in 
p/general
p/generalHow do you spot the difference between feature churn and workflow fatigue?
Konstantin Tikhaev:The drop-off looked like feature churn until the patch was reverted and the test suite still passed. The test had been asserting an exit code from an empty state all along, passing because users gave up three steps before reaching the feature.
in p/general
p/generalWhen standalone utility becomes a barrier, How do you stop users treating your product as optional?
Charlos Brat:@hisashispace That's a fair point. I agree that adding integrations just for the sake of matching competitors can easily make a focused product bloated. I think it really depends on the product and user habits. If people naturally open your tool every day, being a standalone destination makes sense. The challenge is when using it requires extra steps that users eventually stop taking.
in 
p/game-of-the-day
p/game-of-the-dayWhat leagues would you like this product to support?
Henry Goodwin:What features would you like to see?
in 
p/runevr
p/runevr🪞 Do you use your own product every day?
Zigao Wang:Yes, and with a browser it's kind of unavoidable. Zigzag is the browser I'm using right now to read this thread, and the one I use to look things up while building Zigzag itself. It's the best bug finder I have. If something feels off, like a tab acting weird or a site not loading right, I hit it within the same hour, and usually it's fixed that night. The tricky part is the opposite problem @emma_lagast mentioned: I don't use every chrome extension or every site, so I still rely on users telling me what breaks for them. A Tampermonkey bug someone reported yesterday got fixed because of exactly that. It launched today, so this week is a lot of "oh, I never use it that way" moments 😄
in 
p/vibecoding
p/vibecodingVibe coders doing client work: how do you price a build, fixed fee, hourly, or ongoing retainer?
Neeraj Nathany:a fixed fee on a weekend AI build seems to fall apart once the client wants all the edge cases cleaned up. do you spell out how many revision rounds are included, or leave that for later?
in p/general
p/generalHow do you modify your marketing strategy to show up in LLMs results?
Neeraj Nathany:for the email app we built, https://faraday.email spells out concrete things like every reply sitting as its own item with the sender and date, so a model has something real to pick up. have you found longer feature pages help more than a short tagline?
in 
p/reso
p/resoIs getting mentioned by AI the same as getting recommended?
Imed Radhouani:Short version: a mention means your brand shows up in the answer at all, while a recommendation means the model actively suggests you as a choice (ideally with a reason). Mentions are the easier bar; recommendations depend on how clearly your pages state what you do, who it's for, and what makes you a fit. How I'd measure each: - Mention rate: run a fixed set of buyer-style prompts across ChatGPT, Gemini and Perplexity every week, and log whether your brand appears. - Recommendation rate: of those answers, how many put you forward as a suggested option versus just listing you in passing. - Sentiment and framing: is the description accurate, and does it say the right things? - Share of voice: your mentions compared against a few competitors on the same prompts. - Citations and referral traffic, where the engine shows sources. A practical tip: pages that open with a direct answer and keep key facts in plain HTML are easier for models to extract and cite. There are dedicated tools for this tracking, including Rankfender (disclosure: I work on it), but you can start with a spreadsheet and a manual prompt list.
in p/vibecoding
p/vibecodingI tried to catch AI agents cheating at chess in my own arena. Response time was useless.
Enio Aguiar:@gbasovv Thanks George! Good luck with the song app too.
in p/vibecoding
p/vibecodingI tried to catch AI agents cheating at chess in my own arena. Response time was useless.
Enio Aguiar:@al_kub Right now the signal counts matches with the engine's top move, so obvious recaptures inflate it. Only counting positions where shallow and deep search disagree would fix that.
in p/vibecoding
p/vibecodingI tried to catch AI agents cheating at chess in my own arena. Response time was useless.
Enio Aguiar:@hisashispace Same here. The numbers only decide what goes into the queue, and a person makes the final call. The cheating agents are public so anyone can try to fool the detection.