site banner
PaperclipPerfector  18hr ago (text post)   2926 thread views

Culture War Roundup for the week of April 6, 2026

This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.

Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.

We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:

  • Shaming.

  • Attempting to 'build consensus' or enforce ideological conformity.

  • Making sweeping generalizations to vilify a group you dislike.

  • Recruiting for a cause.

  • Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.

In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:

  • Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.

  • Be as precise and charitable as you can. Don't paraphrase unflatteringly.

  • Don't imply that someone said something they did not say, even if you think it follows from what they said.

  • Write like everyone is reading and you want them to be included in the discussion.

On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.

Top Bottom Old Controversial Comments
Jump in the discussion.

No email address required.

I have no idea why Claude Code is working so badly for you. I work at a FAANG-level company, and a huge amount of our code is written by Claude. Garry Tan is in AI psychosis, but Claude Code is easily the biggest productivity unlock in CS since I started my career.

Few recommendations:

  • What thinking mode are you using ? Use at least high or max.
  • For the purpose of this test, give it all permissions and link it to an mcp like context7
    • This allows it to independently read documentation on your local and from remote sources
  • Basic, but update the app. This lapse happened to a very smart coworker of mine.
  • Use plan mode. It allows the model to build an intuition for the problem before it goes off on its own
  • If you want specific behaviors, then ask for that. Something like:
    • State and scrutinize your assumptions explicitly
    • Consider and invalidate counter factuals.
    • Utilize coding patterns that have already been established in the repo.
    • Ideally, ask it to go write readme.md files for core utility dirs in your repo, so it doesn't cold start
  • Pair it with a type checker / linter and add it as a post-model hook
    • In python land, ruff & based-pyright are the tools of choice.
    • I have used pre-defined open source linting rules, which allows the model to implement best-practice behaviors (eg: opinionated null checks) without human intervention.

I've noticed that the quality of the codebase plays a huge role in the model's ability to write effective code.

For ex:

It assumes that all endpoints return plaintext or JSON content, even though several return binary data, CSV data, etc.

Ideally, all endpoints will already be typed. The model should not have to guess the request-response types.

Unless there is a specific regression in Claude Code, I don't know why claude failed at your task. It should have worked.

Also, if you're looking for a model that prioritizes meticulousness, then I'd use codex. Codex has a tendency to autistically cover all of your bases, that benefits the sort of problem you're work with (again, Use in high or xhigh mode).

and a huge amount of our code is written by Claude. Garry Tan is in AI psychosis, but Claude Code is easily the biggest productivity unlock in CS since I started my career.

That's weird because in my experience, Codex 5.4 is way better than the most recent Sonnet. Haven't tried Opus though.

I have no idea why Claude Code is working so badly for you.

I'm not @ChickenOverlord, but I'm also seeing unimpressive results. Maybe we can get to the bottom of it.

I've tried Claude (via Claude Code), Gemini (via Gemini CLI), and GPT (via codex).

In all of them, I've used their equivalent of Claude.md/Agents.md to lay ground rules of how we expect the agent to behave. Multiple people have taken multiple shots at this.

We always use plan mode first.

Our documentation is markdown in the same repository, so that should be useful and accessible.

We're using Java, which is strongly typed and all our endpoints are annotated with additional openapi annotations that should provide even more metadata.

We're using a pretty basic bitch tech stack, but it's not spring boot. All three models regularly fight us on that fact.

We have a four levels of validation, each with their own entry point in the build scripts. These are described in a readme.md in the root of the project. The first is a linter. The second is unit tests and code coverage. The third is a single end to end test. The fourth is all end to end tests. We have instructed the models to use these validation targets to check their work.

Despite all this, we see common failure modes across all models we've tested.

  1. Bad assumptions about the tech stack. No, we do not use spring boot.
  2. A tendency to add more code, rather than fix code.
  3. An urge to "fix" "bad tests" that exist for very specific reasons. These specific reasons are usually covered with inline developer documentation as well.
  4. Confusion about what capabilities our version of java has available. Yeah, the pattern matching preview was cool. Stop trying to turn it on with experimental feature flags.
  5. Writing tests that don't actually test the thing it's changing.

I'm sure there are more, but these immediately come to mind. There are four of us trying to make these things work, and we all keep running into the same problems again and again. It's not just me - even people with dramatically different writing styles and thought processes are seeing the same thing. I feel like I'm taking crazy pills, because a lot of people I know in real life are experiencing the same pain, but on the Internet it seems like I'm a huge outlier.

What's the disconnect here?

What's the disconnect here?

It works a lot better if you bend to the AI and use a stack it likes. Why this specific Java stack?

I hope that isn't a serious suggestion. A programming tool which only works if you use a particular framework is a shitty tool.

Legacy concerns. The amount of custom code that has built up over the last 15 years is too big of a shift to deal with right now. It's on the backlog, but not anywhere near the top priority.

I'm sure there are more, but these immediately come to mind. There are four of us trying to make these things work, and we all keep running into the same problems again and again. It's not just me - even people with dramatically different writing styles and thought processes are seeing the same thing. I feel like I'm taking crazy pills, because a lot of people I know in real life are experiencing the same pain, but on the Internet it seems like I'm a huge outlier.

My most competent co-worker, a Russian guy who got his start writing assembly back in the 80's, was the most enthusiastic about/interested in AI person that I knew. He was always trying out the latest models from OpenAI, Google, and Anthropic. He was also running his own local LLMs and diffusion models locally. He even dropped $4-5k on a DGX Spark late last year. And even he seems to be getting disillusioned/losing interest in AI, he doesn't seem to think it's going to be able to achieve anything remotely close to the promises and hype. Though I will note that the push from our upper management to use AI hasn't pleased him much either, especially since the project we've been working on for the past year (modernizing a giant mess created by our Indian coworkers. They weren't using package management at all, they were literally emailing around zip files full of DLLs for years, I got pulled into 4 hour long calls to fix dependency conflict issues in prod once every 2 or 3 months) was very much not aided by AI, but management insisted we find a way to use AI on the project regardless.

What is this place?

This website is a place for people who want to move past shady thinking and test their ideas in a court of people who don't all share the same biases. Our goal is to optimize for light, not heat; this is a group effort, and all commentators are asked to do their part.

The weekly Culture War threads host the most controversial topics and are the most visible aspect of The Motte. However, many other topics are appropriate here. We encourage people to post anything related to science, politics, or philosophy; if in doubt, post!

Check out The Vault for an archive of old quality posts. You are encouraged to crosspost these elsewhere.


Why are you called The Motte?

A motte is a stone keep on a raised earthwork common in early medieval fortifications. More pertinently, it's an element in a rhetorical move called a "Motte-and-Bailey", originally identified by philosopher Nicholas Shackel. It describes the tendency in discourse for people to move from a controversial but high value claim to a defensible but less exciting one upon any resistance to the former. He likens this to the medieval fortification, where a desirable land (the bailey) is abandoned when in danger for the more easily defended motte. In Shackel's words, "The Motte represents the defensible but undesired propositions to which one retreats when hard pressed."

On The Motte, always attempt to remain inside your defensible territory, even if you are not being pressed.


New post guidelines

If you're posting something that isn't related to the culture war, we encourage you to post a thread for it. A submission statement is highly appreciated, but isn't necessary for text posts or links to largely-text posts such as blogs or news articles; if we're unsure of the value of your post, we might remove it until you add a submission statement. A submission statement is required for non-text sources (videos, podcasts, images).

Culture war posts go in the culture war thread; all links must either include a submission statement or significant commentary. Bare links without those will be removed.

If in doubt, please post it!


Rules


Recommended Posts And Communities

Recommended Realtime Chats