News for hackers. Stuff that matters.

biduskamil

Thanks for feedback! A couple of things:

Temperature settings only matter for api usage. Nevertheless, stochastic nature of LLM responses does produce a distribution of responses for a single prompt query. It could potentially be a good idea to run the same query couple of times in each iteration of the monitoring tool to get a better look at the distribution of responses.

As for live search, we focus pretty much only on queries that refer to brands or products. Such queries do use web search tool almost 100% of the time so you will not encounter the stale training data issue in our tool.

Happy to discuss it in more details if you want

Re:Show HN: How Loud Is My Hotel? on Wednesday June 24, 2026 @07:27AM (#48658157)
I want that but for apartments - when you're looking for a flat to have such score next to the flat I'm looking at
Re:Ask HN: How are you finding work/gigs as a SWE? on Wednesday June 24, 2026 @07:24AM (#48658119)
Niche local technology/ai communities seem to be places with a lot of alpha at the moment when it comes to jobs
Current limitation: we don't extract shopping carousels and product suggestions though that's WIP

Show HN: Estimating what users ask ChatGPT about your company

https://github.com/syntropicsignal-ai/ai-visibility-audit

OpenAI, Google and Perplexity don't expose user query analytics. It's difficult to tell if and why your company appears in chat responses.Turns out the only way you find out currently if you're cited at scale is if you reverse engineer the prompts users might be asking about your brand and track ChatGPT/Gemini/etc. responses for them.I've built a pipeline, that:1. crawls your website and your...

( Read More... | 3 comments | on HN )
Working on adding support for shopping carousels now

Show HN: Open-source tool for reverse engineering ChatGPT queries about brands

https://github.com/syntropicsignal-ai/ai-visibility-audit

Hi, I'm Kamil and I'm a founder of Applied AI agency in Warsaw, Poland.There is a lot of projects in AI visibility/GEO/AI SEO space. I wanted to see what is really possible in terms of how and when ChatGPT, Gemini, Perplexity, etc. reference certain brands.Turns out it's all pretty much about reverse engineering the prompts users write because the platforms don't share any analytics as of...

( Read More... | 1 comment | on HN )
forgot about the Show HN prefix.. hiding the post
Re:Show HN: Local personal data redaction for any AI tools on Thursday June 18, 2026 @08:32AM (#48584368)
Local is the way. Any benchmarks on latency it has on CPU?
Hi, I'm Kamil and I'm a founder of Applied AI agency in Warsaw, Poland.

There is a lot of projects in AI visibility/GEO/AI SEO space. We wanted to see what is really possible in terms of how and when ChatGPT, Gemini, Perplexity, ec. reference certain brands.

Turns out it's all pretty much about reverse engineering the prompts users write because the platforms don't share any analytics as of today.

So we've built a tool that does reverse engineering of prompts with the highest accuracy by first examining your website, your top searched keywords and phrases. It then juxtaposes them with the closest WildChat questions (collection of 1 million conversations between human users and ChatGPT) and uses all this to generate the most realistic set of prompts for you to use when checking ai visibility and sentiment for your brand.

The tool is open source and available on Syntropic's github: https://github.com/syntropicsignal-ai/ai-visibility-audit

We target the tool primarily for e-commerce stores as this seems to be a gap in the market.

I'm curious to hear what you think about this.

Show HN: Voice gender classifier for European voice AI (1MB, ONNX, 4ms)

https://huggingface.co/syntropicsignal-ai/gender-voice-classifier

Hi, I'm Kamil and I'm a founder of Applied AI agency in Warsaw, Poland.We've trained a small <1MB voice classifier model that runs on CPU in 4ms. Can be run next to silero VAD in voice AI deployments.What we noticed in production deployments of voice assistants in Contact Centers in EU is that human consultants pick up immediately how to inflect verbs and ajdectives after one utterance from...

( Read More... | 2 comments | 3 points | on HN )