the decision, and recording it","courseTitle":"Choosing and Using the Tools","module":{"index":2,"slug":"testing-it-yourself","title":"Testing a tool yourself"},"answer":null,"keyIdea":"Write down the choice, the exact model strings, the date, and the specific conditions that would change your mind, because a decision you have not recorded gets silently re-litigated every time somebody publishes a benchmark.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"deciding-is-a-separate-act-from-testing","runs":[{"text":"Deciding is a separate act from testing"}]},{"t":"p","runs":[{"text":"People run a comparison, form an impression, and then keep half-deciding for the next three months: trying the loser again, reading a thread about it, wondering. The test was fine; the decision never happened."}]},{"t":"p","runs":[{"text":"A decision has three parts, and all three are written down:"}]},{"t":"ol","start":1,"items":[[{"text":"What you chose","b":true},{"text":", with the exact model strings and the date."}],[{"text":"What you chose it for.","b":true},{"text":" Rarely one tool for everything. Usually a default, a second opinion, and something local for confidential material."}],[{"text":"What would change your mind.","b":true},{"text":" Named in advance. This is the part almost everybody skips and it is the one that saves the most time."}]]},{"t":"h2","id":"the-record","runs":[{"text":"The record"}]},{"t":"code","lang":"","label":"Text","text":"DECISION 2026-03-14\ndefault:    toolB paid ($20/mo) — best on long documents, my main daily work\nsecond:     toolA free tier — factual second opinion, and when B is capped\nlocal:      qwen3-8b via Ollama — anything with a client name in it\nnot chosen: toolC — 14/24, weak in Hindi, refuses ordinary clinical questions\n\nwould change my mind:\n  - toolA reaches parity on long documents (test prompts 4 and 5)\n  - toolB's price rises above $30, or Hindi quality drops\n  - a local model at 8b passes prompts 7-9\n  - my work shifts to mostly code\n\nrevisit: 2026-09-14 (in calendar)"},{"t":"p","runs":[{"text":"Eight lines. Its whole job is to stop you re-running the argument in your head, and it does that because you can look at it."}]},{"t":"h2","id":"why-what-would-change-my-mind-matters","runs":[{"text":"Why “what would change my mind” matters"}]},{"t":"p","runs":[{"text":"Without it, every announcement is potentially decision-relevant, so you read all of them and act on some of them at random. With it, almost everything is filtered instantly: a new model is interesting, but does it bear on prompts 4 and 5? Usually no, and you get on with your work."}]},{"t":"p","runs":[{"text":"It also protects against the opposite failure, which is quieter and more common. Having paid for something, people defend the choice and stop noticing that the free tier caught up eighteen months ago. Writing the condition down while you are still neutral is the only reliable way to notice when it is met."}]},{"t":"h2","id":"reviewing-the-decision-without-redoing-it","runs":[{"text":"Reviewing the decision without redoing it"}]},{"t":"p","runs":[{"text":"Between the six-month checks, most new information should hit the “would change my mind” list and bounce off. Two things deserve an immediate look rather than waiting: a price change in your currency, and a change in the data terms for the tool you paste work material into. Both are decisions somebody else made about your arrangement, and neither waits for your calendar."}]},{"t":"p","runs":[{"text":"Everything else (a launch, a benchmark, a thread saying the model has become lazy) goes on the list to check in September. If it is real, it will still be real then, and you will have your twelve prompts to measure it with instead of an argument."}]},{"t":"h2","id":"deciding-not-to-decide","runs":[{"text":"Deciding not to decide"}]},{"t":"p","runs":[{"text":"Sometimes the honest outcome is that two tools are equivalent for your work. That is a real finding, and the correct response is to pick one on a tiebreak that has nothing to do with quality:"}]},{"t":"ul","items":[[{"text":"Which is easier to pay for from your country, in your currency?"}],[{"text":"Which has better terms for the data you handle?"}],[{"text":"Which works acceptably on a poor connection or an old phone?"}],[{"text":"Which does your team, your school or your client already use?"}],[{"text":"Which one can you leave most easily if you need to?"}]]},{"t":"p","runs":[{"text":"None of these are quality questions and all of them will matter more, over two years, than a two-point difference on twelve prompts."}]},{"t":"h2","id":"the-three-line-version","runs":[{"text":"The three-line version"}]},{"t":"p","runs":[{"text":"Most people will not write eight lines. Write three:"}]},{"t":"code","lang":"","label":"Text","text":"2026-03-14  using: toolB paid + toolA free.  toolC weak in Hindi.\n            switch if toolA handles long docs, or if B goes over $30.\n            recheck: September."},{"t":"p","runs":[{"text":"Anything is enormously better than nothing, because in September you will not remember which model you tested, what score it got, or why you ruled out the one you nearly chose. Nobody does. That is not a memory failing; it is what six months does to a detail that seemed obvious at the time."}]},{"t":"p","runs":[{"text":"Put the file somewhere you will find it: next to your prompt file, in whatever folder you actually open. Then close it and go back to work. The point of doing this properly once is that you get to stop thinking about it."}]}],"check":{"question":[{"text":"Why does naming in advance the conditions that would change your mind matter more than recording which tool won?"}],"options":[{"text":[{"text":"It filters future announcements against a fixed test, and it is how you notice a paid choice going stale."}],"correct":true,"tempting":null},{"text":[{"text":"Because providers change models frequently, so the winner is likely to be obsolete before the conditions are met."}],"correct":false,"tempting":[{"text":"Assumes the record expires quickly, when the model strings and scores are exactly what makes a later change detectable."}]},{"text":[{"text":"Because it produces a rigorous experiment, converting an informal comparison into something with a stated hypothesis."}],"correct":false,"tempting":[{"text":"Borrows the language of experiments for a decision aid; twelve prompts scored by one person is not made rigorous by writing a condition down."}]},{"text":[{"text":"Because it lets you justify the choice to colleagues who prefer a different tool."}],"correct":false,"tempting":[{"text":"Casts the record as an argument to win rather than a trigger to watch for, which is the use that actually saves time."}]}],"order":[3,1,0,2],"why":null},"prev":{"slug":"testing-refusals","title":"Testing the edges: refusals and house manners","minutes":8},"next":{"slug":"web-app-phone-app-desktop-app","title":"The web app, the phone app and the desktop app","minutes":8}}}