<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: tyzoid</title><link>https://news.ycombinator.com/user?id=tyzoid</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Thu, 19 Nov 2026 15:25:21 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=tyzoid" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by tyzoid in "Show HN: AI Red Lines – write down what would make you act on AI"]]></title><description><![CDATA[
<p>Is this it? <a href="https://notes.archie-hall.com/p/drawing-the-line-on-ai-writing-literally" rel="nofollow">https://notes.archie-hall.com/p/drawing-the-line-on-ai-writi...</a></p>
]]></description><pubDate>Thu, 12 Nov 2026 22:31:02 +0000</pubDate><link>https://news.ycombinator.com/item?id=50475308</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=50475308</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50475308</guid></item><item><title><![CDATA[New comment by tyzoid in "Is Opus 5.5 nerfed? Thinking patterns changed on November 3"]]></title><description><![CDATA[
<p>From experience, qwen3.8 27b with gsq/rco at iq3_xxs has good performance with little intelligence loss.</p>
]]></description><pubDate>Mon, 09 Nov 2026 20:32:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=50431616</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=50431616</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50431616</guid></item><item><title><![CDATA[New comment by tyzoid in "MaxxBench – "Nice DeepSWE score. But is it true?""]]></title><description><![CDATA[
<p>Not everyone is operating with the same constraints. As for me, once the models got good enough (~opus 4.5 levels of intelligence) that they could solve most problems that I was dealing with, I stopped following the frontier intelligence trajectory for daily use and instead followed constant intelligence and decreasing cost. Nowadays I primarily use GPT 6.1 luna, and DeepSeek 4.1 flash.<p>I've also been very happy playing around with qwen 3.8 27b running locally for personal projects. It sometimes can't quite get there, and that's when I'll reach for GPT.</p>
]]></description><pubDate>Sat, 07 Nov 2026 20:22:54 +0000</pubDate><link>https://news.ycombinator.com/item?id=50407709</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=50407709</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50407709</guid></item><item><title><![CDATA[New comment by tyzoid in "Transit rewards"]]></title><description><![CDATA[
<p>Are you factoring in fixed costs to that estimate? That may be partially misleading if so.</p>
]]></description><pubDate>Wed, 23 Sep 2026 04:27:39 +0000</pubDate><link>https://news.ycombinator.com/item?id=49811644</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=49811644</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49811644</guid></item><item><title><![CDATA[New comment by tyzoid in "Luanti removed from Google Play due to baseless AI copyright notice"]]></title><description><![CDATA[
<p>I suspect the TOS say google can remove or deny apps for any reason. If in the EU, there might be a DMA claim though.</p>
]]></description><pubDate>Fri, 28 Aug 2026 12:08:40 +0000</pubDate><link>https://news.ycombinator.com/item?id=49477359</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=49477359</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49477359</guid></item><item><title><![CDATA[New comment by tyzoid in "Study reveals UnitedHealth's profit margins four times what it claimed [pdf]"]]></title><description><![CDATA[
<p>No - this is a false equivalence. Transaction processing companies for the most part handle the in-and-out flows as a single transaction. Insurance companies hold on to the premium pool ("float") for long enough that they have time to realize gains from investing portions of it - the inflow and outflow are very separate.</p>
]]></description><pubDate>Wed, 26 Aug 2026 19:33:57 +0000</pubDate><link>https://news.ycombinator.com/item?id=49454586</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=49454586</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49454586</guid></item><item><title><![CDATA[New comment by tyzoid in "Xorg-server 26.1.0 rc1"]]></title><description><![CDATA[
<p>Isn't most of this already in xlibre?</p>
]]></description><pubDate>Thu, 20 Aug 2026 15:22:32 +0000</pubDate><link>https://news.ycombinator.com/item?id=49375935</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=49375935</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49375935</guid></item><item><title><![CDATA[New comment by tyzoid in "Brain turns listening inward during REM sleep, EEG recordings suggest"]]></title><description><![CDATA[
<p>Out of curiousity, how does this apply to nightmares ending in certain death or grevious injury (driving off a bridge, falling down a well, being chased by zombies and unable to move, etc.)?</p>
]]></description><pubDate>Sat, 15 Aug 2026 16:22:34 +0000</pubDate><link>https://news.ycombinator.com/item?id=49311868</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=49311868</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49311868</guid></item><item><title><![CDATA[New comment by tyzoid in "LLMs reward expertise"]]></title><description><![CDATA[
<p>Harness refers to the tooling that allows you to interact with an LLM. Web chat interface is a harness, CLI coding tools (claude code, codex, opencode, pi, etc.) are harnesses, agent systems (openclaw, hermes, etc.) are harnesses. They present different capabilities to the underlying LLM. Codex, for example, is more likely to write code if you say "I want to build an app" than ChatGPT on web will.</p>
]]></description><pubDate>Tue, 04 Aug 2026 06:06:39 +0000</pubDate><link>https://news.ycombinator.com/item?id=49164818</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=49164818</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49164818</guid></item><item><title><![CDATA[New comment by tyzoid in "Show HN: Reame – a CPU inference server that gets faster as it runs"]]></title><description><![CDATA[
<p>Why qwen 2.5 everywhere? Why not 3.5?</p>
]]></description><pubDate>Sat, 11 Jul 2026 19:02:25 +0000</pubDate><link>https://news.ycombinator.com/item?id=48874748</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=48874748</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48874748</guid></item><item><title><![CDATA[New comment by tyzoid in "Show HN: Pulpie – Models for Cleaning the Web"]]></title><description><![CDATA[
<p>How does this work on pages that require JavaScript in order to render?</p>
]]></description><pubDate>Mon, 06 Jul 2026 20:16:51 +0000</pubDate><link>https://news.ycombinator.com/item?id=48809906</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=48809906</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48809906</guid></item><item><title><![CDATA[New comment by tyzoid in "GPT-5.6 Sol Ultra will be in Codex"]]></title><description><![CDATA[
<p>I'll sometimes use agent swarms for triaging quickly through a number of possible issues. For example, a single request for a code review to uncover all possible issues, followed by an agent swarm to dig deep into each issue to confirm or mark false positive.</p>
]]></description><pubDate>Mon, 06 Jul 2026 20:12:14 +0000</pubDate><link>https://news.ycombinator.com/item?id=48809851</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=48809851</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48809851</guid></item><item><title><![CDATA[New comment by tyzoid in "Words Are a Byproduct of Consciousness. For LLMs, It's Backwards"]]></title><description><![CDATA[
<p>I have to disagree with the fundamental premise here. LLMs do indeed have an internal "concept" mapping - the embedding layer. There's some fun demos on youtube[1] where you can manipulate embedding vectors (such as king - man + woman = queen), and these manipulations are done entirely in the embedding/concept space.<p>[1] <a href="https://youtu.be/gQddtTdmG_8" rel="nofollow">https://youtu.be/gQddtTdmG_8</a> (around 14:00)</p>
]]></description><pubDate>Tue, 30 Jun 2026 17:20:40 +0000</pubDate><link>https://news.ycombinator.com/item?id=48735993</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=48735993</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48735993</guid></item><item><title><![CDATA[New comment by tyzoid in "Enough with the AI FOMO, go slow-mo, says Domo CDO"]]></title><description><![CDATA[
<p>Domo Says No Mo' to AI FOMO</p>
]]></description><pubDate>Mon, 18 May 2026 13:51:59 +0000</pubDate><link>https://news.ycombinator.com/item?id=48179922</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=48179922</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48179922</guid></item><item><title><![CDATA[New comment by tyzoid in "Gaining control of every projector and camera on campus"]]></title><description><![CDATA[
<p>Hey, fancy seeing you up here on HN! Fun article</p>
]]></description><pubDate>Sat, 16 May 2026 12:12:06 +0000</pubDate><link>https://news.ycombinator.com/item?id=48159480</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=48159480</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48159480</guid></item><item><title><![CDATA[New comment by tyzoid in "Agentic AI systems violate the implicit assumptions of database design"]]></title><description><![CDATA[
<p>You can use something like flyway on top of your existing git/cicd stack. Write the query as a migration, have it reviewed using your git code review process, and merge to run the migration.</p>
]]></description><pubDate>Sun, 26 Apr 2026 22:58:21 +0000</pubDate><link>https://news.ycombinator.com/item?id=47915690</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=47915690</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47915690</guid></item><item><title><![CDATA[New comment by tyzoid in "US v. Heppner (S.D.N.Y. 2026) no attorney-client privilege for AI chats [pdf]"]]></title><description><![CDATA[
<p>It doesn't even need to be a third party. Documents you make for yourself but never send to a third party could still be seized in a criminal context or subject to discovery in a civil one.<p>Attorney-client privilege is a special carve out because the courts have recognized that clients need to be open and honest with their attorneys to get proper counsel and representation.<p>This ruling is the court declining to extend that special carve out to non-lawyer AI tooling, and keeping the status quo of contemporaneous documents made by someone discoverable,  whether or not shared with a third party. The judge draws from the TOS as an admonishment, effectively saying (my words, not the judge) <i>the TOS should have put you on notice that you have no expectation this data is confidential</i>.</p>
]]></description><pubDate>Thu, 16 Apr 2026 13:59:47 +0000</pubDate><link>https://news.ycombinator.com/item?id=47793090</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=47793090</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47793090</guid></item><item><title><![CDATA[New comment by tyzoid in "SSL certificate requirements are becoming obnoxious"]]></title><description><![CDATA[
<p>Pretty much any legacy system can have a modern reverse proxy in front of it. If the legacy application can't handler certs sanely, use the reverse proxy for terminating TLS.</p>
]]></description><pubDate>Tue, 26 Aug 2025 14:17:13 +0000</pubDate><link>https://news.ycombinator.com/item?id=45026905</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=45026905</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45026905</guid></item><item><title><![CDATA[New comment by tyzoid in "12ft.io Taken Down"]]></title><description><![CDATA[
<p>Upper-end estimates for typical web traffic CPC/CPM rates are around 10¢ per click and 0.5¢ per impression. I never click on ads, but let's hypothesize that with very-well targeted ads 5% of people click through. That ends up at around an estimated 1¢ per page view. Even at 5x that rate, I'd be able to read 1000 articles ($5 USD) before even approaching the $6 USD/mo rate I think is reasonable.<p>That's over 30 articles per day, again at a 5x rate than advertising will return. Will there users that read vastly more than that? Sure. But there's also many readers that will under-utilize the service too.<p>Just take a look at how YouTube Premium is doing, many creators report that their premium revenue vastly outpaces ad-supported viewers on a per-view basis.<p>If the revenue doesn't make sense, then you could supplement the revenue with ads for users who exceed a soft cap, or have tiered subscriptions. Something like a basic (1k articles per month)ad-supported subscription for $4, basic ad-free for $6, and unlimited ad-free for $10.</p>
]]></description><pubDate>Mon, 21 Jul 2025 23:43:41 +0000</pubDate><link>https://news.ycombinator.com/item?id=44641658</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=44641658</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44641658</guid></item><item><title><![CDATA[New comment by tyzoid in "12ft.io Taken Down"]]></title><description><![CDATA[
<p>The problem is there's not really a good way to subscribe to these things. I'd gladly pay a nominal fee (~$6 USD/mo) for access to media, but I'm not about to subscribe individually to each site. Ideally, I'd subscribe to a single service and payment is split across the various sites in proportion to how many articles I read from each site.<p>There was a service that promised this a while back, but IIRC mozilla bought and killed it.</p>
]]></description><pubDate>Mon, 21 Jul 2025 19:47:52 +0000</pubDate><link>https://news.ycombinator.com/item?id=44639641</link><dc:creator>tyzoid</dc:creator><comments>https://news.ycombinator.com/item?id=44639641</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44639641</guid></item></channel></rss>