<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: asaiacai</title><link>https://news.ycombinator.com/user?id=asaiacai</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Sat, 16 May 2026 10:42:45 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=asaiacai" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by asaiacai in "AI Product Graveyard"]]></title><description><![CDATA[
<p>this list has a lot of false positives.</p>
]]></description><pubDate>Tue, 05 May 2026 19:02:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=48026974</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=48026974</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48026974</guid></item><item><title><![CDATA[New comment by asaiacai in "Delve – Fake Compliance as a Service"]]></title><description><![CDATA[
<p>lmao</p>
]]></description><pubDate>Sat, 21 Mar 2026 03:07:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=47463623</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=47463623</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47463623</guid></item><item><title><![CDATA[New comment by asaiacai in "AI coding is gambling"]]></title><description><![CDATA[
<p>it's the perfect drug. You don't know how to code something up. Ask AI to implement it. It's broken? Ask AI to fix it for you. Will people become unable to fix things without it?</p>
]]></description><pubDate>Wed, 18 Mar 2026 22:43:09 +0000</pubDate><link>https://news.ycombinator.com/item?id=47432326</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=47432326</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47432326</guid></item><item><title><![CDATA[New comment by asaiacai in "Fault Tolerance Benchmark: Clockwork TorchPass, TorchFT and Checkpoint Restart"]]></title><description><![CDATA[
<p>have you guys tested this against any resource/batch managers in k8s (i.e. kueue, volcano, apache yunikorn)? seems like this is a good fit for people who have already have a large cluster. does it handle autoscaling environments well in your experience?</p>
]]></description><pubDate>Wed, 11 Mar 2026 20:19:47 +0000</pubDate><link>https://news.ycombinator.com/item?id=47340923</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=47340923</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47340923</guid></item><item><title><![CDATA[New comment by asaiacai in "Learnings from 4 months of Image-Video VAE experiments"]]></title><description><![CDATA[
<p>its cool to see the iterative improvements to your model laid out, but for everything that workedm i imagine there were at least a million other things you also tried but didnt work out. whats your process of trying these different techniques/architectures? do you just wait for one experiment to finish and visually inspect the results everytime. seems hard since these take a while to train. how do you shorten the feedback loop in this space?</p>
]]></description><pubDate>Wed, 25 Feb 2026 23:20:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=47159468</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=47159468</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47159468</guid></item><item><title><![CDATA[How Upc Barcodes Work]]></title><description><![CDATA[
<p>Article URL: <a href="https://craigball.net/2021/01/25/understanding-the-upc-because-you-can/">https://craigball.net/2021/01/25/understanding-the-upc-because-you-can/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=46835386">https://news.ycombinator.com/item?id=46835386</a></p>
<p>Points: 2</p>
<p># Comments: 0</p>
]]></description><pubDate>Sat, 31 Jan 2026 10:41:18 +0000</pubDate><link>https://craigball.net/2021/01/25/understanding-the-upc-because-you-can/</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=46835386</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46835386</guid></item><item><title><![CDATA[Kamaji: Containerized Control Planes for K8s]]></title><description><![CDATA[
<p>Article URL: <a href="https://kamaji.clastix.io/">https://kamaji.clastix.io/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=46804072">https://news.ycombinator.com/item?id=46804072</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Thu, 29 Jan 2026 00:43:06 +0000</pubDate><link>https://kamaji.clastix.io/</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=46804072</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46804072</guid></item><item><title><![CDATA[New comment by asaiacai in "Don't Become the Machine"]]></title><description><![CDATA[
<p>echoing this sentiment. I really do believe at some level there is strength/wisdom in being able to step away from a problem and return to it from a new perspective despite of what narratives are being pushed online by hustle</p>
]]></description><pubDate>Wed, 24 Dec 2025 09:03:39 +0000</pubDate><link>https://news.ycombinator.com/item?id=46373858</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=46373858</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46373858</guid></item><item><title><![CDATA[Debugging TLS failures in distroless containers]]></title><description><![CDATA[
<p>Article URL: <a href="https://lucabaggi.com/posts/ssl-docker/">https://lucabaggi.com/posts/ssl-docker/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=46225406">https://news.ycombinator.com/item?id=46225406</a></p>
<p>Points: 28</p>
<p># Comments: 0</p>
]]></description><pubDate>Wed, 10 Dec 2025 23:16:26 +0000</pubDate><link>https://lucabaggi.com/posts/ssl-docker/</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=46225406</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46225406</guid></item><item><title><![CDATA[Glibc malloc performance degradation with CPU affinity masks]]></title><description><![CDATA[
<p>Article URL: <a href="https://bugs.launchpad.net/ubuntu/+source/glibc/+bug/2089789">https://bugs.launchpad.net/ubuntu/+source/glibc/+bug/2089789</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=45739579">https://news.ycombinator.com/item?id=45739579</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Tue, 28 Oct 2025 21:38:42 +0000</pubDate><link>https://bugs.launchpad.net/ubuntu/+source/glibc/+bug/2089789</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=45739579</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45739579</guid></item><item><title><![CDATA[When Python can't thread: a deep-dive into the GIL's impact]]></title><description><![CDATA[
<p>Article URL: <a href="https://pythonspeed.com/articles/python-gil/">https://pythonspeed.com/articles/python-gil/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=45537494">https://news.ycombinator.com/item?id=45537494</a></p>
<p>Points: 3</p>
<p># Comments: 0</p>
]]></description><pubDate>Fri, 10 Oct 2025 11:08:28 +0000</pubDate><link>https://pythonspeed.com/articles/python-gil/</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=45537494</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45537494</guid></item><item><title><![CDATA[New comment by asaiacai in "A handy metric is needed for gauging if GPUs are being used optimally"]]></title><description><![CDATA[
<p>MFU is probably the best but requires application logic. You can export metrics at the infra level like SM efficiency. We explain it a bit how we used it to do some optimization.<p><a href="https://www.trainy.ai/blog/gpu-utilization-misleading">https://www.trainy.ai/blog/gpu-utilization-misleading</a></p>
]]></description><pubDate>Fri, 23 May 2025 03:06:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=44069494</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=44069494</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44069494</guid></item><item><title><![CDATA[New comment by asaiacai in "Ask HN: What fiction/non-fiction book should everyone read on the topic of CS?"]]></title><description><![CDATA[
<p>John Ousterhout's "A Philosophy of Software Design" I liked. It was supposed to be assigned reading for Berkeley's data structures class CS61B, and I don't think I really internalized the lessons within, but after re-reading it recently, I appreciated it a lot more and found the material transcends how to write code but also how to architect things as well.</p>
]]></description><pubDate>Tue, 18 Feb 2025 01:21:06 +0000</pubDate><link>https://news.ycombinator.com/item?id=43085063</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=43085063</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=43085063</guid></item><item><title><![CDATA[New comment by asaiacai in "GPU utilization can be a misleading metric"]]></title><description><![CDATA[
<p>power is also a good proxy. For example, we've had distributed runs that we monitored on WandB where one of our workers died in the middle and the rest were basically stalling on the dead worker. On WandB, we were only logging GPU stats on one worker and that one had 100% util but basically no excess power draw compared to having nothing running, which is how I found out something was stalling. Restarting fixed it and got the power draw up to normal, but even with high power draw, we were still having some sections of code with low SM efficiency (~20%) for that training.</p>
]]></description><pubDate>Fri, 23 Aug 2024 17:30:26 +0000</pubDate><link>https://news.ycombinator.com/item?id=41331029</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=41331029</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41331029</guid></item><item><title><![CDATA[Skypilot on Kubernetes]]></title><description><![CDATA[
<p>Article URL: <a href="https://blog.skypilot.co/ai-on-kubernetes/">https://blog.skypilot.co/ai-on-kubernetes/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=41323468">https://news.ycombinator.com/item?id=41323468</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Thu, 22 Aug 2024 19:07:11 +0000</pubDate><link>https://blog.skypilot.co/ai-on-kubernetes/</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=41323468</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41323468</guid></item><item><title><![CDATA[New comment by asaiacai in "GPU utilization can be a misleading metric"]]></title><description><![CDATA[
<p>totally agreed. A lot of our findings during this process is that there's still a lot of alpha in finding the right kernels for the job/model. We're hoping that in the future `torch.compile` will become more mature because current docs on performance at least on pytorch side definitely leave us wanting more</p>
]]></description><pubDate>Thu, 22 Aug 2024 17:38:27 +0000</pubDate><link>https://news.ycombinator.com/item?id=41322617</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=41322617</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41322617</guid></item><item><title><![CDATA[GPU utilization is a misleading metric. DCGM and Konduktor]]></title><description><![CDATA[
<p>Article URL: <a href="https://trainy.ai/blog/instrumentation">https://trainy.ai/blog/instrumentation</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=41073891">https://news.ycombinator.com/item?id=41073891</a></p>
<p>Points: 4</p>
<p># Comments: 1</p>
]]></description><pubDate>Thu, 25 Jul 2024 22:07:31 +0000</pubDate><link>https://trainy.ai/blog/instrumentation</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=41073891</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41073891</guid></item><item><title><![CDATA[New comment by asaiacai in "NVIDIA Transitions Fully Towards Open-Source Linux GPU Kernel Modules"]]></title><description><![CDATA[
<p>I really hope this makes it easier to install/upgrade NVIDIA drivers on Linux. It's a nightmare to figure out version mismatches between drivers, utils, container-runtime...</p>
]]></description><pubDate>Wed, 17 Jul 2024 22:27:14 +0000</pubDate><link>https://news.ycombinator.com/item?id=40990936</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=40990936</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40990936</guid></item><item><title><![CDATA[Show HN: OSS tool to finetune and serve LLMs on different clouds]]></title><description><![CDATA[
<p>Article URL: <a href="https://llm-atc.readthedocs.io/en/latest/">https://llm-atc.readthedocs.io/en/latest/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=37126004">https://news.ycombinator.com/item?id=37126004</a></p>
<p>Points: 12</p>
<p># Comments: 0</p>
]]></description><pubDate>Mon, 14 Aug 2023 19:54:58 +0000</pubDate><link>https://llm-atc.readthedocs.io/en/latest/</link><dc:creator>asaiacai</dc:creator><comments>https://news.ycombinator.com/item?id=37126004</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=37126004</guid></item></channel></rss>