<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: goncharom</title><link>https://news.ycombinator.com/user?id=goncharom</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Thu, 19 Nov 2026 21:08:17 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=goncharom" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by goncharom in "Ask HN: What Are You Working On? (December 2025)"]]></title><description><![CDATA[
<p>A multi-purpose scrapper to turn any webpage into structured data: <a href="https://news.ycombinator.com/item?id=45870231">https://news.ycombinator.com/item?id=45870231</a><p>It uses LLMs to generate python code to scrap a webpage to fit any Pydantic model provided:<p><pre><code>  from hikugen import HikuExtractor
  from pydantic import BaseModel
  from typing import List
  
  class Article(BaseModel):
      title: str
      author: str
      published_date: str
      content: str
  
  class ArticlePage(BaseModel):
      articles: List[Article]
  
  extractor = HikuExtractor(api_key="your-openrouter-api-key")
  
  result = extractor.extract(
      url="https://example.com/articles",
      schema=ArticlePage
  )
  
  for a in result.articles:
      print(a.title, a.author)</code></pre></p>
]]></description><pubDate>Sun, 14 Dec 2025 21:32:12 +0000</pubDate><link>https://news.ycombinator.com/item?id=46267169</link><dc:creator>goncharom</dc:creator><comments>https://news.ycombinator.com/item?id=46267169</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46267169</guid></item><item><title><![CDATA[New comment by goncharom in "Schizophrenia sufferer mistakes smart fridge ad for psychotic episode"]]></title><description><![CDATA[
<p>Relevant read (not my own): <a href="https://simone.org/advertising/" rel="nofollow">https://simone.org/advertising/</a></p>
]]></description><pubDate>Sat, 06 Dec 2025 16:01:01 +0000</pubDate><link>https://news.ycombinator.com/item?id=46174311</link><dc:creator>goncharom</dc:creator><comments>https://news.ycombinator.com/item?id=46174311</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46174311</guid></item><item><title><![CDATA[New comment by goncharom in "A new AI winter is coming?"]]></title><description><![CDATA[
<p>it has a proper paper attached right at the beginning of the article</p>
]]></description><pubDate>Mon, 01 Dec 2025 19:37:35 +0000</pubDate><link>https://news.ycombinator.com/item?id=46112041</link><dc:creator>goncharom</dc:creator><comments>https://news.ycombinator.com/item?id=46112041</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46112041</guid></item><item><title><![CDATA[New comment by goncharom in "A new AI winter is coming?"]]></title><description><![CDATA[
<p>As I said then, and probably echoing what other commenters are saying - what do you mean by understanding when you say computers understand nothing? do humans understand anything? if so, how?</p>
]]></description><pubDate>Mon, 01 Dec 2025 18:11:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=46110783</link><dc:creator>goncharom</dc:creator><comments>https://news.ycombinator.com/item?id=46110783</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46110783</guid></item><item><title><![CDATA[New comment by goncharom in "A new AI winter is coming?"]]></title><description><![CDATA[
<p>Every time I see comments like these I think about this research from anthropic: <a href="https://www.anthropic.com/research/mapping-mind-language-model" rel="nofollow">https://www.anthropic.com/research/mapping-mind-language-mod...</a><p>LLMs activate similar neurons for similar concepts not only across languages, but also across input types. I’d like to know if you’d consider that as a good representation of “understanding” and if not, how would you define it?</p>
]]></description><pubDate>Mon, 01 Dec 2025 17:30:19 +0000</pubDate><link>https://news.ycombinator.com/item?id=46110210</link><dc:creator>goncharom</dc:creator><comments>https://news.ycombinator.com/item?id=46110210</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46110210</guid></item><item><title><![CDATA[New comment by goncharom in "Greggit – Google but it's only the Reddit results"]]></title><description><![CDATA[
<p>Yes this is literally just appending site: reddit.com to the query and redirecting to google. This page is a single HTML: <a href="https://github.com/goncharom/greggit" rel="nofollow">https://github.com/goncharom/greggit</a><p>This was meant as a silly project! I caught myself adding site:reddit.com to a lot of google searches so I figured I’d just make a shortcut.</p>
]]></description><pubDate>Fri, 28 Nov 2025 00:37:20 +0000</pubDate><link>https://news.ycombinator.com/item?id=46074488</link><dc:creator>goncharom</dc:creator><comments>https://news.ycombinator.com/item?id=46074488</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46074488</guid></item><item><title><![CDATA[Greggit – Google but it's only the Reddit results]]></title><description><![CDATA[
<p>Article URL: <a href="https://greggit.com">https://greggit.com</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=46070901">https://news.ycombinator.com/item?id=46070901</a></p>
<p>Points: 2</p>
<p># Comments: 6</p>
]]></description><pubDate>Thu, 27 Nov 2025 16:41:21 +0000</pubDate><link>https://greggit.com</link><dc:creator>goncharom</dc:creator><comments>https://news.ycombinator.com/item?id=46070901</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46070901</guid></item><item><title><![CDATA[New comment by goncharom in "Show HN: Turn any webpage into structured data via LLM codegen"]]></title><description><![CDATA[
<p>The regeneration loop was probably the most interesting part to work on: you need very strict constraints on what “good” content looks like and what the specific issue is when codegen fails. I found Pydantic annotations to be specifically useful for this.</p>
]]></description><pubDate>Thu, 13 Nov 2025 16:10:11 +0000</pubDate><link>https://news.ycombinator.com/item?id=45916621</link><dc:creator>goncharom</dc:creator><comments>https://news.ycombinator.com/item?id=45916621</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45916621</guid></item><item><title><![CDATA[Show HN: Turn any webpage into structured data via LLM codegen]]></title><description><![CDATA[
<p>Hey HN! I wanted to share a little library I've been working on to leverage AI to get structured data from arbitrary pages. Instead of sending the page's HTML to an LLM, Hikugen asks it to generate python code to fetch the data and enforces the generated data conforms to a Pydantic schema defined by the user. I'm using this to power yomu (<a href="https://github.com/goncharom/yomu" rel="nofollow">https://github.com/goncharom/yomu</a>), a personal email newsletter built from arbitrary websites.<p>Hikugen's main features:<p>- Automatically generates, runs, regenerates and caches the LLM-generated extraction code.
- It uses sqlite to save the current working code for each page so it can be reused across executions.
- It uses OpenRouter (<a href="https://openrouter.ai/" rel="nofollow">https://openrouter.ai/</a>) to call the LLM.
- It can fetch the page automatically (it can even reuse Netscape-formatted cookies) but you can also just feed it the raw HTML and leverage the rest of its functionalities.<p>Here's a snippet of what it looks like:<p><pre><code>  from hikugen import HikuExtractor
  from pydantic import BaseModel
  from typing import List
  
  class Article(BaseModel):
      title: str
      author: str
      published_date: str
      content: str
  
  class ArticlePage(BaseModel):
      articles: List[Article]
  
  extractor = HikuExtractor(api_key="your-openrouter-api-key")
  
  result = extractor.extract(
      url="https://example.com/articles",
      schema=ArticlePage
  )
  
  for a in result.articles:
      print(a.title, a.author)
</code></pre>
Hikugen is intentionally minimal: it doesn't attempt website navigation, login flows, headless browsers, or large-scale crawling. Just "given this HTML, extract this structured data".<p>A good chunk of this was built with Claude Code (shoutout to Harper’s blog: <a href="https://harper.blog/2025/02/16/my-llm-codegen-workflow-atm/" rel="nofollow">https://harper.blog/2025/02/16/my-llm-codegen-workflow-atm/</a>.<p>Would love feedback or ideas—especially from others playing with codegen for scraping tasks.</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=45916119">https://news.ycombinator.com/item?id=45916119</a></p>
<p>Points: 2</p>
<p># Comments: 2</p>
]]></description><pubDate>Thu, 13 Nov 2025 15:35:51 +0000</pubDate><link>https://github.com/goncharom/hikugen</link><dc:creator>goncharom</dc:creator><comments>https://news.ycombinator.com/item?id=45916119</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45916119</guid></item><item><title><![CDATA[New comment by goncharom in "Ask HN: What Are You Working On? (Nov 2025)"]]></title><description><![CDATA[
<p>I've been working web scraping using LLMs, I just shared one of the libraries I created to get structured data from arbitrary pages: <a href="https://news.ycombinator.com/item?id=45870231">https://news.ycombinator.com/item?id=45870231</a><p>Instead of sending the page's HTML to an LLM, Hikugen asks it to generate python code to fetch the data and enforces the generated data conforms to a Pydantic schema defined by the user. I'm using this to power yomu (<a href="https://github.com/goncharom/yomu" rel="nofollow">https://github.com/goncharom/yomu</a>), a personal email newsletter built from arbitrary websites.</p>
]]></description><pubDate>Mon, 10 Nov 2025 00:46:21 +0000</pubDate><link>https://news.ycombinator.com/item?id=45870880</link><dc:creator>goncharom</dc:creator><comments>https://news.ycombinator.com/item?id=45870880</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45870880</guid></item><item><title><![CDATA[Show HN: Hikugen – minimalistic LLM-generated web scrapers for structured data]]></title><description><![CDATA[
<p>Hey HN! I wanted to share a little library I've been working on to leverage AI to get structured data from arbitrary pages. Instead of sending the page's HTML to an LLM, Hikugen asks it to generate python code to fetch the data and enforces the generated data conforms to a Pydantic schema defined by the user. I'm using this to power yomu (<a href="https://github.com/goncharom/yomu" rel="nofollow">https://github.com/goncharom/yomu</a>), a personal email newsletter built from arbitrary websites.<p>Hikugen's main features:<p>- Automatically generates, runs, regenerates and caches the LLM-generated extraction code.<p>- It uses sqlite to save the current working code for each page so it can be reused across executions.<p>- It uses OpenRouter (<a href="https://openrouter.ai/" rel="nofollow">https://openrouter.ai/</a>) to call the LLM.<p>- It can fetch the page automatically (it can even reuse Netscape-formatted cookies) but you can also just feed it the raw HTML and leverage the rest of its functionalities.<p>Here's a snippet of what it looks like:<p><pre><code>  from hikugen import HikuExtractor
  from pydantic import BaseModel
  from typing import List
  
  class Article(BaseModel):
      title: str
      author: str
      published_date: str
      content: str
  
  class ArticlePage(BaseModel):
      articles: List[Article]
  
  extractor = HikuExtractor(api_key="your-openrouter-api-key")
  
  result = extractor.extract(
      url="https://example.com/articles",
      schema=ArticlePage
  )
  
  for a in result.articles:
      print(a.title, a.author)
</code></pre>
Hikugen is intentionally minimal: it doesn't attempt website navigation, login flows, headless browsers, or large-scale crawling. Just "given this HTML, extract this structured data".<p>A good chunk of this was built with Claude Code (shoutout to Harper’s blog: <a href="https://harper.blog/2025/02/16/my-llm-codegen-workflow-atm/" rel="nofollow">https://harper.blog/2025/02/16/my-llm-codegen-workflow-atm/</a>.<p>Would love feedback or ideas—especially from others playing with codegen for scraping tasks.</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=45870231">https://news.ycombinator.com/item?id=45870231</a></p>
<p>Points: 2</p>
<p># Comments: 0</p>
]]></description><pubDate>Sun, 09 Nov 2025 23:15:09 +0000</pubDate><link>https://github.com/goncharom/hikugen</link><dc:creator>goncharom</dc:creator><comments>https://news.ycombinator.com/item?id=45870231</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45870231</guid></item></channel></rss>