<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: h45x1</title><link>https://news.ycombinator.com/user?id=h45x1</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Thu, 19 Nov 2026 05:22:28 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=h45x1" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by h45x1 in "[dead]"]]></title><description><![CDATA[
<p>I have been working on this one for a few months already, so the timing is only incidental with the recent protests. Mostly these are my thoughts on how to write simulation code in a more declarative style as opposed to a more imperative style that I have been doing for most projects so far. Recently I read some blogs about Haskell, the language was on my mind, and I thought that a declarative style and Haskell likely go well together. That resulted in a quick draft during my morning commute and, much later, in this write up.</p>
]]></description><pubDate>Thu, 05 Nov 2026 20:01:15 +0000</pubDate><link>https://news.ycombinator.com/item?id=50383713</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=50383713</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50383713</guid></item><item><title><![CDATA[New comment by h45x1 in "Ask HN: What's your current AI agentic setup?"]]></title><description><![CDATA[
<p>I wrote my own bash wrapper around bubblewrap and pasta:
<a href="https://github.com/andrei-dubovik/jail" rel="nofollow">https://github.com/andrei-dubovik/jail</a><p>The wrapper also keeps history, memory, etc. split by project folder, so agents working in different projects don't know each others' history. If anything, `--resume` is a split second easier.</p>
]]></description><pubDate>Thu, 05 Nov 2026 06:59:59 +0000</pubDate><link>https://news.ycombinator.com/item?id=50376030</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=50376030</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50376030</guid></item><item><title><![CDATA[New comment by h45x1 in "Should LLMs write in mainstream languages?"]]></title><description><![CDATA[
<p>I guess I'm rather late for a discussion here, came across by accident while searching for an older post I remembered. Still, this is a question I've been pondering about lately, so let me post a few remarks.<p>Modern harnesses are one size fits all. I think I'm not alone in thinking that were I to desing a harness to fit my work habbits, it'll look very different from the mainstream ones. One could also argue, if you design a harness for an LLM, it's best to bootstrap early in the process, i.e. start using your own harness for designing itself, for then your work stays more grounded. So, you have a system that rewrites itself. Intuitively, Lisp and Erlang are very good at changing live systems. So, would you not want to do it in those languages? Nice to see others touching on this topic.<p>For me the counterargument is that Lisp and Erlang are untyped. If I were to work on my own harness, what I would like to try is more type-oriented development, where I sketch the types and maybe some function signatures and let an agent fill the rest.</p>
]]></description><pubDate>Fri, 30 Oct 2026 11:46:06 +0000</pubDate><link>https://news.ycombinator.com/item?id=50302652</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=50302652</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50302652</guid></item><item><title><![CDATA[New comment by h45x1 in "[dead]"]]></title><description><![CDATA[
<p>I might be misjudging the security risks, but I can't see myself running Claude Code with full access to my work machine, simply relying on its built-in sandboxes and permission systems. For a long time I simply ran it under QEMU, but that proved less convenient than desired. So, recently I put together a Bash wrapper around bwrap and pasta for sandboxing Claude Code. And while I was at it, I also organized a per-project split for Claude config files, so that sessions, say, and certain other files are always project-specific and there is no possible cross-contamination between projects. Sharing, for it proved non-trivial to run bwrap and pasta together: bwrap creates nested namespaces and so one needs to switch namespaces as well before pasta can be attached to bwrap. Now I can just do `jail project-folder` and then `claude`. The wrapper is for Arch, other distros might need some path adjustments.</p>
]]></description><pubDate>Fri, 05 Jun 2026 07:14:19 +0000</pubDate><link>https://news.ycombinator.com/item?id=48409085</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=48409085</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48409085</guid></item><item><title><![CDATA[New comment by h45x1 in "Ask HN: What happens when humans become as dumb as AI?"]]></title><description><![CDATA[
<p>Less knowledgeable overall, I would agree. LLMs are, if anything, the ultimate knowledge bases. But within our own areas of expertise, our worldview is typically more coherent and we have often learnt to reason optimally within that world view. To be fair, it is hard for me to generalize. So at least that's what I see in areas which I know. As pointed out by others, however, making discoveries that bridge disciplines is where LLMs might have an edge.</p>
]]></description><pubDate>Thu, 04 Jun 2026 14:16:42 +0000</pubDate><link>https://news.ycombinator.com/item?id=48399018</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=48399018</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48399018</guid></item><item><title><![CDATA[Ask HN: What happens when humans become as dumb as AI?]]></title><description><![CDATA[
<p>The existential risk that has received much attention is machines eventually becoming as smart as people, and then smarter still. What I see in the news, and, anecdotally, around me is rather the opposite. Thinking is hard. People, even those who went through rigorous university training to develop their critical thinking, are increasingly outsourcing thinking to machines. SOTA models don't need to get any better to catch up with us, they just need to wait. And maybe not even that long. I wonder what happens then?<p>As for my opinion, I am an economist by training. There are many people, many educational systems; the risks stemming from the mistakes people make are idiosyncratic, and they average out in aggregate (to an extent). On the other hand, there are just a few SOTA models. If all thinking is done by them, the aggregate risks will be much larger, I'd suppose.</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=48398449">https://news.ycombinator.com/item?id=48398449</a></p>
<p>Points: 6</p>
<p># Comments: 4</p>
]]></description><pubDate>Thu, 04 Jun 2026 13:34:57 +0000</pubDate><link>https://news.ycombinator.com/item?id=48398449</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=48398449</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48398449</guid></item><item><title><![CDATA[New comment by h45x1 in "I'm Tired of Talking to AI"]]></title><description><![CDATA[
<p>My pet theory comes to information density. LLM writing feels too shallow. If you read a text from a person who cares about her subject, there will be more useful detail and insight scatterd throughout. And the information within sentences will be more meaningfully connected. On one hand, LLM would know all that, but it does not seem to know how to distill that knowledge into a written text. I read what people write. If I know the text is from an LLM, I skim.</p>
]]></description><pubDate>Wed, 27 May 2026 17:58:42 +0000</pubDate><link>https://news.ycombinator.com/item?id=48297974</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=48297974</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48297974</guid></item><item><title><![CDATA[New comment by h45x1 in "Half of social-science studies fail replication test in years-long project"]]></title><description><![CDATA[
<p>I am surprised it's just half. The incentives are not there for replication or verification. You need to publish and it is easier to publish something new than to fight over existing results, especially if you're a young researcher.<p>I maintain something akin to google trends but for abstracts in economics papers. The word "novel" is growing rapidly, while "reproduce", "replicate", and "verify" have not changed much over the years:
<a href="https://dubovik.eu/blog/repec?t=replicate&t=reproduce&t=verify&t=novel" rel="nofollow">https://dubovik.eu/blog/repec?t=replicate&t=reproduce&t=veri...</a></p>
]]></description><pubDate>Wed, 01 Apr 2026 18:04:13 +0000</pubDate><link>https://news.ycombinator.com/item?id=47604330</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=47604330</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47604330</guid></item><item><title><![CDATA[New comment by h45x1 in "Take better notes, by hand"]]></title><description><![CDATA[
<p>Sure. Though to each his own, I'd imagine. Mine is quite basic.<p>- 4 pages at the back are reserved for index.<p>- Daily journal starts at the back.<p>- There is no obligation to have regular entries in the daily journal.<p>- ◦ denotes a past event; ◦ hh:mm denotes an upcoming or past event.<p>- → denotes a task.<p>- "circled" → denotes a completed task.<p>- strikeghrough denotes a cancelled or refiled task.<p>- ¿ optional task, not sure about something ?<p>- "-" is for all types of second-level bullets.<p>(As a side note, I mostly do task organization on the computer, but sometimes in a journal as well.)<p>- Topics start at the front.<p>- Topics are free-form.<p>- A new year starts a new journal. (I don't care for the new year resolutions though. At best, a list of side quests I'd like to do.)</p>
]]></description><pubDate>Mon, 30 Mar 2026 20:23:22 +0000</pubDate><link>https://news.ycombinator.com/item?id=47579273</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=47579273</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47579273</guid></item><item><title><![CDATA[New comment by h45x1 in "A Tinkerer's Introduction to Claude Code, by Claude Opus"]]></title><description><![CDATA[
<p>For reasons of just trying stuff out rather than for utility, I have a set of Python scripts to generate books from their titles alone using an LLM. There are hundreds of similarly titled projects on github. This weekend I generated a new book, "A Tinkerer's Introduction to Claude Code". I thought it would be fun to use a Claude model for that, so that's what I've done. (IHMO, Gemini is better at writing and cheaper for that purpose as well.) The book is dull, full disclaimer here, but it has a few bright moments. Not least,<p>"Set aside time—perhaps one small task per session—where you write code unassisted. Treat it like exercise."</p>
]]></description><pubDate>Mon, 30 Mar 2026 19:07:05 +0000</pubDate><link>https://news.ycombinator.com/item?id=47578406</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=47578406</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47578406</guid></item><item><title><![CDATA[A Tinkerer's Introduction to Claude Code, by Claude Opus]]></title><description><![CDATA[
<p>Article URL: <a href="https://github.com/andrei-dubovik/boekwriter/releases/tag/tinkerers-claude-code-202603">https://github.com/andrei-dubovik/boekwriter/releases/tag/tinkerers-claude-code-202603</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47578405">https://news.ycombinator.com/item?id=47578405</a></p>
<p>Points: 2</p>
<p># Comments: 1</p>
]]></description><pubDate>Mon, 30 Mar 2026 19:07:05 +0000</pubDate><link>https://github.com/andrei-dubovik/boekwriter/releases/tag/tinkerers-claude-code-202603</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=47578405</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47578405</guid></item><item><title><![CDATA[New comment by h45x1 in "Take better notes, by hand"]]></title><description><![CDATA[
<p>There is also the question of real estate. I can have several paper notes side by side (when taking notes on loose sheets) but with iPads or ReMarkables that'll be rather decadent.</p>
]]></description><pubDate>Mon, 30 Mar 2026 17:51:39 +0000</pubDate><link>https://news.ycombinator.com/item?id=47577489</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=47577489</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47577489</guid></item><item><title><![CDATA[New comment by h45x1 in "Take better notes, by hand"]]></title><description><![CDATA[
<p>I have a dedicated couple of pages in a notebook, where I write down the note-taking conventions I use. When transitioning to a new notebook, I would copy those pages, possibly making a few improvements based on past usage. A most unhurried release cycle, if I can say so myself.<p>Regarding the space management, there are many solutions straight out of the programming world, of course: utilize both sides of the notebook, reserve a minimum number of pages per topic, keep an index with free pages, etc. But there are some hardware ones as well, I'm trying Atoma notebooks (<a href="https://atoma.be" rel="nofollow">https://atoma.be</a>) these days.</p>
]]></description><pubDate>Mon, 30 Mar 2026 17:46:47 +0000</pubDate><link>https://news.ycombinator.com/item?id=47577423</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=47577423</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47577423</guid></item><item><title><![CDATA[New comment by h45x1 in "The web should remain anonymous by default"]]></title><description><![CDATA[
<p>In my limited understanding, with zero-knowledge proofs. That's also what the blog mentions. Say, a website issues a challenge, you sign the challendge with your private key supplied by either your government (or some other authority). The signature is verified against a big set of public keys using a ZKP without knowing which specific public key matches.<p>I imagine the problem is more economics and politics than technology. Say, personhood verification becomes available. Then it is also that much easier to mandate age verification, and that becomes a slippery slope.</p>
]]></description><pubDate>Thu, 12 Mar 2026 14:08:58 +0000</pubDate><link>https://news.ycombinator.com/item?id=47350737</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=47350737</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47350737</guid></item><item><title><![CDATA[New comment by h45x1 in "The web should remain anonymous by default"]]></title><description><![CDATA[
<p>"Careful design of the protocols maintains privacy by preventing sites from learning any additional information beyond personhood." Is personhood verification desirable? Say, HN introduces personhood verification and budgets the number of posts or comments per person. That would limit the amount of AI posts, which seems to be desirable judging by the recent discussion (<a href="https://news.ycombinator.com/item?id=47340079">https://news.ycombinator.com/item?id=47340079</a>). What are the drawbacks though?</p>
]]></description><pubDate>Thu, 12 Mar 2026 12:32:15 +0000</pubDate><link>https://news.ycombinator.com/item?id=47349689</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=47349689</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47349689</guid></item><item><title><![CDATA[New comment by h45x1 in "Storing Preexisting Passwords on a Yubikey"]]></title><description><![CDATA[
<p>Starting with the CTAP 2.1 standard, there is a largeBlob extension in the standard that allows for storage of a small amount of arbitrary user data. The current libfido2 library implementation uses a largeBlobKey provided by the authenticator to encrypt the data in userspace before storage. However, at least on a Yubikey, obtaining the largeBlobKey does not require User Presence, which IMHO lowers the essential security guarantees that a hardware authenticator normally provides. Recently got myself a set of Yubikeys and tinkered with them for a while. It seems it is also perfectly possible to use the hmac-secret provided by a Yubikey to encrypt one's data in user space and then store it in the largeBlob array on a Yubikey. Importantly, requesting the hmac-secret does require User Presence.<p>That combination is hardly documented anywhere. And so I somewhat skeptical about my unorthodox security solution. Sharing on the off-chance a security expert would comment.</p>
]]></description><pubDate>Thu, 12 Mar 2026 09:05:01 +0000</pubDate><link>https://news.ycombinator.com/item?id=47348176</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=47348176</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47348176</guid></item><item><title><![CDATA[Storing Preexisting Passwords on a Yubikey]]></title><description><![CDATA[
<p>Article URL: <a href="https://dubovik.eu/blog/yubikey">https://dubovik.eu/blog/yubikey</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47348175">https://news.ycombinator.com/item?id=47348175</a></p>
<p>Points: 1</p>
<p># Comments: 1</p>
]]></description><pubDate>Thu, 12 Mar 2026 09:05:01 +0000</pubDate><link>https://dubovik.eu/blog/yubikey</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=47348175</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47348175</guid></item><item><title><![CDATA[Deciphering the HMAC-secret and largeBlob CTAP2 Extensions]]></title><description><![CDATA[
<p>Article URL: <a href="https://dubovik.eu/blog/yubikey">https://dubovik.eu/blog/yubikey</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47298581">https://news.ycombinator.com/item?id=47298581</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Sun, 08 Mar 2026 16:25:33 +0000</pubDate><link>https://dubovik.eu/blog/yubikey</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=47298581</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47298581</guid></item><item><title><![CDATA[New comment by h45x1 in "Gemini and I Wrote a Book: Introduction to Computational Linguistics"]]></title><description><![CDATA[
<p>The prompts are slightly tailored towards a textbook but are otherwise generic. If you have a Gemini API key, you can specify any title and get your own book. It can take several hours for a larger book (can be done faster and slightly cheaper with async queries, but I thought that an unnecessary complication for a small pilot). It is easy to disagree on the resulting quality of the book; for me it feels more like an extended Google query than anything. But as an exercise where the size of your prompt stack is larger than 1, it was fun and, in parts, thought-provoking.<p>If I am to name just one prompting trick that I discovered to be useful, that'll be hierarchical budgeting. I have a total word budget for the whole book and I ask Gemini to allocate that budget for chapters, and so on recursively. Fun part, the sum of parts does not always equal the total, but the difference was never bigger than 5%. That's some stochastic summing happening.</p>
]]></description><pubDate>Sun, 19 Oct 2025 08:10:15 +0000</pubDate><link>https://news.ycombinator.com/item?id=45632715</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=45632715</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45632715</guid></item><item><title><![CDATA[Gemini and I Wrote a Book: Introduction to Computational Linguistics]]></title><description><![CDATA[
<p>Article URL: <a href="https://dubovik.eu/blog/computational-linguistics">https://dubovik.eu/blog/computational-linguistics</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=45632714">https://news.ycombinator.com/item?id=45632714</a></p>
<p>Points: 1</p>
<p># Comments: 1</p>
]]></description><pubDate>Sun, 19 Oct 2025 08:10:15 +0000</pubDate><link>https://dubovik.eu/blog/computational-linguistics</link><dc:creator>h45x1</dc:creator><comments>https://news.ycombinator.com/item?id=45632714</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45632714</guid></item></channel></rss>