Search Ideas

I think you defeated my criticism. That’s kind of cool, I get it.

This could extend beyond fixing errors to improving page load times, search quality, and reducing churn, provided you can measure whether a change helped. My friend Samay co-wrote this about a similar loop using production feedback and evals to improve tax agents:
https://openai.com/index/building-self-improving-tax-agents-with-codex/

Kitty Comes Home: my first game made with Claude Opus 5.5.

You play as a little kitten trying to find your way back to your mom.

Play for free here: https://blog.dennishackethal.com/posts/kitty-comes-home

Teaser trailer: https://youtu.be/bcM2BI-41cY

This takes me back to my startup days in Berlin. It was common knowledge among founders that anyone overly worried about telling people about their great idea because someone might ‘steal’ it was a newbie founder. It’s almost always a mistake not to tell people, for several reaasons:

1. Your idea probably isn’t new anyway. Somebody has probably already tried it before you and realized why it doesn’t work.
2. If you don’t share your idea, you don’t get any critical feedback. You’ll make more mistakes. It will be harder to launch it.
3. Most people don’t care. They have their own problems. They’re not looking to copy ideas.

Tik Tok did reels and then Instagram copied the feature. Arguably, Instagram does it better now. Imagine the founder of Tik Tok getting angry with Zuckerberg and trying to shame him online. That would be ridiculous.

As Ben points out, reels go all the way back to Vine! Which corroborates point 1 above: the idea probably isn’t new. And your customers don’t care either way. They only care that your product solves their problems better than the competitors’ products.

People get angry because they hope to control more than they actually can. Ideas spread; we can’t control people’s speech. One can’t expect secrecy from people who didn’t agree to secrecy. If it’s important, like industry secrets or product launches, one should make people sign NDAs. If something is truly secret, the only person you should tell is your lawyer. Aside from that, sharing generously and openly is more profitable for you 99% of the time anyway.

It also pays to be brief: the more you write, the more text there is to criticize. It’s in Vincent’s own interest to be less verbose. Not sure he realizes that.

It also pays to be concise: the more you write, the more text there is to criticize. It’s in Vincent’s own interest to be less verbose. Not sure he realizes that.

I guess, the point I'm trying to make is really that it should not be coming with any form of punishment (like that this does not count for badges) IF the AI assist is clearly done in good faith.

Absence of a reward != punishment

These three paragraphs could be shortened to: ‘basing badges on non-AI use may discourage non-native speakers.’ Verbosity is a recurring issue

Fixed. The discussion title now uses title case: ‘Immortality, Billionaires, and Copying Business Ideas Is Not Immoral’.

Midnight Ride: Production Errors That Fix Themselves (Almost)

blog.dennishackethal.com

Midnight Ride: Production Errors That Fix Themselves (Almost)

Midnight Ride, my new Rails gem, sends production errors from Solid Errors to Telegram and starts a Claude Code session that opens a PR with the fix.

Reels go all the way back to Vine!

This takes me back to my startup days in Berlin. It was common knowledge among founders that anyone overly worried about telling people about their great idea because someone might ‘steal’ it was a newbie founder. It’s almost always a mistake not to tell people, for several reaasons:

1. Your idea probably isn’t new anyway. Somebody has probably already tried it before you and realized why it doesn’t work.
2. If you don’t share your idea, you don’t get any critical feedback. You’ll make more mistakes. It will be harder to launch it.
3. Most people don’t care. They have their own problems. They’re not looking to copy ideas.

Tik Tok came up with reels and then Instagram copied the feature. Arguably, Instagram does it better now. Imagine the founder of Tik Tok getting angry with Zuckerberg and trying to shame him online. That would be ridiculous.

People get angry because they hope to control more than they actually can. Ideas spread; we can’t control people’s speech. One can’t expect secrecy from people who didn’t agree to secrecy. If it’s important, like industry secrets or product launches, one should make people sign NDAs. If something is truly secret, the only person you should tell is your lawyer. Aside from that, sharing generously and openly is more profitable for you 99% of the time anyway.

We’re talking about a friend of yours who you think might have a problem, but who hasn’t even signed up never mind voiced that concern to me directly. This is all hypothetical. I’m not gonna relax the rules just because one person who isn’t even a member might possibly not feel accommodated. If some day ten non-native members actually try and make an effort and then all tell me the same thing, now that would be a good reason to relax the rules.

Eh. Being a non-native speaker myself, I have no sympathy for other non-native speakers. Tell your friend to work harder and get better at English. Badges are rewards for personal effort, and I can’t reliably tell that apart from AI usage. The badge logic stays until further notice.

That all makes sense to me, thanks. But there is a cost: my previous badges were removed which means that my contributions would have stayed limited to 3 per day if I understand the rules correctly. This should change of course, now, that I do not use AI for responses, so I should be fine in the log run, but my friend, who would then disclose that he uses AI for all his discussions, would never be able to submit more than 3 ideas per day, and would also never be able to join more than one discussion per week because none of his responses would count towards earning the necessary badges https://veritula.com/badges.

My criticism goes deeper though: I really think that there is a case to make that for the goal you describe, to find a bit of truth (or maybe just the current best explanation), it is actually better to use AI, simply because we can spend less time correcting bad faith, misunderstandings, unclear language etc. (Also, one might just say "well, so it takes a bit more time, no problem" - but it's more complicated, because bad faith is often provoking emotions etc which often ends debates before finding a bit more truth)
In our sample discussion about animal consciousness, if I would correct for the "too long" and "not simple enough", I think it would be helpful. But of course, now that I know that you don't appreciate it and since you are currently the only one responding, I will gladly follow your wish not to use it.

I guess, the point I'm trying to make is really that it should not be coming with any form of punishment (like that this does not count for badges) IF the AI assist is clearly done in good faith.

I’m not using AI for this response and will not do so moving forward on Veritula to respect the rules and what seems to be the sentiment behind these rules (unless they change).

As written, the forum rules don’t ban AI. They say to disclose it. Someone may voice a preference about AI use and it’s up to you to honor it. Their preference doesn’t bind you like the forum rules do. But members will make decisions about who to spend time discussing with. Our discussion about animals, if you choose to continue it, will probably go better if you don’t use AI.

That said, it’s a good idea, especially for new members, to get a feel for the platform, study the rules, and understand the culture better before going full-steam ahead and submitting dozens of AI-assisted ideas. It’s best to exercise care when entering a new space and prove that you’re worth engaging with. (I’m not only talking about you specifically, I mean new members in general.) Checking a checkbox indicating AI use is a small cost, and it’s a good way to be transparent with readers of one’s ideas.

To that end, it pays to show an understanding that Veritula isn’t a debate club, and that the goal isn’t to win an argument (although winning and losing do happen here, in an objectively recognizable sense – though finding an error is rarely an actual ‘loss’). The goal is to find a bit of truth. It pays to show that you can change your mind, eg by saying things like ‘I never thought of it that way – you’ve given me a new way to look at things.’ Having your AI say something like that on your behalf obviously carries less weight than if you said it yourself. And excessive use of AI gives the impression that someone is just trying to win an argument. (Not saying you specifically – again, this is about the forum rules in general.)

My English is not great but the English from a Spanish friend of mine is considerably worse, and I wonder if I should tell him about this platform, because he would usually use AI to polish his English.

Your English is fine. Clearly good enough to have discussions. As for your friend, he can enlist the help of AI to translate ideas. Again, he’d just need to disclose it.

AI-generated text is notoriously verbose regardless of who came up with an idea.

Summary feature proposed in #4727 is for discussions not ideas.

I think this did not address what I think is the key part of the question which is 'how can something have an intrinsic value if value is subjective'.

I simply do not think all value is subjective. One can alter value through their own subjective experience. I personally still hold on to a paper napkin from a sentimental trip that I went on, and that creates a personal premium value for that napkin that goes above the utility and scarcity conditions of that napkin's value.

The contrary is obviously true. I doubt there are many monks that intend on buying Ferraris, even if they had the means to do so, so their subjective value impacts what would be considered the total of a Ferrari.

However, to speak of the 'intrinsic' side. I think value is best split into a multiplicative relationship between utility and scarcity. A toll road is a great example. Many times, there aren't many alternative roads, if any, and its utility at getting you to where you need to go should be non-zero assuming you intend on going where that road takes you. So, if a toll road were to crop up on the market selling for $1 dollar, assuming no 'strings attached', that would be a clear mispricing one could diagnose on an 'intrinsic value basis'.

I think it's incorrect to say that something cannot be intrinsically valued due to subjectivity being the primary motivator for humans, and so since subjectivity drives exchange, valuing that relationship is not possible. We can have long lasting theories of human behaviour or social systems that allow us to assign an objective/intrinsic value.

It doesn’t play the whole codebase at once. It walks through the code line by line, and when a file changes, it jumps to the changed lines and plays an accent made from them, rising or falling with that file’s health. So a small change in a big codebase is as audible as one in a small codebase. What’s quieter in a big codebase is the overall stage, and that’s on purpose: one long function shouldn’t make a whole healthy codebase sound sick. Bugs are the exception, since they count in full. One syntax error drops even a huge project from beautiful to meh.

I think this is actually more urgent for the new crit tree view on the right side.

For ex., this just shows:

1. Not if the...
2. A discussion can...
3. How is this a...

It could say something like:

1. Not if concise
2. Long discussions
3. No real criticism

How is this a criticism? My point was not about preventing long discussions but that we don't need to ai-summarize ideas that are already super concise.

Relatedly, I think in the future, software engineering will mostly be you talking to voice agents and looking at diagrams the agent is generating and constantly updating to help you understand the code/ideas. This is kind of how students already learn in the classroom, where there's a teacher who talks them through a drawing on the board, etc.

 It responds live as you make changes to your code

This seems useful only for small codebases. If the music is composed of all the code in a large codebase, each PR would change only a small subset, making the difference in the music almost inaudible.

I think what you actually care about is whether the ideas in the content were generated by AI. Whether the text was written or edited by AI should not be a concern. Therefore, I suggest the checkbox should read: "Were the core ideas of this submission generated by AI?"

Asking whether the text was written by AI is equivalent to asking a musician whether they used Ableton (music software) or only analog instruments.