{"slug": "2026-01-06-ensure-your-work-is-a-meaningful-contribution", "title": "2026-01-06 ensure your work is a meaningful contribution", "summary": "A developer reflects on the importance of delivering proven, working code rather than merely generating large patches with AI assistance. The post emphasizes that agent-generated code is cheap and that meaningful contributions require testing, honest naming, and self-consistent abstractions. It warns that sloppy PRs cost more than they save and that human accountability remains essential.", "body_md": "| # meaningful contribution | |\n| Based on embarassing feedback on a bad PR for AXM-10598, and written from my beliefs; with only a little bit of copy pasting, because to be human is to copy from a shitload of places until it becomes your own. | |\n| ## A contribution is not code, it's proven working code | |\n| > \"your job is to deliver code you have proven to work.\" — [simon willison](https://simonwillison.net/2025/Dec/18/code-proven-to-work/) | |\n| agent-generated code is cheap. anyone can prompt an LLM to produce a thousand-line patch. that's not valuable. what's valuable is contributing code that demonstrably works, has been tested, and doesn't shift burden to reviewers. | |\n| ## what qualifies as meaningful | |\n| ### 1. proven correctness | |\n| if you haven't tested it, it doesn't work. | |\n| if it happens to work, that's luck. | |\n| to actually prove something works, we: | |\n| - you tested manually, and saw it work. | |\n| - you wrote an automated test for what you did manually. | |\n| - you saw this automated test fail when you reverted you change. | |\n| - you tested the edges, you explicitly defined what happens outside the happy path. | |\n| ### 2. self-consistent abstractions | |\n| you created a mental model, great, | |\n| does it make sense when you look closer? | |\n| does it fit with the wider mental model? | |\n| naming something `VersionedStructuredRequestWithOptions` and then passing unversioned requests through it is confusing. names are contracts. | |\n| i don't know how to prove that something makes sense, but a good start is: | |\n| - you can explain it in plain english; from start to finish, and each part in isolation. | |\n| - you can explicitly articulate the expected inputs and outputs, and what happens if one of those is unexpected. | |\n| ## what doesn't qualify | |\n| ### slop indicators | |\n| - PR descriptions that read like \"summarize this for me\" prompts | |\n| - missing tests | |\n| - contradictions in abstractions, both against themselves and against the wider context. | |\n| - names that lie about what they contain, if you need a comment or \"x but it's actually y or x+y\", its lying. | |\n| ### the speed trap | |\n| speed without quality is negative value. a sloppy PR costs more reviewer and maintenance time than it saves. the \"i'll fix it later\" debt compounds negativelly, the \"i made it well\" value compounds positivelly. | |\n| > we lost track of the goal. we sacrificed quality in pursuit of speed, and for what? | |\n| ## the review standard | |\n| ask yourself before submitting: | |\n| 1. have i seen this work? not \"does the code look right\"—have i actually run it? | |\n| 2. do the types tell the truth? or am i lying to the compiler and hoping reviewers don't notice? | |\n| 3. is the naming honest? would someone reading this in six months be confused? | |\n| 4. did i test the edges? what happens when usage deviates from the happy path? when it's the WORST path? | |\n| 5. would i be confident to walk colleagues through my changes? | |\n| if the answer to any of these is \"no\" or \"i'm not sure,\" the contribution isn't ready. | |\n| ## agent-assisted work requires more scrutiny, not less | |\n| agents make it easy to produce large volumes of code quickly. this doesn't reduce your responsibility—it increases it. the code still has your name on it. | |\n| one pass from an agent is rarely enough. read it over. improve it. run it. test it. do this multiple times if needed. | |\n| ## the accountability loop | |\n| a computer can never be held accountable. that's your job as the human. almost anyone can generate a thousand-line patch. what's valuable is proving it works. | |\n| next time you submit a PR, include the evidence. |", "url": "https://wpnews.pro/news/2026-01-06-ensure-your-work-is-a-meaningful-contribution", "canonical_source": "https://gist.github.com/bdsqqq/1e7e6f454271d5f856a1176d0e800d89", "published_at": "2026-08-24 23:45:36+00:00", "updated_at": "2026-08-25 03:13:37.968884+00:00", "lang": "en", "topics": ["developer-tools", "ai-agents", "ai-ethics"], "entities": ["Simon Willison"], "alternates": {"html": "https://wpnews.pro/news/2026-01-06-ensure-your-work-is-a-meaningful-contribution", "markdown": "https://wpnews.pro/news/2026-01-06-ensure-your-work-is-a-meaningful-contribution.md", "text": "https://wpnews.pro/news/2026-01-06-ensure-your-work-is-a-meaningful-contribution.txt", "jsonld": "https://wpnews.pro/news/2026-01-06-ensure-your-work-is-a-meaningful-contribution.jsonld"}}