<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: nickpsecurity</title><link>https://news.ycombinator.com/user?id=nickpsecurity</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Wed, 18 Nov 2026 04:34:55 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=nickpsecurity" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by nickpsecurity in "Show HN: A terminal-native idle factory game, written in Rust with ratatui"]]></title><description><![CDATA[
<p>If it runs fast and low resource, it could be a nice testbed for agent architectures. I'm thinking old-school planners, genetic programming, ML techniques, etc.<p>Especially integrating planners or logic-based approaches with ML models. The old architectures for StarCraft were great at building and micro-managing units but terrible at assessing or adapting to enemy strategy. Hybrid architectures can address that.<p>Games like yours, esp different styles or requirements, can be used to prove how well the techniques generalize.</p>
]]></description><pubDate>Sat, 14 Nov 2026 18:44:29 +0000</pubDate><link>https://news.ycombinator.com/item?id=50499613</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=50499613</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50499613</guid></item><item><title><![CDATA[New comment by nickpsecurity in "Proving and improving the Platonic Representation Hypothesis with JEPAs"]]></title><description><![CDATA[
<p>Math is a social process that involves mental concepts, symbols for them, and discussion about them. It also needs peer review. If you see math in a paper, that's only one part of the process.<p>Further, in machine learning, we see lots of papers with math that doesn't lead to anything useful. Most papers claiming advances over prior architectures don't pan out in later testing. So, there will be skepticism when people make performance claims like the authors'. The best response is just independent testing.</p>
]]></description><pubDate>Mon, 02 Nov 2026 00:35:43 +0000</pubDate><link>https://news.ycombinator.com/item?id=50331544</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=50331544</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50331544</guid></item><item><title><![CDATA[New comment by nickpsecurity in "Proving and improving the Platonic Representation Hypothesis with JEPAs"]]></title><description><![CDATA[
<p>The math can be misinterpreted, misapplied, or not fit reality. That's why high-assurance security mandated a combination of things:<p>1. English specification of the system and its goals<p>2. Formal, mathematical specification of the system and its goals<p>3. Mathematical proof the system met its goals<p>4. Code implementing that system with a proof, often informal, that the code embodied the informal and formal design.<p>5. Tests of all functionality to prove the design and code worked.<p>6. Independent analysis of the above to find flaws in them.<p>7. If the implementation has concurrency or parallelism, more errors can occur due to that. There were techniques to mitigate it.<p>Each step found problems in systems made by professionals. Mathematical work and coding both have a long history of errors in their on fields. Combining them increases the likelihood of error.</p>
]]></description><pubDate>Sun, 01 Nov 2026 21:23:24 +0000</pubDate><link>https://news.ycombinator.com/item?id=50330044</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=50330044</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50330044</guid></item><item><title><![CDATA[New comment by nickpsecurity in "A Manchurian Candidate Hidden in Qwen's Weights"]]></title><description><![CDATA[
<p>Author thought nobody was playing chess with a beach-side view?<p>But God gave us beautiful oceans and intellectually-stimulating games. I'm sure some chess players aren't wasting the opportunity to enjoy both. Now, they're at risk of being taken out!</p>
]]></description><pubDate>Wed, 28 Oct 2026 18:07:01 +0000</pubDate><link>https://news.ycombinator.com/item?id=50278239</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=50278239</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50278239</guid></item><item><title><![CDATA[New comment by nickpsecurity in "Securing our frontier model training environment"]]></title><description><![CDATA[
<p>You raise fair concerns. So, we need to get data to systems they're reading from. But, they can't connect back over those lines.<p>The safest way to do that is a high-assurance, data diode (one-way link). One builds hardware that, physically and by software, only allows information to flow one day. Many are available commercially.<p>Sometimes you need proof of acknowledgement of each packet but nothing else. Naval Research Laboratory had The [Network] Pump that allowed only Ack's back and with algorithm to throw off covert, timing channels.<p>Next, one might want to let people send them requests over an API which they can respond to in limited ways. One can use a high-assurance guard with secure, hardware/software architecture. It uses simplified I/O with a message format that's easy to parse securely.<p>At this point, I'm also thinking it might be better to use the closed-source, proprietary solutions. The AI's are rapidly scanning the open-source code for vulnerabilities which I'm sure they're not always sharing. They'd be more likely to have attacks on it.<p>So, don't tell the untrusted model how the guard works, use private models + traditional verification to secure its code, and include both HIDS and NIDS monitoring. Also, make the systems as predictable and boring as possible to make anomalies easier to spot.</p>
]]></description><pubDate>Sat, 24 Oct 2026 03:09:26 +0000</pubDate><link>https://news.ycombinator.com/item?id=50218047</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=50218047</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50218047</guid></item><item><title><![CDATA[New comment by nickpsecurity in "Securing our frontier model training environment"]]></title><description><![CDATA[
<p>Even if it was, covert, channel analysis from the 1980's and 1990's would've taught it to do or prevent that. It was required in all TCSEC-certified systems high, security classes. Many papers were published describing those analyses, including cache channel, which are in their pretraining data.<p>Most of IT security ignored that work and never used the methods. That's why their syatems can't defend against them. There are hardware and software groups that built countermeasures. They're not mainstream. I don't know if the hardware needed is even commercially available (eg partitioning caches, async & randomized CPU's).</p>
]]></description><pubDate>Sat, 24 Oct 2026 03:00:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=50218011</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=50218011</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50218011</guid></item><item><title><![CDATA[New comment by nickpsecurity in "The RAM shortage is bringing back DDR4"]]></title><description><![CDATA[
<p>My idea was taking 20+ year old designs that are patent-free, implementing them on any modern nodes that work (esp MOSIS-supported), their performance per watt will probably be better, and just use those. Some of the old designs, like NonStop architecture or MPP clusters, still have amazing benefits.</p>
]]></description><pubDate>Sun, 11 Oct 2026 14:38:07 +0000</pubDate><link>https://news.ycombinator.com/item?id=50043778</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=50043778</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50043778</guid></item><item><title><![CDATA[New comment by nickpsecurity in "Unikernels were hard. key word: were"]]></title><description><![CDATA[
<p>The two choices offered are a false delimma. The MILS kernels that came before seL4, like INTEGRITY RTOS and LynxSecure, had middleware to make deplpyments easier. The open-source solutions mixed standalone apps with user-mode VM's for Linux compatibility. OKL4 let one use drivers across them.<p>Looking at what's out there, I think GenodeOS should be mentioned for application-specific systems. They can even build some of their stuff on seL4 if you want.<p><a href="https://genode.org/" rel="nofollow">https://genode.org/</a><p>I'll also add there's been unikernel projects that aren't in exotic languages. Even C++ (IncludeOS?). There's probably one in Rust by now. If not, it could probably build on Redox's components.</p>
]]></description><pubDate>Sun, 11 Oct 2026 03:12:25 +0000</pubDate><link>https://news.ycombinator.com/item?id=50039492</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=50039492</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50039492</guid></item><item><title><![CDATA[New comment by nickpsecurity in "Ask HN: What do you run on a $5 VPS that's worth keeping online 24/7?"]]></title><description><![CDATA[
<p>I actually considered doing that with my droplet. Do you just use a web browser on your phone or specific apps?</p>
]]></description><pubDate>Sat, 10 Oct 2026 13:22:58 +0000</pubDate><link>https://news.ycombinator.com/item?id=50032705</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=50032705</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50032705</guid></item><item><title><![CDATA[New comment by nickpsecurity in "Ask HN: What do you run on a $5 VPS that's worth keeping online 24/7?"]]></title><description><![CDATA[
<p>Mine are on DigitalOcean. Someone else said VPSDime.com.</p>
]]></description><pubDate>Sat, 10 Oct 2026 13:22:00 +0000</pubDate><link>https://news.ycombinator.com/item?id=50032694</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=50032694</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50032694</guid></item><item><title><![CDATA[New comment by nickpsecurity in "Ask HN: What do you run on a $5 VPS that's worth keeping online 24/7?"]]></title><description><![CDATA[
<p>All my domains run on different, $5 droplets. Even if money is tight, I keep two up that lead to saved souls and transformed lives:<p><a href="https://www.gethisword.com" rel="nofollow">https://www.gethisword.com</a><p><a href="https://www.givethemlife.com" rel="nofollow">https://www.givethemlife.com</a><p>I'm also using BunnyCDN right now.</p>
]]></description><pubDate>Fri, 09 Oct 2026 14:51:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=50021344</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=50021344</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50021344</guid></item><item><title><![CDATA[New comment by nickpsecurity in "US Suspends Visa Program for Tech Firms Including Microsoft"]]></title><description><![CDATA[
<p>They were also often doing basic stuff, like database programming. Even in AI, most of what they produce didn't have marketable value. So, I don't buy that they couldn't find <i>or train</i> anyone here to do the job.</p>
]]></description><pubDate>Thu, 08 Oct 2026 16:04:53 +0000</pubDate><link>https://news.ycombinator.com/item?id=50007535</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=50007535</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50007535</guid></item><item><title><![CDATA[New comment by nickpsecurity in "A browser-native classic Visual Basic VB6 IDE"]]></title><description><![CDATA[
<p>I made a utility, only partly done, to convert Avalonia files to HTML. It was intended to make cross-platform, Python/Tkinkter GUI's with HTML/Flask being a pit stop. Also, run them on QPython on Android. I got too busy to improve it past a few, form controls.<p>So, I think there's still potential to throw together an interface even if it can't do all of it. We do need a placeholder box that can be labeled for the controls we add manually but use its sizing info. Also, a converter that autoscales them in HTML or Qt to maintain the ratios that were in the VB6 app. For the placeholder, I was temporarily using that one control that's like a giant frame or box meant to contain others.<p>One reason I abandoned the idea is the AI's are so good now that we can probably just generate better GUI's from scratch or a basic, VB6 sketch. I couldn't justify the work until I try an AI with such projects. To avoid hallucinations or garbagr designs, might make sense to do the original program but AI-generate the transpiler.</p>
]]></description><pubDate>Tue, 06 Oct 2026 03:54:15 +0000</pubDate><link>https://news.ycombinator.com/item?id=49973957</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=49973957</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49973957</guid></item><item><title><![CDATA[New comment by nickpsecurity in "Beam: Reflection's 501B open-weight model"]]></title><description><![CDATA[
<p>Sometimes we just enjiy having a conversation with people. It's how we got many answers before Google exists. I believe such choices have many, positive effects on people that society is losing.</p>
]]></description><pubDate>Tue, 06 Oct 2026 03:45:57 +0000</pubDate><link>https://news.ycombinator.com/item?id=49973914</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=49973914</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49973914</guid></item><item><title><![CDATA[New comment by nickpsecurity in "Run Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/s"]]></title><description><![CDATA[
<p>That's cheap, too. Which hosting service are you using?</p>
]]></description><pubDate>Sun, 04 Oct 2026 20:18:17 +0000</pubDate><link>https://news.ycombinator.com/item?id=49957405</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=49957405</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49957405</guid></item><item><title><![CDATA[New comment by nickpsecurity in "GLM-5.3 and the spread of advanced cyber capabilities"]]></title><description><![CDATA[
<p>I seriously thought that's what they were going to say. It seemed like it was setting up the reader to think the agent was about to use the exploit to escape the sandbox. Then, it was contained by Anthropic's security practices. But, if random people run things things, they couldn't be contained! They'll be a hoard of agents attacking the whole Internet!<p>Then, it was just that it made an exploit, and works really well, and people might use it instead of their products. Tragic for their investors I guess...</p>
]]></description><pubDate>Tue, 29 Sep 2026 22:41:21 +0000</pubDate><link>https://news.ycombinator.com/item?id=49901806</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=49901806</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49901806</guid></item><item><title><![CDATA[New comment by nickpsecurity in "Trusting-Trust Attack against an Entire Linux Distribution"]]></title><description><![CDATA[
<p>It's very exciting to see the work on combining AI models with static analysis, test generation, formal proof, and refactoring. All of these suggests we might see high assurance (EAL6+) developed rapidly in the future. At least for combinations of well-understood concepts.</p>
]]></description><pubDate>Tue, 08 Sep 2026 00:08:10 +0000</pubDate><link>https://news.ycombinator.com/item?id=49604326</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=49604326</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49604326</guid></item><item><title><![CDATA[New comment by nickpsecurity in "Trusting-Trust Attack against an Entire Linux Distribution"]]></title><description><![CDATA[
<p>It goes back to Paul Karger's MULTICS Security Evaluation where he invented and described the attack. Thompson learned it from him. Karger invented a lot of attacks and security techniques a decade or more ahead of the hacking community.<p><a href="https://gwern.net/doc/cs/security/2002-karger.pdf" rel="nofollow">https://gwern.net/doc/cs/security/2002-karger.pdf</a></p>
]]></description><pubDate>Mon, 07 Sep 2026 21:40:56 +0000</pubDate><link>https://news.ycombinator.com/item?id=49603254</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=49603254</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49603254</guid></item><item><title><![CDATA[New comment by nickpsecurity in "Trusting-Trust Attack against an Entire Linux Distribution"]]></title><description><![CDATA[
<p>The solution to this in the Orange Book (TCSEC) days in the 1980's-1990's was a system fully traceable from requirements to code, proven to embed a security policy, and analyzable and buildable from source locally by the customer using existing, trusted tools. Eventually, people added hashes for the code and data.<p>So, your program that combines source files or checks dependencies would be fully specified in its success and failure states. Only combinations of functions leading to a provably-secure state are even allowed. If you can't do that, the feature is too complex to allow. Human pentesters review it from design to algorithms to building it to spot ways attacks might happen.<p>That's what it takes to build software that <i>usually</i> resists subversion. Most software isn't built that way. It can't be because the priorities of developers and customers work against it. So, we'll continue to see clever attacks that exploit systems not designed to high security standards.<p>For this topic, I recommend David A. Wheeler's page on Software, Configuration Management Security because it covers many issues with it in mostly-centralized systems.</p>
]]></description><pubDate>Mon, 07 Sep 2026 21:39:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=49603236</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=49603236</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49603236</guid></item><item><title><![CDATA[New comment by nickpsecurity in "GPT-6 Astra"]]></title><description><![CDATA[
<p>Yeah, GPT4 was one-shotting utilities that GPT3 Davinci couldn't. So, I'd have my limited tokens on GPT4 crank out the initial program before iterating with my abundant, GPT3 tokens.</p>
]]></description><pubDate>Fri, 04 Sep 2026 02:15:05 +0000</pubDate><link>https://news.ycombinator.com/item?id=49559706</link><dc:creator>nickpsecurity</dc:creator><comments>https://news.ycombinator.com/item?id=49559706</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49559706</guid></item></channel></rss>