<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: lukebuehler</title><link>https://news.ycombinator.com/user?id=lukebuehler</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Thu, 19 Nov 2026 22:18:26 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=lukebuehler" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by lukebuehler in "Goldman Sachs is pulling Claude Code from its bankers and traders"]]></title><description><![CDATA[
<p>Exactly. People want to run agents where all their data is. It’s a very hard problem. Just yelling sandbox does not solve it.</p>
]]></description><pubDate>Mon, 09 Nov 2026 07:46:27 +0000</pubDate><link>https://news.ycombinator.com/item?id=50422388</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=50422388</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50422388</guid></item><item><title><![CDATA[New comment by lukebuehler in "Goldman Sachs is pulling Claude Code from its bankers and traders"]]></title><description><![CDATA[
<p>Remote-hosted agents are available in most places. BUT they are not interesting, because they do not have access to all the user's data.<p>People really want to run agents on their data.<p>Re sandboxing: I dont know Goldman's particulars, but most enterprises lock down all sandboxing tech on workstations. So I assume this is actually harder than it seems. So people just run it directly on the machine.<p>The article is referring to the dilemma of the trifecta twice.</p>
]]></description><pubDate>Sun, 08 Nov 2026 20:32:56 +0000</pubDate><link>https://news.ycombinator.com/item?id=50416871</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=50416871</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50416871</guid></item><item><title><![CDATA[New comment by lukebuehler in "Goldman Sachs is pulling Claude Code from its bankers and traders"]]></title><description><![CDATA[
<p>Cloud agents or even sandboxes are too technical for most. And most corporate workstations do not allow sandboxing of any kind (virtualization or containers).</p>
]]></description><pubDate>Sun, 08 Nov 2026 20:26:25 +0000</pubDate><link>https://news.ycombinator.com/item?id=50416794</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=50416794</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50416794</guid></item><item><title><![CDATA[New comment by lukebuehler in "Goldman Sachs is pulling Claude Code from its bankers and traders"]]></title><description><![CDATA[
<p>Yeah, I'm working on an open source product in this space [0]--particularly in finance--that's why I know about this problem quite a bit.<p>I think the solutions is to give people the ability to run agents that are "coding agent like" but in managed and governed infrastructure (i.e. not on their workstations). Basically, Meta Muse or OAI dots for enterprise.<p>It will take some time to find the right architectures and patterns. IMO, coding agents were just the first wave, much more interesting stuff and systems are coming.<p>[0] <a href="https://ls.bot" rel="nofollow">https://ls.bot</a>, <a href="https://github.com/smartcomputer-ai/lightspeed" rel="nofollow">https://github.com/smartcomputer-ai/lightspeed</a></p>
]]></description><pubDate>Sun, 08 Nov 2026 20:22:35 +0000</pubDate><link>https://news.ycombinator.com/item?id=50416759</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=50416759</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50416759</guid></item><item><title><![CDATA[New comment by lukebuehler in "Goldman Sachs is pulling Claude Code from its bankers and traders"]]></title><description><![CDATA[
<p>People are giving all kinds of explanations here why Goldman is doing this, but they are not addressing what the article (implicitly) says: operating coding agents on the primary machines of finance professionals runs directly into the lethal trifecta. It's as simple as that.<p>You cannot have semi-unattended agents running on the same workstations that have access to client/trade/prop information and email client, and internet, and internal networks, it's too dangerous--a major leak is just one small mistake or prompt injection away. Banks are not stupid, they are getting it and shutting it down.</p>
]]></description><pubDate>Sun, 08 Nov 2026 20:03:11 +0000</pubDate><link>https://news.ycombinator.com/item?id=50416595</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=50416595</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50416595</guid></item><item><title><![CDATA[New comment by lukebuehler in "Using AI for just 10 minutes erodes your ability to persist at hard things"]]></title><description><![CDATA[
<p>Exactly! The sheer effort of having AI produce good results, from writing, to planning, to coding is immense. I have consistently gather all my energies and focus to guide agents to the ideal, maintainable, and usable solution. Often it is me that makes the crucial design contribution.<p>Yes, I could be lazy and just vibe piles of slop. But anyone who spends any meaningful amount of time using AI (more than the articles 10 minutes), knows that quite extreme concentration and effort is required to build ambitiously with agents.<p>I feel like I work harder and and with more difficult problems than before AI, just at a much faster pace and scope.</p>
]]></description><pubDate>Sun, 11 Oct 2026 11:33:48 +0000</pubDate><link>https://news.ycombinator.com/item?id=50042098</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=50042098</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50042098</guid></item><item><title><![CDATA[New comment by lukebuehler in "What is Codemode"]]></title><description><![CDATA[
<p>I agree. Harnesses are quickly becoming the next systems programming frontier. The OS analogy is apt and often made--just like browsers have been compared to becoming "platforms" and "OSes". Another analogy that is obvious is threads/processes and agents and sub-agents: all the concurrency, shared resource  access, cooperative multitasking, etc problems are emerging again.<p>It will take some time to shake out, but lots of old ideas are new again. Personally, I'm mining lisp machines, homoiconicity, actor models, and other ideas for my harness designs.</p>
]]></description><pubDate>Wed, 07 Oct 2026 09:22:10 +0000</pubDate><link>https://news.ycombinator.com/item?id=49990254</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49990254</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49990254</guid></item><item><title><![CDATA[New comment by lukebuehler in "Sites in ChatGPT"]]></title><description><![CDATA[
<p>oof, the post smelled AI generated to me, more in content than voice. so instead of writing software, his agents are actually karma farming</p>
]]></description><pubDate>Sat, 03 Oct 2026 15:35:24 +0000</pubDate><link>https://news.ycombinator.com/item?id=49945185</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49945185</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49945185</guid></item><item><title><![CDATA[New comment by lukebuehler in "Pi Durable"]]></title><description><![CDATA[
<p>Not infinitely, but 6-12 hour long individual runs: complex analysis in enterprise across many data sources. Basically, long running investigations that touch databases, many files, apps via computer use, and so on.</p>
]]></description><pubDate>Fri, 02 Oct 2026 07:28:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=49930772</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49930772</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49930772</guid></item><item><title><![CDATA[New comment by lukebuehler in "Pi Durable"]]></title><description><![CDATA[
<p>What is interesting here is the _library_ approach to durable agents. All the other options I listed, including mine, take more of an SDK/batteries included approach. So this is very much in line with the Pi philosophy in general.<p>So, it'll be interesting to see if these durable agent setups need more of a complete product approach, or if people want to compose them as libraries.</p>
]]></description><pubDate>Fri, 02 Oct 2026 07:24:52 +0000</pubDate><link>https://news.ycombinator.com/item?id=49930748</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49930748</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49930748</guid></item><item><title><![CDATA[New comment by lukebuehler in "Pi 1.0"]]></title><description><![CDATA[
<p>I think you should look into Pi Durable wich was just released right now too. seems like it is made for your use-case.</p>
]]></description><pubDate>Thu, 01 Oct 2026 21:27:43 +0000</pubDate><link>https://news.ycombinator.com/item?id=49927275</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49927275</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49927275</guid></item><item><title><![CDATA[New comment by lukebuehler in "Pi Durable"]]></title><description><![CDATA[
<p>Yes, from your release post I can tell that you put a lot of thought into it!<p>I like your structured concurrency approach with tasks, which is similar to how I do it in Lightspeed too.<p>Also, the durable state implementation as documents is elegant! Question, though: why directly write/read to the store, why not abstract it and do more of a reducer/redux pattern and hide the persistence of the documents?</p>
]]></description><pubDate>Thu, 01 Oct 2026 21:08:43 +0000</pubDate><link>https://news.ycombinator.com/item?id=49927113</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49927113</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49927113</guid></item><item><title><![CDATA[New comment by lukebuehler in "Pi Durable"]]></title><description><![CDATA[
<p>Very cool to see Pi build a durable agent harness too. I've been building in this space for quite some time myself [0][1] and it is a super interesting place of innovation. Less hype-y that on-your-machine coding agents, but all major players are building products in this space: LangChain Deep Agents, Vercel Eve, OpenAI Agents API, Anthropic Managed Agents, etc.<p>The main reasons are:<p>1) they are "durable", i.e. easier to make long-running in an unattended way, and easier to implement recovery, monitoring, etc<p>2) separating the harness from the compute brings safety and scaling benefits<p>3) easier to make multi-player.<p>[0] <a href="https://github.com/smartcomputer-ai/lightspeed" rel="nofollow">https://github.com/smartcomputer-ai/lightspeed</a><p>[1] <a href="https://github.com/smartcomputer-ai/agent-os/" rel="nofollow">https://github.com/smartcomputer-ai/agent-os/</a></p>
]]></description><pubDate>Thu, 01 Oct 2026 20:57:12 +0000</pubDate><link>https://news.ycombinator.com/item?id=49927001</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49927001</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49927001</guid></item><item><title><![CDATA[New comment by lukebuehler in "Dots: Always-on agents"]]></title><description><![CDATA[
<p>yeah, it's not 100% clear, but notice nothing you quoted indicates that the dot harness itself runs inside said workspace. If you look at where OAI agent architecture has been going, they increasingly separate the harness from the compute env. See, "separating harness from compute" articles, recent "managed agents" offering, and so on.<p>If I'm reading between the lines correctly, the core Dot agent loop does not run in the workspace, but outside it.</p>
]]></description><pubDate>Tue, 29 Sep 2026 18:45:48 +0000</pubDate><link>https://news.ycombinator.com/item?id=49898406</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49898406</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49898406</guid></item><item><title><![CDATA[New comment by lukebuehler in "Dots: Always-on agents"]]></title><description><![CDATA[
<p>Dots are not remote agents _in_ a sandbox. They use a sandboxes/environments, but they are, what is now called, "managed agents", meaning they run in a distributed harness and utilize environments when they need on.<p>At least that is what I can ascertain from this article: <a href="https://openai.com/index/how-we-build-safety-security-and-privacy-into-dots/" rel="nofollow">https://openai.com/index/how-we-build-safety-security-and-pr...</a>
(see first diagram when scrolling down)</p>
]]></description><pubDate>Tue, 29 Sep 2026 18:28:32 +0000</pubDate><link>https://news.ycombinator.com/item?id=49898143</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49898143</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49898143</guid></item><item><title><![CDATA[New comment by lukebuehler in "AI has no intent and no motivation"]]></title><description><![CDATA[
<p>There is a really good discussion on exactly this analogy in the Jonas paper I posted in a top-level comment. Jonas considers Torpedoes (starting on p. 182, section III). Unfortunately too long to post here. But he asks where to locate purpose and agency in these automated systems. So as with the article, the question is less in "can these things be dangerous" but more where does the agency and intent reside. The answer is of course they can be used as weapons, but the intent and motivation does not really reside in the mechanism (drone, torpedo, AI agent), but in the humans behind them. The fallacy is trying to attribute full autonomy (and hence responsibility).</p>
]]></description><pubDate>Thu, 24 Sep 2026 10:23:28 +0000</pubDate><link>https://news.ycombinator.com/item?id=49828597</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49828597</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49828597</guid></item><item><title><![CDATA[New comment by lukebuehler in "AI has no intent and no motivation"]]></title><description><![CDATA[
<p>I think the "human needs" point is understated here. It would be better to say agents do not really have a self-contained metabolic engine that is required to keep going. Which bubbles up as that what we interpret as "drive", "will", "agency"... basically the will to live, and being willing to do _a lot_ to live if push comes to shove. We can't really identify a mechanism of similar complexity and integration in agents or LLMs. I think the article correctly points into that direction.<p>But Hans Jonas has made this point much better than the article or me, in "Critique of Cybernetics" (1953). PDF: <a href="https://s3.amazonaws.com/arena-attachments/892605/f0747c7943bec99f4891969d4a808ecd.pdf" rel="nofollow">https://s3.amazonaws.com/arena-attachments/892605/f0747c7943...</a></p>
]]></description><pubDate>Thu, 24 Sep 2026 10:02:58 +0000</pubDate><link>https://news.ycombinator.com/item?id=49828470</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49828470</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49828470</guid></item><item><title><![CDATA[New comment by lukebuehler in "Stripe's Knowledge AI Platform"]]></title><description><![CDATA[
<p>:-) <a href="https://github.com/smartcomputer-ai/lightspeed/pull/100" rel="nofollow">https://github.com/smartcomputer-ai/lightspeed/pull/100</a></p>
]]></description><pubDate>Wed, 23 Sep 2026 18:41:40 +0000</pubDate><link>https://news.ycombinator.com/item?id=49820572</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49820572</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49820572</guid></item><item><title><![CDATA[New comment by lukebuehler in "Stripe's Knowledge AI Platform"]]></title><description><![CDATA[
<p>k, I just looked through it for where the latest is. First of all, the earliest version were fully written by me, and then lightly edited with AI. The current state are as follows: intro sentences written by me, "why lightspeed" section too. Quickstart AI written, Features mostly written by me, but AI keeps interfering there. Design was also written by me several times, it also used to be much longer, but I just noticed AI put one  a few paragraph of slop in there, I think that's the most egregious--shame on me for not caching it. So, if I look at the readme right now, I would say it's about 60% me. It can be better, I'll make the effort because it really bothers me if readmes are not (mostly) written by humans.</p>
]]></description><pubDate>Wed, 23 Sep 2026 18:20:35 +0000</pubDate><link>https://news.ycombinator.com/item?id=49820309</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49820309</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49820309</guid></item><item><title><![CDATA[New comment by lukebuehler in "Stripe's Knowledge AI Platform"]]></title><description><![CDATA[
<p>Thank you. Yes, I did. There might be some AI-isms in there, because agents just can’t help themselves to dump their arcanae in there.</p>
]]></description><pubDate>Wed, 23 Sep 2026 16:39:52 +0000</pubDate><link>https://news.ycombinator.com/item?id=49818765</link><dc:creator>lukebuehler</dc:creator><comments>https://news.ycombinator.com/item?id=49818765</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49818765</guid></item></channel></rss>