<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: rdli</title><link>https://news.ycombinator.com/user?id=rdli</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Thu, 19 Nov 2026 21:32:20 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=rdli" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by rdli in "Getting the most out of Opus 5.5 in Claude and Claude Code"]]></title><description><![CDATA[
<p>I would also say: I think this works great because the goal is well-defined and measurable.<p>I’ve also used Opus 5.5 on some hill-climbing, and a lot more steering is required here, because … eval is hard.</p>
]]></description><pubDate>Sat, 03 Oct 2026 21:47:22 +0000</pubDate><link>https://news.ycombinator.com/item?id=49948111</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=49948111</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49948111</guid></item><item><title><![CDATA[New comment by rdli in "Getting the most out of Opus 5.5 in Claude and Claude Code"]]></title><description><![CDATA[
<p>I’m on the $100/month subscription; this session took about $500 in token-equivalent costs.<p>(Note that it wasn’t all Opus 5.5; I have a setup that uses Fable 5.1 as an advisor, Sonnet 5.5 for mechanical changes, etc.)</p>
]]></description><pubDate>Sat, 03 Oct 2026 21:44:46 +0000</pubDate><link>https://news.ycombinator.com/item?id=49948095</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=49948095</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49948095</guid></item><item><title><![CDATA[New comment by rdli in "Getting the most out of Opus 5.5 in Claude and Claude Code"]]></title><description><![CDATA[
<p>Yes. It spawned multiple subagents to run different experiments to benchmark a lot of different things, reviewed CI logs from past runs, etc. In the end, there were changes to what/how we cached, various code quality checks, speeding up test runners, and many other things.</p>
]]></description><pubDate>Sat, 03 Oct 2026 20:59:42 +0000</pubDate><link>https://news.ycombinator.com/item?id=49947728</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=49947728</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49947728</guid></item><item><title><![CDATA[New comment by rdli in "Getting the most out of Opus 5.5 in Claude and Claude Code"]]></title><description><![CDATA[
<p>It’s a really good model. Over the past few days, I give Opus some general directives to basically speed up our CI, and telling it I care both about billing minutes and wall clock time. I told it to create a plan after analyzing everything in our CI, run the plan by a Fable subagent, and then focus on low-risk, high-reward changes.<p>9 hours later, I had 12 PRs ready to be merged, and the net result is CI time has dropped from ~10 minutes to ~4 minutes, and billing minutes have dropped around 60%. Less than an hour of my attention.</p>
]]></description><pubDate>Sat, 03 Oct 2026 20:49:34 +0000</pubDate><link>https://news.ycombinator.com/item?id=49947634</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=49947634</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49947634</guid></item><item><title><![CDATA[New comment by rdli in "DeepMind's WeatherNext model achieves breakthrough forecasting cyclones"]]></title><description><![CDATA[
<p>I couldn't find a client of any sort for WeatherNext, so I had Claude write a simple one: <a href="https://github.com/richarddli/weatherodds" rel="nofollow">https://github.com/richarddli/weatherodds</a>.</p>
]]></description><pubDate>Sun, 09 Aug 2026 19:44:23 +0000</pubDate><link>https://news.ycombinator.com/item?id=49235020</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=49235020</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49235020</guid></item><item><title><![CDATA[New comment by rdli in "Claude Tag"]]></title><description><![CDATA[
<p>One thing that wasn't obvious in the post: it's usage (token) based pricing, not including in your existing Claude subscriptions.</p>
]]></description><pubDate>Tue, 23 Jun 2026 21:21:17 +0000</pubDate><link>https://news.ycombinator.com/item?id=48651631</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=48651631</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48651631</guid></item><item><title><![CDATA[New comment by rdli in "Ask HN: Who is hiring? (February 2026)"]]></title><description><![CDATA[
<p>Cybersecurity/AI seed startup | Founding AI Engineer | Bay Area | Full-time<p>We're seed-staged, 3 people, building an AI for cybersecurity, looking for a founding AI engineer who wants to learn/apply SOTA techniques for AI. Ideal background is experience building production agentic AI systems (by this I mean something like Simon's definition: <a href="https://simonwillison.net/2025/Sep/18/agents/" rel="nofollow">https://simonwillison.net/2025/Sep/18/agents/</a> definition) who also likes to think about WHAT to build and not just how.<p>We are a Golang/Python shop (although I'm not sure that matters so much any more).<p>Email jobs@polarsky.ai with subject HN.</p>
]]></description><pubDate>Mon, 02 Feb 2026 19:17:53 +0000</pubDate><link>https://news.ycombinator.com/item?id=46859992</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=46859992</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46859992</guid></item><item><title><![CDATA[New comment by rdli in "Notion AI: Unpatched data exfiltration"]]></title><description><![CDATA[
<p>I get that but just not entirely obvious how you do that for the Notion AI.</p>
]]></description><pubDate>Thu, 08 Jan 2026 01:28:52 +0000</pubDate><link>https://news.ycombinator.com/item?id=46535888</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=46535888</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46535888</guid></item><item><title><![CDATA[New comment by rdli in "Notion AI: Unpatched data exfiltration"]]></title><description><![CDATA[
<p>Securing LLMs is just structurally different. The attack space is "the entirety of the human written language" which is effectively infinite. Wrapping your head around this is something we're only now starting to appreciate.<p>In general, treating LLM outputs (no matter where) as untrusted, and ensuring classic cybersecurity guardrails (sandboxing, data permissioning, logging) is the current SOTA on mitigation. It'll be interesting to see how approaches evolve as we figure out more.</p>
]]></description><pubDate>Wed, 07 Jan 2026 22:39:47 +0000</pubDate><link>https://news.ycombinator.com/item?id=46534195</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=46534195</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46534195</guid></item><item><title><![CDATA[New comment by rdli in "Skills Officially Comes to Codex"]]></title><description><![CDATA[
<p>Nothing super-fancy. We have a common GitHub repo in our org for skills, and everyone checks out the repo into their preferred setup locally.<p>(To clarify, I meant that some engineers mostly use CC while others mostly use Codex, as opposed to engineers using both at the same time.)</p>
]]></description><pubDate>Sat, 20 Dec 2025 13:21:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=46335998</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=46335998</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46335998</guid></item><item><title><![CDATA[New comment by rdli in "Skills Officially Comes to Codex"]]></title><description><![CDATA[
<p>This is great. At my startup, we have a mix of Codex/CC users so having a common set of skills we can all use for building is exciting.<p>It’s also interesting to see how instead of a plan mode like CC, Codex is implementing planning as a skill.</p>
]]></description><pubDate>Sat, 20 Dec 2025 12:27:42 +0000</pubDate><link>https://news.ycombinator.com/item?id=46335695</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=46335695</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46335695</guid></item><item><title><![CDATA[Security Flaws in DeepSeek-Generated Code Linked to Political Triggers]]></title><description><![CDATA[
<p>Article URL: <a href="https://www.crowdstrike.com/en-us/blog/crowdstrike-researchers-identify-hidden-vulnerabilities-ai-coded-software/">https://www.crowdstrike.com/en-us/blog/crowdstrike-researchers-identify-hidden-vulnerabilities-ai-coded-software/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=46005687">https://news.ycombinator.com/item?id=46005687</a></p>
<p>Points: 3</p>
<p># Comments: 0</p>
]]></description><pubDate>Fri, 21 Nov 2025 15:59:23 +0000</pubDate><link>https://www.crowdstrike.com/en-us/blog/crowdstrike-researchers-identify-hidden-vulnerabilities-ai-coded-software/</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=46005687</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46005687</guid></item><item><title><![CDATA[New comment by rdli in "Ask HN: Who is hiring? (November 2025)"]]></title><description><![CDATA[
<p>Polar Sky | Bay Area | Full-time | Founding AI Engineer<p>Generative AI is rewriting how organizations use data, and breaking traditional security models in the process. We’re a team of cybersecurity, AI, and systems experts building the foundation for secure, trustworthy AI in the enterprise.<p>We're looking for a Founding AI Engineer who loves building with AI -- crafting context pipelines, integrating and evaluating LLMs into production systems, and delivering AI-native product experiences. You'll work on all parts of Polar Sky, from the data and eval systems to the reasoning, retrieval, and orchestration systems.<p>Apply online here: <a href="https://ats.rippling.com/polar-sky/jobs/a04ed5b7-6202-45e6-b6e5-389a75d01e20" rel="nofollow">https://ats.rippling.com/polar-sky/jobs/a04ed5b7-6202-45e6-b...</a>.</p>
]]></description><pubDate>Mon, 03 Nov 2025 16:39:04 +0000</pubDate><link>https://news.ycombinator.com/item?id=45801006</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=45801006</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45801006</guid></item><item><title><![CDATA[New comment by rdli in "Ask HN: Who is hiring? (October 2025)"]]></title><description><![CDATA[
<p>Polar Sky | Founding AI Lead | Bay Area/Seattle | Hybrid/Onsite | Full-time<p>We're a well-funded, pre-seed cybersecurity startup focused on data security. I'm looking for a founding AI lead with experience in fine-tuning LLMs (expertise around RL + reasoning models a big plus). This person would own the full AI stack from data to training to eval to test-time compute.<p>Who's a good fit:<p>* If you've always thought about starting a company, but for whatever reason (funding, life, idea), this is a great opportunity to be part of the founding team. We're 2 people right now.<p>* You enjoy understanding customer problems and their use cases, and then figuring out the best solution (sometimes technical, sometimes not) to their problems.<p>* You want to help figure out what a company looks like in this AI era.<p>* You enjoy teaching and sharing knowledge.<p>Questions, interest, just email jobs@polarsky.ai.</p>
]]></description><pubDate>Wed, 01 Oct 2025 23:47:41 +0000</pubDate><link>https://news.ycombinator.com/item?id=45444965</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=45444965</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45444965</guid></item><item><title><![CDATA[Secret Service dismantles telecom threat capable of crippling cell service in NY]]></title><description><![CDATA[
<p>Article URL: <a href="https://www.politico.com/news/2025/09/23/secret-service-un-nyc-telecom-00576100">https://www.politico.com/news/2025/09/23/secret-service-un-nyc-telecom-00576100</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=45354995">https://news.ycombinator.com/item?id=45354995</a></p>
<p>Points: 2</p>
<p># Comments: 4</p>
]]></description><pubDate>Wed, 24 Sep 2025 01:06:55 +0000</pubDate><link>https://www.politico.com/news/2025/09/23/secret-service-un-nyc-telecom-00576100</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=45354995</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45354995</guid></item><item><title><![CDATA[New comment by rdli in "Ask HN: Who is hiring? (September 2025)"]]></title><description><![CDATA[
<p>Polar Sky | Founding AI Lead | Bay Area/Seattle | Hybrid/Onsite | Full-time<p>We're a well-funded, pre-seed cybersecurity startup focused on data security. I'm looking for a founding AI lead with experience in fine-tuning LLMs (expertise around RL + reasoning models a big plus). This person would own the full AI stack from data to training to eval to test-time compute.<p>Who's a good fit:<p>* If you've always thought about starting a company, but for whatever reason (funding, life, idea), this is a great opportunity to be part of the founding team. We're 2 people right now.<p>* You enjoy understanding customer problems and their use cases, and then figuring out the best solution (sometimes technical, sometimes not) to their problems.<p>* You want to help figure out what a company looks like in this AI era.<p>* You enjoy teaching and sharing knowledge.<p>Questions, interest, just email jobs@polarsky.ai.</p>
]]></description><pubDate>Tue, 02 Sep 2025 10:56:37 +0000</pubDate><link>https://news.ycombinator.com/item?id=45101341</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=45101341</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45101341</guid></item><item><title><![CDATA[New comment by rdli in "OpenAI to buy AI startup from Jony Ive"]]></title><description><![CDATA[
<p>Seems that OpenAI is acquiring Io for $6.4B in an all-equity deal.</p>
]]></description><pubDate>Wed, 21 May 2025 17:04:41 +0000</pubDate><link>https://news.ycombinator.com/item?id=44053550</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=44053550</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44053550</guid></item><item><title><![CDATA[New comment by rdli in "llm-d, Kubernetes native distributed inference"]]></title><description><![CDATA[
<p>I would think that that the NVidia Dynamo SDK (pipelines) is a big difference as well (<a href="https://github.com/ai-dynamo/dynamo/tree/main/deploy/sdk/docs/sdk">https://github.com/ai-dynamo/dynamo/tree/main/deploy/sdk/doc...</a>), or am I missing something?</p>
]]></description><pubDate>Tue, 20 May 2025 17:02:37 +0000</pubDate><link>https://news.ycombinator.com/item?id=44043676</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=44043676</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44043676</guid></item><item><title><![CDATA[New comment by rdli in "llm-d, Kubernetes native distributed inference"]]></title><description><![CDATA[
<p>In this analogy, Dynamo is most definitely not like Django. It includes inference aware routing, KV caching, etc. -- all the stuff you would need to run a modern SOTA inference stack.</p>
]]></description><pubDate>Tue, 20 May 2025 16:59:38 +0000</pubDate><link>https://news.ycombinator.com/item?id=44043644</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=44043644</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44043644</guid></item><item><title><![CDATA[New comment by rdli in "llm-d, Kubernetes native distributed inference"]]></title><description><![CDATA[
<p>This is really interesting. For SOTA inference systems, I've seen two general approaches:<p>* The "stack-centric" approach such as vLLM production stack, AIBrix, etc. These set up an entire inference stack for you including KV cache, routing, etc.<p>* The "pipeline-centric" approach such as NVidia Dynamo, Ray, BentoML. These give you more of an SDK so you can define inference pipelines that you can then deploy on your specific hardware.<p>It seems like LLM-d is the former. Is that right? What prompted you to go down that direction, instead of the direction of Dynamo?</p>
]]></description><pubDate>Tue, 20 May 2025 16:04:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=44043135</link><dc:creator>rdli</dc:creator><comments>https://news.ycombinator.com/item?id=44043135</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44043135</guid></item></channel></rss>