<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Nick Wilkens | CTO &amp; Hands-On Software Architect</title><link>https://wilkens.io/</link><description>Recent content on Nick Wilkens | CTO &amp; Hands-On Software Architect</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Wed, 11 Nov 2026 17:00:00 +0000</lastBuildDate><atom:link href="https://wilkens.io/index.xml" rel="self" type="application/rss+xml"/><item><title>Asking Questions Outside My Field</title><link>https://wilkens.io/blog/asking-questions-outside-my-field/</link><pubDate>Wed, 11 Nov 2026 17:00:00 +0000</pubDate><guid>https://wilkens.io/blog/asking-questions-outside-my-field/</guid><description>&lt;p>I build cloud infrastructure. I have never taken an astronomy course. This
week I published
&lt;a href="https://wilkens.io/research/nea-pair-2018-nc15-2026-ld1/">a candidate asteroid pair&lt;/a> that I
could not find in the scientific literature. This post is about how that
happened, because I think the how matters more than the asteroid.&lt;/p>
&lt;h2 id="the-question-was-vague">The question was vague&lt;/h2>
&lt;p>I started with one sentence: NASA publishes a lot of data, so pick a dataset
and look for something nobody has found. That is not a research question. It
is curiosity, and until recently curiosity in a field you do not know went
nowhere. You would need months to learn which datasets exist, what the
standard methods are, and what has already been done.&lt;/p></description></item><item><title>A Candidate Asteroid Pair: 2018 NC15 and 2026 LD1</title><link>https://wilkens.io/research/nea-pair-2018-nc15-2026-ld1/</link><pubDate>Wed, 11 Nov 2026 12:00:00 +0000</pubDate><guid>https://wilkens.io/research/nea-pair-2018-nc15-2026-ld1/</guid><description>&lt;p>Two asteroids are circling the Sun together. They follow the same path, they
are about as far apart as the Earth and the Moon, and one is about half the
size of the other. The simplest explanation is that they used to be one rock.
I could not find any record of anyone pointing this out before.&lt;/p>
&lt;div class="explainer">
&lt;div class="explainer-row">
&lt;svg viewBox="0 0 200 130" role="img" aria-label="Diagram of the Sun, Earth's orbit, and a larger orbit with two dots close together on it.">
&lt;circle cx="92" cy="68" r="34" fill="none" stroke="#a6a6a6" stroke-width="1"/>
&lt;ellipse cx="108" cy="62" rx="56" ry="52" fill="none" stroke="#3987E5" stroke-width="1.5" opacity="0.8"/>
&lt;circle cx="92" cy="68" r="5" fill="#e8c547"/>
&lt;circle cx="126" cy="68" r="3.5" fill="#fafafa"/>
&lt;circle cx="150" cy="96" r="4" fill="#3987E5"/>&lt;circle cx="143" cy="102" r="3" fill="#EF5F2E"/>
&lt;text x="92" y="84" text-anchor="middle">Sun&lt;/text>&lt;text x="126" y="60" text-anchor="middle">Earth&lt;/text>&lt;text class="ink" x="160" y="118" text-anchor="middle">the pair&lt;/text>
&lt;/svg>
&lt;div>&lt;h3>Two asteroids, one path&lt;/h3>&lt;p>2018 NC15 and 2026 LD1 follow the same orbit around the Sun, a little outside Earth's. One lap takes two years. They never come closer to Earth than about 18 million km.&lt;/p></description></item><item><title>Hot-Patching in the Agentic CVE Age</title><link>https://wilkens.io/blog/hot-patching-in-the-agentic-cve-age/</link><pubDate>Sat, 31 Oct 2026 15:17:11 +0000</pubDate><guid>https://wilkens.io/blog/hot-patching-in-the-agentic-cve-age/</guid><description>&lt;p>The window between a kernel CVE going public and a working exploit can be
measured in hours, with AI making exploit development faster. Vendor patch
cadence does not always keep up. Supported kernels may get a fix quickly,
especially after coordinated disclosure. Kernels a few releases behind may
wait longer or get nothing, and the
machines running them are exactly the ones nobody wanted to reboot in the
first place.&lt;/p></description></item><item><title>Context Engineering: The Skill That Actually Matters for LLM Applications</title><link>https://wilkens.io/blog/context-engineering-llms/</link><pubDate>Sun, 01 Jun 2025 00:00:00 +0000</pubDate><guid>https://wilkens.io/blog/context-engineering-llms/</guid><description>&lt;p>Everyone talks about prompt engineering. The real skill for building useful LLM applications is context engineering.&lt;/p>
&lt;h2 id="whats-the-difference">What&amp;rsquo;s the Difference?&lt;/h2>
&lt;p>&lt;strong>Prompt engineering&lt;/strong> is about how you phrase requests to the model. Use chain-of-thought. Be specific. Give examples. Assign a role. These things matter, but they&amp;rsquo;re table stakes.&lt;/p>
&lt;p>&lt;strong>Context engineering&lt;/strong> is about what information you give the model to work with. What does it need to know to do the task well? How do you get that information to it efficiently? How do you manage context windows across a long conversation or multi-step task?&lt;/p></description></item><item><title>Building an AI-Native Engineering Platform</title><link>https://wilkens.io/blog/building-ai-native-engineering-platform/</link><pubDate>Sat, 15 Mar 2025 00:00:00 +0000</pubDate><guid>https://wilkens.io/blog/building-ai-native-engineering-platform/</guid><description>&lt;p>When I joined Edgecast as CTO, one of my first priorities was transforming how engineering worked with AI. Not adding an AI feature to the product — rethinking how the engineering organization itself operates.&lt;/p>
&lt;h2 id="the-problem-with-most-ai-tooling-rollouts">The Problem With Most AI Tooling Rollouts&lt;/h2>
&lt;p>Most enterprise AI adoption looks like this: someone buys a Copilot license, rolls it out, declares victory, and moves on. Engineers use it for autocomplete, maybe some basic generation, and that&amp;rsquo;s the end of it. The ROI is real but modest.&lt;/p></description></item><item><title>Acquiring Triton DataCenter from Samsung</title><link>https://wilkens.io/blog/acquiring-triton-datacenter-from-samsung/</link><pubDate>Tue, 20 Aug 2024 00:00:00 +0000</pubDate><guid>https://wilkens.io/blog/acquiring-triton-datacenter-from-samsung/</guid><description>&lt;p>In May 2022, MNX.io completed the acquisition of the Triton DataCenter and SmartOS platforms from Samsung. This is the story of how that happened and what we learned.&lt;/p>
&lt;h2 id="background">Background&lt;/h2>
&lt;p>Joyent built Triton DataCenter — one of the most technically sophisticated cloud platforms ever built. Native container support before Docker existed. Hardware virtualization (bhyve, KVM) on illumos. Distributed object storage (Manta). Software-defined networking. The engineering was exceptional.&lt;/p>
&lt;p>Samsung acquired Joyent in 2016 for approximately $170M. By 2022, Samsung&amp;rsquo;s priorities had shifted and they were looking to transition stewardship of the open-source platforms.&lt;/p></description></item><item><title>Resume</title><link>https://wilkens.io/resume/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://wilkens.io/resume/</guid><description/></item></channel></rss>