<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: ctoth</title><link>https://news.ycombinator.com/user?id=ctoth</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Tue, 14 Apr 2026 11:21:35 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=ctoth" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by ctoth in "AI Will Be Met with Violence, and Nothing Good Will Come of It"]]></title><description><![CDATA[
<p>Brought to you from the same guy who said in April 2025: "Google Is Winning on Every AI Front" <a href="https://www.thealgorithmicbridge.com/p/google-is-winning-on-every-ai-front" rel="nofollow">https://www.thealgorithmicbridge.com/p/google-is-winning-on-...</a></p>
]]></description><pubDate>Sun, 12 Apr 2026 18:17:06 +0000</pubDate><link>https://news.ycombinator.com/item?id=47742653</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47742653</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47742653</guid></item><item><title><![CDATA[New comment by ctoth in "Small models also found the vulnerabilities that Mythos found"]]></title><description><![CDATA[
<p>> It's weird that Aisle wrote this.<p>No, writing an advertisement is not weird. What's weird is that it's top of HN. Or really, no, this isn't weird either if you think about it -- people lookin for a gotcha "Oh see, that new model really isn't that good/it's surely hitting a wall/plateau any day now" upvoted it.</p>
]]></description><pubDate>Sat, 11 Apr 2026 17:39:44 +0000</pubDate><link>https://news.ycombinator.com/item?id=47732447</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47732447</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47732447</guid></item><item><title><![CDATA[New comment by ctoth in "Small models also found the vulnerabilities that Mythos found"]]></title><description><![CDATA[
<p>> They recovered much of the same analysis<p>Really?<p>> We isolated the vulnerable vc_rpc_gss_validate function, provided architectural context (that it handles network-parsed RPC credentials, that oa_length comes from the packet), and asked eight models to assess it for security vulnerabilities.<p>No.</p>
]]></description><pubDate>Sat, 11 Apr 2026 17:37:21 +0000</pubDate><link>https://news.ycombinator.com/item?id=47732427</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47732427</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47732427</guid></item><item><title><![CDATA[New comment by ctoth in "Molotov cocktail is hurled at home of Sam Altman"]]></title><description><![CDATA[
<p>Hi! I love to cook! I also use AI to brainstorm recipes sometimes! Wanna try asking Claude, ChatGPT, Gemini, or even Grok what temperature chicken needs to be cooked to? I just asked Claude: 165°F (74°C) internal temperature.<p>Where does this come from?</p>
]]></description><pubDate>Fri, 10 Apr 2026 19:34:20 +0000</pubDate><link>https://news.ycombinator.com/item?id=47722643</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47722643</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47722643</guid></item><item><title><![CDATA[New comment by ctoth in "Research-Driven Agents: When an agent reads before it codes"]]></title><description><![CDATA[
<p>I'm actually sort of working on this! <a href="https://github.com/ctoth/propstore" rel="nofollow">https://github.com/ctoth/propstore</a> -- it's like Cyc, but there is no one answer. Plus knowledge bases are literally git repos that you can fork/merge. Research-papers-plugin is the frontend, we extract the knowledge, then we need somewhere to put it :)</p>
]]></description><pubDate>Thu, 09 Apr 2026 20:22:35 +0000</pubDate><link>https://news.ycombinator.com/item?id=47709338</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47709338</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47709338</guid></item><item><title><![CDATA[New comment by ctoth in "Research-Driven Agents: When an agent reads before it codes"]]></title><description><![CDATA[
<p>For each paper, have your agent extract a three sentence description, create a description.md, then concat those with the paper names into an INDEX.md which it should consult to find appropriate papers. Also: have your agent tag papers, then autogenerate your tagged collection on the filesystem. Then you get nice things like <a href="https://github.com/ctoth/Qlatt/tree/master/papers/tagged" rel="nofollow">https://github.com/ctoth/Qlatt/tree/master/papers/tagged</a><p>Then something in your {CLAUDE,AGENTS}.md that says: when working on something with relevant context supplied by papers, read the papers before doing the work. You can find all papers plus their descriptions in ./papers/INDEX.md and papers by tag in ./papers/tagged</p>
]]></description><pubDate>Thu, 09 Apr 2026 20:17:56 +0000</pubDate><link>https://news.ycombinator.com/item?id=47709255</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47709255</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47709255</guid></item><item><title><![CDATA[New comment by ctoth in "Research-Driven Agents: What Happens When Your Agent Reads Before It Codes"]]></title><description><![CDATA[
<p>Check out the Researcher and Process Leads skill in ctoth/research-papers-plugin. I have basically completely automated the literature review.</p>
]]></description><pubDate>Thu, 09 Apr 2026 19:56:09 +0000</pubDate><link>https://news.ycombinator.com/item?id=47708921</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47708921</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47708921</guid></item><item><title><![CDATA[New comment by ctoth in "Research-Driven Agents: When an agent reads before it codes"]]></title><description><![CDATA[
<p>I used <a href="https://github.com/ctoth/research-papers-plugin" rel="nofollow">https://github.com/ctoth/research-papers-plugin</a> to produce the annotations. The thing that's really cool is how they surface the cross-links in the collection, for instance look at <a href="https://github.com/ctoth/Qlatt/blob/master/papers/Fant_1988_LFFrequencyDomainInterpretation/notes.md" rel="nofollow">https://github.com/ctoth/Qlatt/blob/master/papers/Fant_1988_...</a><p>Claude is much faster and better at reading papers than Codex (some of this is nested skill dispatch) but they both work quite incredibly for this. Compile your set of papers, queue it up and hit /ingest-collection and go sleep, and come back to a remarkable knowledge base :)</p>
]]></description><pubDate>Thu, 09 Apr 2026 19:53:28 +0000</pubDate><link>https://news.ycombinator.com/item?id=47708889</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47708889</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47708889</guid></item><item><title><![CDATA[New comment by ctoth in "Research-Driven Agents: When an agent reads before it codes"]]></title><description><![CDATA[
<p>I've been working on ctoth/research-papers-plugin, the pipeline to actually get LLMs to extract the notes. I really like your insight re RST over Markdown! It sounds like we're working on similar stuff and I'll absolutely reach out :)</p>
]]></description><pubDate>Thu, 09 Apr 2026 19:43:40 +0000</pubDate><link>https://news.ycombinator.com/item?id=47708759</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47708759</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47708759</guid></item><item><title><![CDATA[New comment by ctoth in "Research-Driven Agents: When an agent reads before it codes"]]></title><description><![CDATA[
<p>I've been very interested in this recently. I'm pretty sure that <i>every</i> project should have a ./papers directory of annotated papers in it like I do in Qlatt[0].<p>Literally every project. If it's something that's been done a million times then that means it has good literature on it? If not, then even more important to find related stuff! And not just crunchy CS stuff like databases or compilers or whatever. Are you creating a UI? There's probably been great UI research you can base off of! Will this game loop be fun in the game you're building? There's probably been research about it!<p>[0]: <a href="https://github.com/ctoth/Qlatt/blob/master/papers/" rel="nofollow">https://github.com/ctoth/Qlatt/blob/master/papers/</a></p>
]]></description><pubDate>Thu, 09 Apr 2026 19:41:41 +0000</pubDate><link>https://news.ycombinator.com/item?id=47708734</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47708734</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47708734</guid></item><item><title><![CDATA[The 200k Ghost: Instruction Degradation in Long-Context LLM Sessions]]></title><description><![CDATA[
<p>Article URL: <a href="https://github.com/WaspBeeNSOSWE/the-200k-ghost">https://github.com/WaspBeeNSOSWE/the-200k-ghost</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47706689">https://news.ycombinator.com/item?id=47706689</a></p>
<p>Points: 3</p>
<p># Comments: 1</p>
]]></description><pubDate>Thu, 09 Apr 2026 17:35:20 +0000</pubDate><link>https://github.com/WaspBeeNSOSWE/the-200k-ghost</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47706689</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47706689</guid></item><item><title><![CDATA[New comment by ctoth in "System Card: Claude Mythos Preview [pdf]"]]></title><description><![CDATA[
<p>My favorite example of this from last night:<p>Me: Let's figure out how to clone our company Wordpress theme in Hugo. Here're some tools you can use, here's a way to compare screenshots, iterate until 0% difference.<p>Codex: Okay Boss! I did the thing! I couldn't get the CSS to match so I just took PNGs of the original site and put them in place! Matches 100%!</p>
]]></description><pubDate>Tue, 07 Apr 2026 19:35:51 +0000</pubDate><link>https://news.ycombinator.com/item?id=47680290</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47680290</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47680290</guid></item><item><title><![CDATA[New comment by ctoth in "Issue: Claude Code is unusable for complex engineering tasks with Feb updates"]]></title><description><![CDATA[
<p>Fair point on tone.  It's a bit of a bind isn't it? When you come with a well-researched issue as OP did, you get this bland corporate nonsense "don't believe your lyin' eyes, we didn't change anything major, you can fix it in settings."<p>How should you actually communicate in such a way that you are actually heard when this is the default wall you hit?<p>The author is in this thread saying every suggested setting is already maxed. The response is "try these settings." What's the productive version of pointing out that the answer doesn't address the evidence? Genuine question. I linked my repo because it's the most concrete example I have.</p>
]]></description><pubDate>Mon, 06 Apr 2026 18:32:32 +0000</pubDate><link>https://news.ycombinator.com/item?id=47664933</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47664933</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47664933</guid></item><item><title><![CDATA[New comment by ctoth in "Issue: Claude Code is unusable for complex engineering tasks with Feb updates"]]></title><description><![CDATA[
<p>[flagged]</p>
]]></description><pubDate>Mon, 06 Apr 2026 18:20:32 +0000</pubDate><link>https://news.ycombinator.com/item?id=47664793</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47664793</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47664793</guid></item><item><title><![CDATA[New comment by ctoth in "Embarrassingly simple self-distillation improves code generation"]]></title><description><![CDATA[
<p>> Also, we have full control over their operation and we can trace every signal. Are you surprised we understand them better than brains?<p>Very, monsieur Laplace.</p>
]]></description><pubDate>Sat, 04 Apr 2026 18:55:33 +0000</pubDate><link>https://news.ycombinator.com/item?id=47642077</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47642077</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47642077</guid></item><item><title><![CDATA[New comment by ctoth in "OpenClaw privilege escalation vulnerability"]]></title><description><![CDATA[
<p>Those numbers aren't in the CVE. You introduced them, attributed them to a source that doesn't contain them, and now you're disclaiming them. Where did they come from, and what was the goal of sharing them?</p>
]]></description><pubDate>Sat, 04 Apr 2026 18:36:43 +0000</pubDate><link>https://news.ycombinator.com/item?id=47641928</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47641928</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47641928</guid></item><item><title><![CDATA[New comment by ctoth in "Show HN: Castra – Strip orchestration rights from your LLMs"]]></title><description><![CDATA[
<p>I've thought a lot about this problem. Is a state machine what you want? Or is it actually a Behavior Tree which can construct itself on-the-fly?</p>
]]></description><pubDate>Wed, 01 Apr 2026 15:45:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=47602435</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47602435</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47602435</guid></item><item><title><![CDATA[New comment by ctoth in "ChatGPT won't let you type until Cloudflare reads your React state"]]></title><description><![CDATA[
<p>The ZK part isn't the problem. The "attestation of recent humanity" part is. Who attests? What happens when someone can't get attested?<p>You've been to the doctor recently, right? Given them your SSN? Every identity system ever built was going to be scoped || voluntary. None of them stayed that way.<p>Once you have the identity mechanism, "Oh it's zero knowledge! So let's use it for your age! Have you ever been convicted?" which leads to "mandated by employers" which leads to...<p>We've seen this goddamn movie before. Let's just skip it  this time? Please?</p>
]]></description><pubDate>Sun, 29 Mar 2026 22:28:28 +0000</pubDate><link>https://news.ycombinator.com/item?id=47568083</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47568083</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47568083</guid></item><item><title><![CDATA[New comment by ctoth in "ChatGPT won't let you type until Cloudflare reads your React state"]]></title><description><![CDATA[
<p>No, we really don't. We don't need worldcoin, we don't need papers, please. We just don't.<p>"Prove your humanity/age/other properties" with this mechanism quickly goes places you do not want it to go.</p>
]]></description><pubDate>Sun, 29 Mar 2026 22:17:09 +0000</pubDate><link>https://news.ycombinator.com/item?id=47567983</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47567983</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47567983</guid></item><item><title><![CDATA[New comment by ctoth in "The bot situation on the internet is worse than you could imagine"]]></title><description><![CDATA[
<p>If you don't mind me asking, what sort of data are you licensing? I noticed that you explicitly don't mention it.</p>
]]></description><pubDate>Sun, 29 Mar 2026 20:45:08 +0000</pubDate><link>https://news.ycombinator.com/item?id=47567119</link><dc:creator>ctoth</dc:creator><comments>https://news.ycombinator.com/item?id=47567119</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47567119</guid></item></channel></rss>