I was testing how LLMs rank products in our space by asking vague buyer questions like "what's the best tool for X." We showed up on Gemini, but ChatGPT kept listing older competitors that haven't updated in years.

It made me realize how inconsistent these AI recommendations are compared to traditional Google SEO.

If you added comparison pages, structured data, or an llms.txt file recently, did you notice any real difference in how models cite you?

An enlightening conversation with my eldest cousin sister currently living abroad who used to work for Infosys 15 years ago:
Me: "Ya so Infosys is still doing mass hiring in Indian colleges, so is Wipro."
Her: "Oh okay, what's the package these days?"
Me: "3-6 LPA CTC."
Her: "What?? This is what WE were paid!"
My other cousin sister worked for Wipro. She said the exact same.
And then Narayana Murthy, Billionaire, Founder of Infosys, says that Indians should have a 72-hour work week, i.e, 12 hours for 6 days.
This is the current state of India's tech sector. Freshers, experienced developers, everyone is struggling to find a job in this market, having to settle for absolute outdated minimum wage stipends. The country's premier engineering institutes are an export sector for global tech companies contributing to the GDP of first world countries, all while said companies setup shop in Bengaluru, Pune, Hyderabad, Gurgaon and outsource majority of their work to cheap Indian labor.
But by all means, lets purchase AI tokens and further add to the problem.

I spent an entire afternoon trying to get an LLM to write API documentation for a payment gateway. The first three outputs were full of summaries and fake endpoint URLs that looked professional but were completely useless.

The breakthrough happened when I stopped asking for text and fed the model raw JSON payloads along with three specific error response codes. Giving it constraints forced it to write accurate reference guides instead of high-level marketing fluff.

Documentation turnaround time went from four hours down to twenty minutes per release. The secret is treating the model like a technical editor rather, than an open-ended creative writer.

How do you structure your input prompts to get technical output from generative AI models?

I found that my automated coding assistants kept resetting my main environment template file every time I asked them to update dependencies. They acted like my config files were scratchpads erasing custom keys I needed for production. That was a headache.

The fix was simple: I set file-level permissions and added a clear exclusion list in the root instruction file. I made sure to tell the AI exactly which files were, off-limits. After that unintended changes stopped completely.

It saved me hours of debugging and made my local test runs consistent again. Now every time I use an agent the prompt includes a reminder to check file boundaries before running any script.

What method do you use to keep your coding tools from making unauthorized edits to sensitive system files?

The hardest thing to get right was getting people successfully through sign-up.

Initially I started with one time links, had a few successful sign ups and inbetween them would be a couple 'unauthenticated' emails. I would reach out via email to try find out why they didn't complete the process, but get no response. I eventually found out that some email providers burn the one time link when they scan the email.

One thing I've learned while building software is that getting a feature to work and making it reliable are two different challenges.

A feature might work perfectly during testing but behave differently when real users interact with it.

For example:

• A form works, but doesn't handle unexpected input properly.

• An API responds correctly, but fails when an external service becomes unavailable.

• A dashboard displays accurate information, but doesn't distinguish between current and outdated data.

• An AI feature produces a convincing answer, but doesn't have enough context to support its conclusion.

Hey PH

I'm 6 months into building my first product (still pre-PMF, ~20 active users), and user feedback has been the hardest thing to crack. Here's what I've tried so far and what I learned from each:

Post-mortem from last week.

My site is translated into 10 languages. The language switcher and the internal links on every localised page were supposed to keep visitors in their language. For months they did not. Every locale page quietly linked back to the English homepage.

First post here. Been browsing Product Hunt for a while without an account and finally made one.

I keep noticing apps adding an AI button to something that didn't need one. A calendar app with "AI scheduling." A to-do list with "AI prioritization." A notes app with "AI summaries." Most of the time I poke at it once, it does something vaguely fine, and then I go back to using the app exactly the way I did before. The button just sits there.

I'm not against AI features. Some of them are genuinely good and I use them. But there's a difference between a tool that was built around it and a tool that stuck a chat box on top because it was trendy. From the listing alone, those two look identical.

Curious if anyone else feels this. What's an AI feature you actually use regularly, versus one you tried once and forgot about? And does a listing mentioning AI make you more likely to try something, or less at this point?

I ve been trying different tools to find competitors for my product, but most either show irrelevant results or miss similar products entirely.

Curious how other founders approach this. Do you rely on tools, Google searches, or manual research?

I m also wondering if there s a better way to solve this. Would love to hear your thoughts and recommendations.