<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Hacker News: vood</title><link>https://news.ycombinator.com/user?id=vood</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Fri, 20 Nov 2026 22:24:45 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=vood" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[Show HN: I analyzed 1,300 Google searches – Reddit appeared in 83%]]></title><description><![CDATA[
<p>Built a scraper over a few days, spent a few hundred dollars on API costs. Queried 1,300 keywords across 26 industries against Google and several AI search platforms.<p><pre><code>  Google Search results:

  - Reddit in 83% of product-related queries
  - 62% of those in positions 1-3
  - Commercial intent queries ("best X", "X vs Y"): 93.6%

  AI Search was more interesting:

  - ChatGPT: ~20% Reddit citations
  - YouTube cited more often than Reddit
  - Perplexity: nearly zero

  That last one surprised me until I remembered Reddit sued Perplexity in October 2025. External research from Q2 2025 showed 47% Reddit citations on Perplexity. My data from January 2026 shows it's basically gone.

  My hypothesis on why Google is doing this:

  74% of new webpages now contain AI-generated content. Google seems to be responding by weighting sources that are harder to fake. Reddit has real users with post history. YouTube requires actual video production. Both are expensive to game at scale.

  This isn't necessarily good — Reddit has its own problems with astroturfing and mod capture. But the signal-to-noise ratio is apparently still better than the open web.

  Raw data and methodology - follow the link.

  Curious if others are seeing similar patterns, or if my methodology is flawed somewhere.</code></pre></p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=46510233">https://news.ycombinator.com/item?id=46510233</a></p>
<p>Points: 2</p>
<p># Comments: 0</p>
]]></description><pubDate>Tue, 06 Jan 2026 09:14:04 +0000</pubDate><link>https://mentioned.to/research</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=46510233</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46510233</guid></item><item><title><![CDATA[Ask HN: Best back end to run models on Google TPU?]]></title><description><![CDATA[
<p>So, I got Pixel 10 Pro, and I'd like to run parakeet (or whisper) model on it for voice to text. I'm building an ai dictation app (aidictation.com). I'm struggling to find a way to run this model on device. I have to reserve to use groq API, which is suboptimal.<p>Any recommendations?</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=46268766">https://news.ycombinator.com/item?id=46268766</a></p>
<p>Points: 8</p>
<p># Comments: 0</p>
]]></description><pubDate>Mon, 15 Dec 2025 00:24:20 +0000</pubDate><link>https://news.ycombinator.com/item?id=46268766</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=46268766</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46268766</guid></item><item><title><![CDATA[Show HN: AIDictation – zero data retention dictation app]]></title><description><![CDATA[
<p>Hi HN,<p>I built AIDictation.com, a voice to text app written in Swift. It sends audio to my own backend, runs it through a Whisper-based pipeline, and returns a transcription you can then send straight into an AI chat like ChatGPT or Claude.<p>I’ve been building full‑stack apps for ~20 years, but this is my first Swift application. I leaned heavily on AI coding tools to get from zero Swift to a working app and backend in a couple of weeks.<p>What it does<p>Records audio and sends it to my server.
The backend runs a pipeline using Whisper V3 Turbo + OpenAI GPT OSS 120B.<p>I intentionally went with a cloud pipeline instead of on‑device models so I can:
- Parallelize work on the backend and tune the pipeline.
- Mix and match providers and models.
- Improve latency without shipping new app versions.<p>After transcription, there’s a “share to AI chat” flow so you can send it with one tap to ChatGPT, Claude, etc.<p>Context rules
One feature I missed in Whisper Flow was configurable context rules (similar to the Super Whisper Modes). AIDictation lets you define how transcription should behave depending on what you’re doing.<p>For example:
- Meetings: keep speaker names and timestamps.
- Coding: preserve technical terms and code formatting.
- Journaling: be more forgiving, add punctuation, make the text more readable.
- You can configure different presets and switch between them.<p>Why cloud instead of on‑device<p>A lot of apps focus on running models locally. I chose the opposite trade‑off:
- Provider flexibility: right now I’m using the Groq API because, in my tests, it had the best end‑to‑end latency (700-800ms), but the backend is built to swap providers and models.
- This does mean audio leaves the device, so I tried to be very explicit about data handling.<p>No registration needed. You get about 2,000 words per month for free without creating an account or giving an email.<p>Tech stack
Client: Swift (first real Swift/iOS app I’ve shipped).
Backend: NodeJS on Vercel.
Models: Whisper V3 Turbo + OpenAI GPT OSS 120B.
Provider: Groq API at the moment, mainly for latency reasons.<p>I’ve been using AIDictation daily for the past couple of weeks, and I’m happy with it so far, but I’d really like candid feedback from HN—both on the product and on the implementation.</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=46094955">https://news.ycombinator.com/item?id=46094955</a></p>
<p>Points: 2</p>
<p># Comments: 0</p>
]]></description><pubDate>Sun, 30 Nov 2025 08:40:30 +0000</pubDate><link>https://aidictation.com</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=46094955</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46094955</guid></item><item><title><![CDATA[New comment by vood in "Show HN: Nano PDF – A CLI Tool to Edit PDFs with Gemini's Nano Banana"]]></title><description><![CDATA[
<p>Congratulations on the release; that's a really good job.</p>
]]></description><pubDate>Sun, 30 Nov 2025 06:38:14 +0000</pubDate><link>https://news.ycombinator.com/item?id=46094402</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=46094402</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46094402</guid></item><item><title><![CDATA[New comment by vood in "Ask HN: What Are You Working On? (Nov 2025)"]]></title><description><![CDATA[
<p><a href="https://voicesinmyhead.co/" rel="nofollow">https://voicesinmyhead.co/</a> - AI-powered voice dictation. 5x faster than typing.</p>
]]></description><pubDate>Mon, 10 Nov 2025 07:00:04 +0000</pubDate><link>https://news.ycombinator.com/item?id=45873192</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=45873192</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45873192</guid></item><item><title><![CDATA[New comment by vood in "Ask HN: How would you set up a child’s first Linux computer?"]]></title><description><![CDATA[
<p>This is such a weird thing to read. You try to project your ideas on the kids, thinking that is the best thing to do. Let them be.<p>I learned Linux when I was like 13 or 14 and not because my father told me. He didn't know much about computers in early 2000/late 90ties.<p>The curiosity, the desire to learn, the need to set up my own isp, the need to start to make money, the curiosity of how html, php and other stuff worked let me to Linux.<p>Teach them how to be curious and feed that curiosity, the rest will happen.<p>And if they choose Mac over Linux, just get of their way, otherwise they will rebel.</p>
]]></description><pubDate>Sun, 09 Nov 2025 15:52:24 +0000</pubDate><link>https://news.ycombinator.com/item?id=45866419</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=45866419</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45866419</guid></item><item><title><![CDATA[New comment by vood in "We built a modern data stack from scratch and reduced our bill by 70%"]]></title><description><![CDATA[
<p>Rolling your own generally has mainly downsides in the context they are in. 
1. This is clearly a small team with very little spend
2. Tomorrow someone leaves and next engineer will have to manage all of this. 
3. I don't think they realize that they actually increased cost of this service not decreased it. Now they need to manage their own Kafka monthly. Engineering time is expensive.</p>
]]></description><pubDate>Sun, 09 Mar 2025 22:12:20 +0000</pubDate><link>https://news.ycombinator.com/item?id=43314436</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=43314436</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=43314436</guid></item><item><title><![CDATA[Show HN: Open-Source H-Index]]></title><description><![CDATA[
<p>Have you heard of Hirsch index for academic researchers and publications?<p>We adapted the algorithm for open source contributors. Fun project!<p>Open Source Hirsch Index is equal to N if a developer has at least N projects, each with at least N commits from the developer, and each project has at least N stars.<p>Hirsch ≥ 27 is like winning a Nobel Prize—only a few hundred people reach this level.
Hirsch ≥ 13 puts you in the top 10,000, which is on par with earning an Olympic medal.
Hirsch ≥ 8 places you in the top 100,000, where you’ll find people making a serious impact in tech—think billionaires and CTOs of unicorn companies.
Hirsch ≥ 3 lands you in the top 1 million, the marathoners of the industry.</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=41987491">https://news.ycombinator.com/item?id=41987491</a></p>
<p>Points: 4</p>
<p># Comments: 0</p>
]]></description><pubDate>Tue, 29 Oct 2024 17:59:33 +0000</pubDate><link>https://www.getprog.ai/open-source-h-index</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=41987491</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41987491</guid></item><item><title><![CDATA[Show HN: I've built a split view UI to measure LLM performance]]></title><description><![CDATA[
<p>Checkout the video where I measure Mixtral 8x22 vs DBRX Instruct. TLDR; the latter is 3x faster.</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=40017203">https://news.ycombinator.com/item?id=40017203</a></p>
<p>Points: 2</p>
<p># Comments: 0</p>
]]></description><pubDate>Fri, 12 Apr 2024 20:18:40 +0000</pubDate><link>https://twitter.com/avysotsky/status/1778513840977973365</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=40017203</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40017203</guid></item><item><title><![CDATA[New comment by vood in "Ask HN: Is anybody getting value from AI Agents? How so?"]]></title><description><![CDATA[
<p>I do. I use assistants as containers for different conversations for my GTM work:
An assistant for marketing and copywriting
An assistant for customer support
An assistant sales conversations.<p>These agents aren't super smart: just few PDFs for context plus a few sentences system prompts.<p>I do get what I want in 80% of use cases (not measured, just a feeling).</p>
]]></description><pubDate>Sun, 31 Mar 2024 18:08:28 +0000</pubDate><link>https://news.ycombinator.com/item?id=39886532</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=39886532</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=39886532</guid></item><item><title><![CDATA[New comment by vood in "Claude 3 model family"]]></title><description><![CDATA[
<p>This is a weird demand to have in my opinion. You have plenty of applications on your computer and they only do what they were designed for. You can't ask a note taking app (even if it's open soured) to do video editing, unless you modify the code.</p>
]]></description><pubDate>Tue, 05 Mar 2024 03:30:49 +0000</pubDate><link>https://news.ycombinator.com/item?id=39599123</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=39599123</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=39599123</guid></item><item><title><![CDATA[New comment by vood in "Show HN: I made an app to use local AI as daily driver"]]></title><description><![CDATA[
<p>Thanks for sharing ChatbotUI. While I'm not an author, I use it extensively and contribute to it. Thanks to the permissive license, I could offer ChatbotUI as a hosted solution with our API keys. <a href="https://labs.writingmate.ai" rel="nofollow">https://labs.writingmate.ai</a>.</p>
]]></description><pubDate>Wed, 28 Feb 2024 12:03:54 +0000</pubDate><link>https://news.ycombinator.com/item?id=39536805</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=39536805</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=39536805</guid></item><item><title><![CDATA[New comment by vood in "Representation Engineering: Mistral-7B on Acid"]]></title><description><![CDATA[
<p>This is very well written and entertaining post. I enjoyed reading it.<p>Selfishly, would you mind sharing literature or blog posts that led you to this level of understanding of LLMs? I'm trying hard to understand the inner workings via experiments but definitely far behind your expertise.<p>Thanks</p>
]]></description><pubDate>Sun, 18 Feb 2024 06:54:53 +0000</pubDate><link>https://news.ycombinator.com/item?id=39416844</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=39416844</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=39416844</guid></item><item><title><![CDATA[New comment by vood in "Veritasium: The SAT Question Everyone Got Wrong [video]"]]></title><description><![CDATA[
<p>I'm surprised that only 3 people reported the error.<p>While problem is not trivial, my initial reaction was "it can't be 3, it's too obvious". And it seems that a lot of people in the comments here do get to the right answer on their own.<p>So, why not hundreds of reports?</p>
]]></description><pubDate>Wed, 27 Dec 2023 04:24:40 +0000</pubDate><link>https://news.ycombinator.com/item?id=38779083</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=38779083</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=38779083</guid></item><item><title><![CDATA[New comment by vood in "Ask HN: What are you grateful for?"]]></title><description><![CDATA[
<p>I'm grateful to have parents who in their age, have energy, health and financial freedom to join us for celebrations from overseas. I'm grateful for my parents to be friends to me and all my friends.<p>As someone said here, I'm extremely grateful for free quality higher education I have received in my home country. And I'm saddened that many are in student debts for decades.<p>I'm grateful to be in the best health I've been in last 10 years.<p>I'm grateful to myself for quitting alcohol more than a year ago and sticking with it.<p>I'm grateful, for the first time, to feel financially stable for many years to come.<p>I'm grateful to United States of America for being our home and safe harbor for the past 10 years, especially last 3, given the global instability.</p>
]]></description><pubDate>Mon, 25 Dec 2023 03:39:56 +0000</pubDate><link>https://news.ycombinator.com/item?id=38759481</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=38759481</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=38759481</guid></item><item><title><![CDATA[New comment by vood in "Show HN: nbi.ai – Generative Business Intelligence"]]></title><description><![CDATA[
<p>Hey folks, I'm a free customer here and a big-big fan of their project. I've been using narrative.bi for my project and sharing the results on Twitter [1] lately.<p>How I'm using narrative.bi:
- A high-level report of WoW, and MoM changes on our website traffic that land in my email.
- Alternative to GA4 UI, which drives me nuts.<p>What I don't get yet and I want to see implemented (I'll definitely pay for this):
- Dimensional analysis: Which dimensions contribute to the topline change the most. I.e. Say unique users increased by 30% WoW, what top 5 dimensions have contributed to this change? Country? Device? Etc
- Global event correlation: are there any global events that influence the metrics change? Maybe a holiday in the Country?<p>[1] <a href="https://twitter.com/avysotsky/status/1723403222906286185" rel="nofollow noreferrer">https://twitter.com/avysotsky/status/1723403222906286185</a></p>
]]></description><pubDate>Fri, 17 Nov 2023 22:41:40 +0000</pubDate><link>https://news.ycombinator.com/item?id=38311583</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=38311583</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=38311583</guid></item><item><title><![CDATA[Show HN: Create Google Docs, Slides, Sheets with GPT-4 Turbo]]></title><description><![CDATA[
<p>A chrome extension that allows you to communicate with GPT-4 Turbo in a context of the webpage and export results in Google Workspace tools</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=38202833">https://news.ycombinator.com/item?id=38202833</a></p>
<p>Points: 5</p>
<p># Comments: 1</p>
]]></description><pubDate>Thu, 09 Nov 2023 09:27:47 +0000</pubDate><link>https://writingmate.ai/</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=38202833</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=38202833</guid></item><item><title><![CDATA[New comment by vood in "How to build and grow Chrome extension to over 300k active users"]]></title><description><![CDATA[
<p>An interview about challenges and learnings of building chrome extensions.</p>
]]></description><pubDate>Thu, 07 Sep 2023 18:06:34 +0000</pubDate><link>https://news.ycombinator.com/item?id=37423096</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=37423096</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=37423096</guid></item><item><title><![CDATA[How to build and grow Chrome extension to over 300k active users]]></title><description><![CDATA[
<p>Article URL: <a href="https://thesaasengineer.substack.com/p/building-and-promoting-chrome-extensions">https://thesaasengineer.substack.com/p/building-and-promoting-chrome-extensions</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=37423095">https://news.ycombinator.com/item?id=37423095</a></p>
<p>Points: 13</p>
<p># Comments: 3</p>
]]></description><pubDate>Thu, 07 Sep 2023 18:06:34 +0000</pubDate><link>https://thesaasengineer.substack.com/p/building-and-promoting-chrome-extensions</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=37423095</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=37423095</guid></item><item><title><![CDATA[Show HN: WritingMate – Send ChatGPT Answers to Google Docs, Sheets, Gmail]]></title><description><![CDATA[
<p>WritingMate.ai has neat new feature - it makes possible to share ChatGPT responses (in it's own interface) to Google Sheets, Docs and Gmail.<p>WritingMate.ai is a popular minimalistic ChatGPT extension for Chrome.</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=37156470">https://news.ycombinator.com/item?id=37156470</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Thu, 17 Aug 2023 02:32:38 +0000</pubDate><link>https://twitter.com/writingmateai/status/1691984071583560029</link><dc:creator>vood</dc:creator><comments>https://news.ycombinator.com/item?id=37156470</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=37156470</guid></item></channel></rss>