<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Nathan on Medium]]></title>
        <description><![CDATA[Stories by Nathan on Medium]]></description>
        <link>https://medium.com/@nathan_14268?source=rss-ed7ec72ee4ed------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*CLiRI-DTfZnXRZqP</url>
            <title>Stories by Nathan on Medium</title>
            <link>https://medium.com/@nathan_14268?source=rss-ed7ec72ee4ed------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 12:39:00 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@nathan_14268/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[If your government customer says, “Don’t give me another spinning globe,” this is what they mean.]]></title>
            <link>https://medium.com/@nathan_14268/if-your-government-customer-says-dont-give-me-another-spinning-globe-this-is-what-they-mean-efc0115eab5b?source=rss-ed7ec72ee4ed------2</link>
            <guid isPermaLink="false">https://medium.com/p/efc0115eab5b</guid>
            <category><![CDATA[united-states-space-force]]></category>
            <category><![CDATA[space]]></category>
            <dc:creator><![CDATA[Nathan]]></dc:creator>
            <pubDate>Tue, 14 Jul 2026 21:25:25 GMT</pubDate>
            <atom:updated>2026-07-14T21:44:09.838Z</atom:updated>
            <content:encoded><![CDATA[<p>Almost every day I log into LinkedIn and see some variation of the same post:</p><blockquote>“Excited to announce my new space platform, which visualizes close approaches, conjunctions, launches, debris, &lt;insert-space-problem-here&gt;.”</blockquote><p>A 5 minute scroll down my LinkedIn looks something like this</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yG4p98mgvT13FMR82i3Y7w.png" /></figure><p>And honestly, I love it. As a fellow tinkerer, builder, and someone who has spent years developing these kinds of systems, I genuinely enjoy seeing more people enter the space domain. AI-assisted development has made it possible to move from an idea to a polished 3D visualization faster than ever before.</p><p>But many of these projects are <strong>solving the same first problem</strong>: putting objects on a globe. Yes, that is a valuable starting point. It is just not where the difficult work begins.</p><p>So, rather than criticizing another generation of spinning globes, I thought I would put together a few lessons that may help early-stage founders, developers, and startups move beyond the visualization and toward something an operator might actually use.</p><h3>A little about me.</h3><p>I began my space domain awareness journey in 2019, before the United States Space Force had formally stood up. At the time, many space capabilities, datasets, and operational concepts still lived largely within classified environments (and on whiteboards and spreadsheets).</p><p>There was, however, a concerted effort by people across the community to declassify appropriate capabilities, open more data to commercial industry, and create pathways for companies to contribute meaningfully to national security space missions and being located in Colorado Springs, I was at the forefront of many of these discussions and meetings, behind the curtain if you will.</p><p>That shift created an opportunity for commercial organizations to come in and innovate in areas where the Air Force and, later, the Space Force had previously spent billions of dollars on large, slow-moving waterfall programs (CAVEnet, JMS etc) with mixed success.</p><p>My own journey began in this world of whiteboards, spreadsheets &amp; prototypes, and assumptions about what operators might need. It eventually led to the <a href="https://sbms.saberastro.com/"><strong>Space Battle Management System (SBMS)</strong></a><strong> </strong>as it exists today, A system that is adopted and used by over 3,000 Operators and has become one of the standards for Space Domain Awareness Analysis &amp; visualization.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hv9CzmvrdcA3DLtJutds5w.jpeg" /><figcaption><a href="https://www.vandenberg.spaceforce.mil/News/Article-Display/Article/3918814/delta-8-enhances-situational-awareness-through-space-cockpit-cop-powered-by-jco/">Delta 8 enhances situational awareness through Space Cockpit</a></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ik5HO0S6sifgg5N5A80_ig.png" /><figcaption><a href="https://sbms.saberastro.com/">https://sbms.saberastro.com</a></figcaption></figure><h3><strong>Operations FIRST, Developer Second.</strong></h3><p><strong><em>You will learn more from one six hour live exercise with actual operators, than you will from six years of sitting in a room imagining what they might want.</em></strong></p><p>You need to close the gap between your development lifecyle and operational experience. Operators will use your system in ways you did not expect. They will ignore features you thought were essential. They will perform mental calculations you did not realize they were doing. They will combine data sources in ways your architecture did not anticipate. They will immediately expose which parts of your interface support a real decision and which parts simply look good in a demonstration.</p><p>That feedback loop is where operational software is really built. The spinning globe is the starting line, not the finish line.</p><p>The most important step for any early-stage space startup is not adding another visualization layer. It is getting the software in front of real operators, during a real exercise, against a real mission problem.</p><p>Once the visualization has to represent the real operational environment, an entirely different set of problems start to appear. In part this is the beauty of AI vibe coding, a lot of these operators are now able to quickly prototype and develop their own set of tools and widgets.</p><h3><strong>Moving beyond Celestrak</strong></h3><p>In SBMS, we ingest roughly <strong>two million records</strong> per day from <strong>multiple </strong>sensor and data providers, often describing slightly different versions of the same reality. That immediately creates difficult questions:</p><ul><li>Which source should be shown and when?</li><li>Which provider is strongest for a particular mission, orbital regime, time-frame, or phenomenology?</li><li>How should conflicting data be normalized, standardized, enriched, and presented?</li><li>How does the operator know what they can trust?</li></ul><p>Moving beyond TLEs creates another major step change. State vectors and ephemerides can provide much higher fidelity, but they also introduce questions around propagation consistency, force models, covariance, epoch age, and how far a state can reasonably be propagated before confidence begins to degrade.</p><p>The operational perspective also changes by orbital regime.</p><p>For example,<strong> </strong>in <strong>LEO</strong>, plane matching can be more important than current proximity. Two spacecraft may be on opposite sides of the Earth, but if their RAAN and inclination are closely matched, the time and energy required to create a future interaction can be substantially reduced.</p><p>In <strong>GEO</strong>, longitude, relative longitude drift rates, and time to drift around the belt often become more operationally meaningful.</p><p>MEO and highly elliptical regimes combine aspects of both. Cislunar operations are creating an entirely new set of visualization and reasoning challenges.</p><ul><li>Context matters. Is an object truly debris, or is it an operational payload being represented incorrectly? Who made this satellite, what are its capabilities? Do they match with what hte SDA picture is telling us?</li><li>Is the orbit state based on a TLE, a state vector, or a fitted solution?</li><li>What propagation model was used?</li><li>Does covariance exist?</li><li>How old is the underlying data?</li><li>How far are we propagating that state from it’s epoch?</li></ul><p>An operator should be able to understand all these things and more at a glance.</p><p>It is also not just about orbit states. Electro-optical, radar, passive RF, active RF, and other metric observations each tell part of the story. The challenge is combining them without overwhelming the user or creating a visually impressive but operationally confusing display.</p><p>One design principle I have tried to follow is simple:</p><h4>The user should never have to perform ANY math in their head.</h4><ul><li>Do not just show the Longitude drift rate of each satellite, calculate and prioritize the relative drift rate, especially if the user has one object selected and hovers over a second satellite etc.</li><li>Do not only show two timestamps, show the time offset between them (<strong>1h 14m 23s ago</strong> etc)</li><li>Do not only show the age of a state, show its age relative to both simulation time (if propagating) and the operator’s current time.</li><li>Do not force users to mentally convert between longitude conventions (-180/180, 0–360 etc) when the software can present both.</li></ul><p>Every calculation the interface can perform reliably is one less source of operator delay or error. And there are literally 20–30 of these rules of thumb I can think of.</p><h3>Oh man, this is both a BIG DATA and BIG COMPUTE problem (both server and client side).</h3><p>A modern catalog may include approximately 100,000 payloads and hundreds of thousands of additional tracked objects. Each can have repeated state updates and potentially thousands of observations across multiple phenomenology’s EVERY DAY.</p><p>Now propagate those objects, maintain histories, calculate conjunctions, and allow users to query the data without accidentally requesting two years of observations and overwhelming the system. Plotting 100K dots spinning around the globe is fine, but how do you render an actual orbit trajectory for all those objects and make that look smooth when propagating at 100x real-time, do you need to? Now throw in ground-sites, and calculate groundsite to satellite overpasses and things blow out of control rapidly.</p><p>That is where the problem becomes big data, plus big compute, not simply astrodynamics and 3D rendering.</p><h3>Astrodynamic Library Selection</h3><p>Propagating a TLE with SGP4 is the starting point, not the end state. It is accessible, computationally efficient, and often good enough for applications such as catalog visualization, general situational awareness, and coarse sensor tasking. But TLEs are compressed mean-element products designed for use with SGP4, not precision state estimates with well-characterized uncertainty. They are therefore insufficient for many higher-consequence SDA activities, including detailed rendezvous and proximity operations analysis, maneuver characterization, precision conjunction assessment, and other problems where small errors in position, velocity, timing, or reference-frame handling materially affect the answer.</p><p>Below is a list of some of the state representations propagated using different methods and their relative accuracy. If you are using TLE’s from Space-Track or Celestrak your accuracy representation you’re giving your users is in that 1 star band.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cHpETRhzCFnElNl-YrhNJA.png" /></figure><p>This may be OK depending on what you want to do (ground-site overpass calculations etc), but for things like conjunction assesment you need to be using one of the more accurate methods.</p><p>Fortunately, builders do not need to create the astrodynamics stack from scratch (Like I did back in 2019). The community already has tools such as the <a href="https://celestrak.org/software/tutorials/sgp4.php">Kelso/Vallado SGP4</a> implementations, USSF Astro Standards (if you’re on Government Contract), <a href="https://www.orekit.org/">Orekit</a>, <a href="https://etd.gsfc.nasa.gov/capabilities/capabilities-listing/general-mission-analysis-tool-gmat/">GMAT</a>, <a href="https://github.com/david-rc-dayton/pious_squid">Pious Squid</a>, and <a href="https://www.npmjs.com/package/satellite.js/v/1.3.0">Satellite.js</a>. The important question is not simply which library you use, but whether the propagation method is appropriate for the source data and mission. State vectors, ephemerides, covariance, reference frames, time systems, gravity models, atmospheric models, solar activity, and force-model assumptions all matter.</p><p>Two applications can begin with what appears to be the same spacecraft state and produce different trajectories because they made different choices about propagation, interpolation, Earth orientation, drag, or numerical integration. Those assumptions, and the confidence of the resulting solution, should be visible to the operator rather than hidden behind a smooth orbit line LERPING between propagated positions.</p><h3>Efforts to standardize</h3><p>There are also important efforts underway to standardize both how space data is exchanged and how operators interact with it. You (or your AI) should be aware of these efforts and incorporate them or their learnings. On the user-experience side, <a href="https://www.astrouxds.com/"><strong>Astro UXDS</strong></a> provides shared design guidance, components, interaction patterns, and best practices specifically for space applications, helping reduce the need for every program to invent its own terminology, controls, and visual language.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*amkK_BcRBNC_R4GB5QL03Q.png" /></figure><p>On the data side, the <a href="https://unifieddatalibrary.com/"><strong>Unified Data Library</strong></a> provides a centralized environment for ingesting and distributing hundreds of different space data types from government and commercial sensors.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*SAt9BHPLjpj52B3C" /></figure><p>while open-source efforts such as <strong>TJ Koury’s </strong><a href="https://spacedatastandards.org/"><strong>Space Data Standards</strong></a> are developing modern, high-performance schemas for exchanging space situational awareness and space traffic management data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Tc_Ljdwtp-ccpIzW7sNN5w.png" /></figure><p>These initiatives matter because operators should not have to relearn the basics every time they open a new tool, and developers should not have to build a new data model every time they integrate another provider.</p><h3>Deployment</h3><p>You might have the greatest tool in the world, but if the operators can’t use it in their environment, it’s useless. If your customer is the USSF or Airforce, you will want to ensure that your software meets all the requirements for an Authority to Operate (ATO) so that it can be deployed onto their ops floors for use at different Impact Levels, Non Classified (NIPRnet) (IL2), SECRET (SIPRnet) (IL6) and TOP SECRET (JWICS) enclaves (the classifications do go higher depending on your customer).</p><p>This is traditionally a long and complex process, of meeting a long list of stringent cyber security requirements encapsulated here: DoD Enterprise <a href="https://dodcio.defense.gov/Portals/0/Documents/DoD%20Enterprise%20DevSecOps%20Reference%20Design%20v1.0_Public%20Release.pdf">DevSecOPs Reference Design</a>. There have been numerous efforts to streamline this process over the years such as the <a href="https://p1.dso.mil/">DoD Platform One</a> whose products (PARTYBUS, IRONBANK and BIGBANG) are a great accelerator and step in the right direction.</p><p>If I was to summarize my advice would be as follows:</p><ul><li>Use a containerized microservices approach (e.g deployed into AWS managed by Kubernetes stack) perferably BigBang or Defense Unicorns Big Bang.</li><li>Base your software off hardened container images available on <a href="https://ironbank.dso.mil/">IronBank</a> or <a href="https://images.chainguard.dev/">ChainGuard</a></li><li>Incorporate CI/CD Pipelines for all your microservices that include at least all of the stages of Partybus (Static code analysis, pen testing, unit testing, linting, dependency scans etc, etc, etc).</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DuFP1b91DPLlLgASkXyQOA.png" /></figure><p>Good luck!</p><h3>Closing</h3><p>So yes: congratulations to everyone building a globe, loading a catalog, and experimenting with AI-generated space interfaces. This is a great way to enter the problem space, communicate ideas, and rapidly test concepts.</p><p>Just recognize that beyond the first impressive visualization is a deep set of challenges involving astrodynamics, sensor fusion, data provenance, trust, human factors, storage, compute, and operational decision-making.</p><p>Many of us have spent years discovering problems we did not initially know existed. I am always open to constructive discussion, comparing lessons learned, and hearing where others agree or disagree. The more openly we can share the problems, even when we keep some of the implementation details proprietary, the faster the entire community will advance.</p><p>I hope this helps some of the builders entering this space move beyond the first impressive visualization and toward tools that genuinely support operators and mission decisions. None of us has all the answers, and many of these challenges are still evolving, especially as new sensors, data standards, orbital regimes, and AI-assisted workflows come online.</p><p>Good luck out there!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=efc0115eab5b" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>