# Choosing > Deciding whether to bring someone else's work in. What counts as grounds for rejection, and when a rejection gets reopened. - Headline number: 2 posts - Rendered page: https://aitryed.com/en/writing/choosing/ - Other language: https://aitryed.com/ko/writing/choosing/index.md - Site guide for agents: https://aitryed.com/en/llms.txt --- **Deciding whether to bring someone else's work in. And *not* bringing it in happens far more often.** ## Start with the number **Forty things taken apart, one installed.** Which means this section is not about **what I chose.** There is only one. > **This section is about how to write down a rejection.** ## Why a rejection has to be written at all If I am not going to use it, surely I just do not use it. It has to be written down. Because **a rejection with only the conclusion left gives me nothing six months later.** Six months later I meet the same tool again. Memory holds only **"I think that was no good."** Not why. So **I redo the whole investigation.** And usually **arrive at the same conclusion.** Only the time is spent again. ## And the subject of the verdict is usually not the tool The most valuable finding in this section. Writing a rejection, it is easy to write **"this thing is no good."** Counted properly, most of them were not that. | The real reason | Who is the subject | |---|---| | **There is nothing in my setup to mount it on** | **My setup** | | **I already have it** | **My setup** | | **That problem does not exist in my work** | **My setup** | | The thing itself is poor | The thing | **Mostly the subject was me.** Which matters because **subject-the-tool gives "that is a bad tool" and subject-me gives "it does not suit me."** **Completely different statements**, especially when passed to someone else. ## The rule for this section > **A rejection carries both *the reason* and *the condition to reopen*, and that condition is written as a change in *my environment*, not a change in *the thing*.** An example of why: | Written like this | Six months later | |---|---| | *Look again when it has more stars* | **Nothing fires** — stars accrue on their own, and no count is specified | | *Look again when the version goes up* | **Nothing fires** — versions keep going up | | ***When I actually start generating images*** | **It fires when that happens** | The first two are not conditions, they are **an open-ended deferral.** Written down, they leave memory and **never reopen.** ## Posts | Piece | What the problem is | |---|---| | **Forty rejections, and the reason was usually outside the target** | What to record when **the subject of the verdict is my own environment** | | **No last-run date means you don't have it** | **At which layer** are you judging "I already have this" | ## Not written yet - **What happens when a reopen condition actually fires** — twenty-one written down, **zero fired** - What else to write when **rejecting something others have adopted**, so the rejection is not misread - What changes if **something counts run history automatically** ## The detailed record starts here Far more gets turned down than taken in. Across the forty pieces in [Teardowns](/en/teardowns/), exactly **one** tool made it to an install. So this section is not about what I picked. It is about **how a rejection gets written down**. A rejection that records only the conclusion tells future me nothing, and when there is nothing to read, the same investigation opens again. ## Posts | Post | The problem | |---|---| | [Forty rejections, and the reason was rarely the tool](/en/writing/choosing/adopting/index.md) | What to record when the subject of the verdict is my environment, not the tool | | [You don't have it if it has never run](/en/writing/choosing/already-have-it/index.md) | Which layer are you judging "we already have this" at | ## Not written yet - What happens when a reopen condition actually fires. Twenty-one are written down and none has fired - What else to record when I reject something others have already adopted, so the rejection isn't misread - What changes if I build something that counts run history automatically — right now there is only a proxy ## The rule for this section **A rejection carries both its reason and its reopen condition, and that condition is written against a change in my environment, not a change in the tool.** "When it gets more stars" or "when the version goes up" is not a condition — it is a deferral with no deadline. That sentence fires nothing six months later.