<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: auspiv</title><link>https://news.ycombinator.com/user?id=auspiv</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Sat, 13 Jun 2026 02:28:34 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=auspiv" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by auspiv in "Claude Code – Everything You Can Configure That the Docs Don't Tell You"]]></title><description><![CDATA[
<p>Wow, not one mention of the env vars that have a far greater influence on how the models actually work under the hood - <a href="https://code.claude.com/docs/en/env-vars" rel="nofollow">https://code.claude.com/docs/en/env-vars</a><p>Very important for bedrock deployments and other not-as-standard deployments<p>Key for how I've deployed it - disable adaptive thinking, max thinking tokens, disable telemetry, etc</p>
]]></description><pubDate>Fri, 29 May 2026 13:44:32 +0000</pubDate><link>https://news.ycombinator.com/item?id=48323034</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=48323034</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48323034</guid></item><item><title><![CDATA[New comment by auspiv in "GPT-5.5 Price Increase: What It Costs"]]></title><description><![CDATA[
<p>GPT-5.5 is a solid leap with Codex or other harnesses. Opus 4.7 I still don't understand how people use... I tried it for a day or two, have tried it for a few hours every week or so since release, and still use 4.6 as daily driver (with xhi thinking).</p>
]]></description><pubDate>Fri, 08 May 2026 14:30:21 +0000</pubDate><link>https://news.ycombinator.com/item?id=48063780</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=48063780</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48063780</guid></item><item><title><![CDATA[New comment by auspiv in "When everyone has AI and the company still learns nothing"]]></title><description><![CDATA[
<p>And their spreadsheets don't?</p>
]]></description><pubDate>Tue, 05 May 2026 18:10:48 +0000</pubDate><link>https://news.ycombinator.com/item?id=48026302</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=48026302</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48026302</guid></item><item><title><![CDATA[I swept all 667 layer duplication configs for Qwen3-4B on 1x RTX 3090]]></title><description><![CDATA[
<p>Article URL: <a href="https://austinsnerdythings.com/2026/04/14/rys-layer-duplication-qwen3-4b/">https://austinsnerdythings.com/2026/04/14/rys-layer-duplication-qwen3-4b/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47778817">https://news.ycombinator.com/item?id=47778817</a></p>
<p>Points: 3</p>
<p># Comments: 0</p>
]]></description><pubDate>Wed, 15 Apr 2026 13:39:13 +0000</pubDate><link>https://austinsnerdythings.com/2026/04/14/rys-layer-duplication-qwen3-4b/</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=47778817</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47778817</guid></item><item><title><![CDATA[One Layer, +12%: What 667 Configs Reveal About Small LLM Anatomy]]></title><description><![CDATA[
<p>Article URL: <a href="https://austinsnerdythings.com/2026/04/14/rys-layer-duplication-qwen3-4b/">https://austinsnerdythings.com/2026/04/14/rys-layer-duplication-qwen3-4b/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47771990">https://news.ycombinator.com/item?id=47771990</a></p>
<p>Points: 2</p>
<p># Comments: 0</p>
]]></description><pubDate>Tue, 14 Apr 2026 21:54:05 +0000</pubDate><link>https://austinsnerdythings.com/2026/04/14/rys-layer-duplication-qwen3-4b/</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=47771990</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47771990</guid></item><item><title><![CDATA[New comment by auspiv in "A macOS bug that causes TCP networking to stop working after 49.7 days"]]></title><description><![CDATA[
<p>probably same thing for boeing 787 jets - <a href="https://www.theregister.com/2020/04/02/boeing_787_power_cycle_51_days_stale_data" rel="nofollow">https://www.theregister.com/2020/04/02/boeing_787_power_cycl...</a><p>says 51 days, which would be an interesting number of (milli)seconds</p>
]]></description><pubDate>Mon, 06 Apr 2026 21:15:24 +0000</pubDate><link>https://news.ycombinator.com/item?id=47667232</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=47667232</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47667232</guid></item><item><title><![CDATA[New comment by auspiv in "Embarrassingly simple self-distillation improves code generation"]]></title><description><![CDATA[
<p>apparently you can straight up duplicate/add/rearrange layers without changing any of the weights and get better results as well - <a href="https://dnhkng.github.io/posts/rys/" rel="nofollow">https://dnhkng.github.io/posts/rys/</a></p>
]]></description><pubDate>Sat, 04 Apr 2026 16:16:40 +0000</pubDate><link>https://news.ycombinator.com/item?id=47640387</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=47640387</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47640387</guid></item><item><title><![CDATA[Show HN: Trained YOLOX from scratch to avoid Ultralytics (aircraft detection)]]></title><description><![CDATA[
<p>Article URL: <a href="https://austinsnerdythings.com/2026/02/13/training-yolox-aircraft-detection-mit-license/">https://austinsnerdythings.com/2026/02/13/training-yolox-aircraft-detection-mit-license/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47049703">https://news.ycombinator.com/item?id=47049703</a></p>
<p>Points: 2</p>
<p># Comments: 1</p>
]]></description><pubDate>Tue, 17 Feb 2026 16:58:39 +0000</pubDate><link>https://austinsnerdythings.com/2026/02/13/training-yolox-aircraft-detection-mit-license/</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=47049703</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47049703</guid></item><item><title><![CDATA[Show HN: Trained YOLOX from scratch to avoid Ultralytics (iOS aircraft detect)]]></title><description><![CDATA[
<p>Article URL: <a href="https://austinsnerdythings.com/2026/02/13/training-yolox-aircraft-detection-mit-license/">https://austinsnerdythings.com/2026/02/13/training-yolox-aircraft-detection-mit-license/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47014877">https://news.ycombinator.com/item?id=47014877</a></p>
<p>Points: 3</p>
<p># Comments: 0</p>
]]></description><pubDate>Sat, 14 Feb 2026 14:40:53 +0000</pubDate><link>https://austinsnerdythings.com/2026/02/13/training-yolox-aircraft-detection-mit-license/</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=47014877</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47014877</guid></item><item><title><![CDATA[New comment by auspiv in "Qwen3-Max-Thinking"]]></title><description><![CDATA[
<p>There have been a couple "studies" and comparing various frontier-tier AIs that have led to the conclusion that Chinese models are somewhere around 7-9 months behind US models. Other comment says that Opus will be at 5.2 by the time Qwen matches Opus 4.5. It's accurate, and there is some data to show by how much.</p>
]]></description><pubDate>Mon, 26 Jan 2026 16:57:00 +0000</pubDate><link>https://news.ycombinator.com/item?id=46768063</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=46768063</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46768063</guid></item><item><title><![CDATA[New comment by auspiv in "Building a 25 Gbit/s workstation for the SCION Association"]]></title><description><![CDATA[
<p>Wow, 249 CHF for 8x fans is insane. The grip Noctua has on people! Nice workstation.</p>
]]></description><pubDate>Mon, 12 Jan 2026 18:13:54 +0000</pubDate><link>https://news.ycombinator.com/item?id=46592095</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=46592095</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46592095</guid></item><item><title><![CDATA[New comment by auspiv in "Stop Doom Scrolling, Start Doom Coding: Build via the terminal from your phone"]]></title><description><![CDATA[
<p>Using this with tmux and various VPN tech. Main issue is scrolling. Termius + tmux don't scroll very well. And I've been led to believe tmux is necessary to keep sessions open when I turn off my phone screen</p>
]]></description><pubDate>Tue, 06 Jan 2026 20:01:21 +0000</pubDate><link>https://news.ycombinator.com/item?id=46517803</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=46517803</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46517803</guid></item><item><title><![CDATA[New comment by auspiv in "Show HN: I vibe-coded an aircraft AR tracking app and wasted weeks on AI bugs"]]></title><description><![CDATA[
<p>Thanks! How I eventually found it was stripping stuff back layer by layer. And by that I mean I started with the raw camera feed and got to where things worked well in a different swift view. And then from there, peeled stuff back from the main process feature by feature. And then bam, aircraft were exactly where they should be (minus the compass inaccuracy). I even had stuff like drawing mountain peaks (I live near Denver) as "aircraft" to figure things out, determining different FOV at different zoom levels (a lot of AI keyed in where the boxes would move in one direction at low zooms, be completely correct at some middle zoom, and then in the opposite direction at high zoom).<p>And that peeling back was me looking at each function to see what it did (I am a dev, but not for SwiftUI). So yep, can't vibe code it all!</p>
]]></description><pubDate>Fri, 19 Dec 2025 17:10:36 +0000</pubDate><link>https://news.ycombinator.com/item?id=46328226</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=46328226</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46328226</guid></item><item><title><![CDATA[Show HN: I vibe-coded an aircraft AR tracking app and wasted weeks on AI bugs]]></title><description><![CDATA[
<p>Built an app entirely with Claude/AI assistance – backend (Django + C#), iOS frontend, server deployment, CI/CD pipeline, the works. Hosted on a single VPS. Postgres on VPS, Redis on VPS, Django on VPS, etc. The VPS is a VM I have in a proxmox server I have sitting in a datacenter (Dell R630, 1x Xeon 2697v4, 128GB memory, 6x 960GB Intel D3-S4610 with Optane SLOG, etc). No AWS/GCP/Vercel, etc. Incremental cost to me = $0/month. I skipped using CloudFlare tunnels for this - hoping I don't regret that.<p>What it does: Point your phone at the sky, see real-time aircraft info overlaid on your camera. ADS-B data from community feeders, WebSocket streaming, kinematic prediction to smooth positions between updates. No ARKit – just AVFoundation camera + CoreLocation/CoreMotion + math. SwiftUI overlays positioned via GPS/heading projection.<p>The humbling part: Spent 2 months debugging "FOV calibration issues." Built an 800-line calibration UI, a Flask debug server, Jupyter notebooks for pipeline analysis, extensive logging infrastructure. Hung a literal picture on the wall with a black rectangle of specific size to "calibrate" the FOV reported by my phone. The AI helped me build all of it beautifully.<p>The actual bug? A UI scale factor on the overlay boxes. Not FOV math. Not coordinate projection. Just scaleEffect() making things the wrong size. Commit message when I found it: "scale may have been fucking me over for a long time for 'fov issues'". Guess where the scaleEffect() function was introduced? That's right - AI generated. I asked it at one point something along the lines of "ok when you draw the boxes around the aircraft, make them smaller when the aircraft is farther away".<p>I went through 2-3 major model releases that I tested on this "hey I've been fighting a FOV bug for a while - can you please take a look and let me know if any issues jump out". Gemini 3 Pro, Opus 4.5, none of them found the "bug".<p>Takeaways from vibe-coding a full product:<p>- AI is incredible at building things fast – entire features in minutes. The entire UI, website, logo, etc, all AI. Claude Opus 4.5 kind of sucks at UI. Gemini 3 cleaned all that up.<p>- AI will also confidently help you debug the wrong thing for weeks<p>- Still need to know when to step back and question your assumptions<p>- Deleted 2,700 lines of debug infrastructure once I found the real bug<p>- Low performance? Just tell AI to rewrite it in a more performance language (load tested the process with 1000 connections - with Python/Django, tons of drops and latency spikes to 5000ms. Switched to c# and now it'll do 1000 and keep latency under 300ms)<p>Release process: painless, except for the test RevenueCat SDK key causing instacrash. Didn't test release locally. Approved in 6 minutes 2nd submission.<p>Question: what are people using to get super accurate heading out of Apple devices? The heading estimated error never drops below 10°. It's about 50/50 on being spot on vs not that close for the projections.<p>App link: <a href="https://apps.apple.com/us/app/skyspottr/id6756084687">https://apps.apple.com/us/app/skyspottr/id6756084687</a></p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=46327046">https://news.ycombinator.com/item?id=46327046</a></p>
<p>Points: 3</p>
<p># Comments: 2</p>
]]></description><pubDate>Fri, 19 Dec 2025 15:44:22 +0000</pubDate><link>https://news.ycombinator.com/item?id=46327046</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=46327046</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46327046</guid></item><item><title><![CDATA[New comment by auspiv in "Carrier Landing in Top Gun for the NES"]]></title><description><![CDATA[
<p>ZScaler (at work) classifies this whole site as<p>Website blocked<p>Not allowed to browse Weapons/Bombs category<p>You tried to visit:<a href="https://relaxing.run/blag/posts/top-gun-landing/" rel="nofollow">https://relaxing.run/blag/posts/top-gun-landing/</a></p>
]]></description><pubDate>Mon, 15 Dec 2025 21:00:50 +0000</pubDate><link>https://news.ycombinator.com/item?id=46280575</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=46280575</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46280575</guid></item><item><title><![CDATA[New comment by auspiv in "New benchmark shows top LLMs struggle in real mental health care"]]></title><description><![CDATA[
<p>And real therapists are good right?</p>
]]></description><pubDate>Wed, 10 Dec 2025 15:06:23 +0000</pubDate><link>https://news.ycombinator.com/item?id=46218486</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=46218486</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46218486</guid></item><item><title><![CDATA[New comment by auspiv in "A supersonic engine core makes the perfect power turbine"]]></title><description><![CDATA[
<p>a 16" natural gas pipeline moving 200 MMscf/d at pressures >1000 psi (relatively standard numbers, 16" can go up to ~600 MMscf/d) has a power transport capability of 2.5 GW (thermal). burn that in 60% efficient combined cycle turbines and you get 1 GW of electricity. that's a lot easier than building 1 GW of electric transmission lines.</p>
]]></description><pubDate>Wed, 10 Dec 2025 14:39:14 +0000</pubDate><link>https://news.ycombinator.com/item?id=46218206</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=46218206</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46218206</guid></item><item><title><![CDATA[New comment by auspiv in "Running on Empty: Copper"]]></title><description><![CDATA[
<p>Other countries are very much the same. Almost always located near giant hydroelectric generation facilities. Brazil + Russia are two big ones that come to mind. Probably China too.</p>
]]></description><pubDate>Mon, 08 Dec 2025 22:11:56 +0000</pubDate><link>https://news.ycombinator.com/item?id=46198342</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=46198342</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46198342</guid></item><item><title><![CDATA[New comment by auspiv in "RAM is so expensive, Samsung won't even sell it to Samsung"]]></title><description><![CDATA[
<p>DDR4 prices are up 2-6x in the last couple months depending on frequency. High end, high speed modules (e.g. 128GB 3200MHz LRDIMM) are super expensive.</p>
]]></description><pubDate>Thu, 04 Dec 2025 17:15:45 +0000</pubDate><link>https://news.ycombinator.com/item?id=46150052</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=46150052</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46150052</guid></item><item><title><![CDATA[Ask HN: Who's figured out using Claude Code via voice on mobile? e.g. on a walk]]></title><description><![CDATA[

<p>Comments URL: <a href="https://news.ycombinator.com/item?id=46121653">https://news.ycombinator.com/item?id=46121653</a></p>
<p>Points: 4</p>
<p># Comments: 0</p>
]]></description><pubDate>Tue, 02 Dec 2025 14:40:50 +0000</pubDate><link>https://news.ycombinator.com/item?id=46121653</link><dc:creator>auspiv</dc:creator><comments>https://news.ycombinator.com/item?id=46121653</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46121653</guid></item></channel></rss>