<?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 Sebastian Schuon on Medium]]></title>
        <description><![CDATA[Stories by Sebastian Schuon on Medium]]></description>
        <link>https://medium.com/@sebastian.schuon?source=rss-a8d6117e36a9------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*lz6T8eBDQ4ueuOeRybsSWw.jpeg</url>
            <title>Stories by Sebastian Schuon on Medium</title>
            <link>https://medium.com/@sebastian.schuon?source=rss-a8d6117e36a9------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 05:35:39 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@sebastian.schuon/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[When an engineer becomes product manager…]]></title>
            <link>https://medium.com/alasco-tech/when-an-engineer-becomes-product-manager-cf65536aace9?source=rss-a8d6117e36a9------2</link>
            <guid isPermaLink="false">https://medium.com/p/cf65536aace9</guid>
            <category><![CDATA[cloudformation]]></category>
            <category><![CDATA[circleci]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[continous-integration]]></category>
            <dc:creator><![CDATA[Sebastian Schuon]]></dc:creator>
            <pubDate>Thu, 14 Feb 2019 00:00:00 GMT</pubDate>
            <atom:updated>2019-03-07T15:39:10.608Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*P3UVDupKm_qbR6kQ.png" /></figure><p>In this post, I would like to talk about our infrastructure and how we deploy. I, Sebastian, had been coding the initial version of Alasco. As time went on and the team grew, I moved into the product manager role. Today I am responsible for the product, but more from a strategic point and less from a technical point. My transition was gradual, at first I stopped writing features and the last thing I worked on was infrastructure — both because I enjoy it for the empowerment it gives and because it’s quite decoupled from the daily product development work.</p><p>An issue I often have as product manager is “can you give feedback / approve this feature?”. Obviously if I do it on an engineers computer, the engineer is blocked. Therefore I had a build environment checkout on my computer but as stuff got more complex I was often stuck for hours to catch up with my setup so the latest features would work. As I want to block my engineering team as little as possible, we needed a better solution! Here’s the story how each branch today gets its own infrastructure and can readily be accessed by <a href="https://some-nice-url.alasco.de./">https://some-nice-url.alasco.de.</a> Super-easy to test!</p><h3>Our initial CI pipeline</h3><p>Alasco is built in the backend with a python / django application and react in the frontend. We already had a quite good CI pipeline in place before we ventured on this quest. In a nutshell, whenever changes get pushed to GitHub, a CircleCI build is triggered. Our CI pipeline builds a docker container that contains both backend and frontend code and run tests on it. Once good, it’s pushed into a docker registry. We host our application on AWS ElasticBeanstalk, where we deploy it as a docker container. Beanstalk gives us good additional tooling (like rolling updates, monitoring, logging…) and we feel with using a docker container we are independent enough to switch later to more sophisticated setups. In the end, the process from committing code to having it live is quite straightforward. We take pride that new team members commit and deploy to production on their first day in the team! I still remember copying php files to ftp servers ten years ago, what a dramatic change!</p><h3>An idea is born: never checkout a branch again to demo it</h3><p>Being a young startup, we did not provision our infrastructure using code in the first days. As we now have a significant customer base on our platform, we wanted to change this for enhanced reliability and reproducibility. Here the idea was born: why don’t we spin up a dedicated setup for each branch pushed? Obviously this would be something that should be easily doable once all our infrastructure was in code. Here’s how we did that:</p><p>As we host on AWS, we looked at different options to describe our infrastructure: <a href="https://aws.amazon.com/cloudformation/">CloudFormation</a>, <a href="https://www.terraform.io/">Terraform</a> or <a href="https://github.com/cloudtools/troposphere">troposhere</a>. We ended up using troposhere, as pure CloudFormation json files seemed tedious to maintain, terraform had a complex syntax and we didn’t need multi-cloud support. Troposhere didn’t require us writing pure json files and had the benefit of being in python too (thus could well integrate into our existing infrastructure).</p><h3>AWS Beanstalk as the nucleus of a deployment</h3><p>ElasticBeanstalk has the concept of an application and an environment. Think of an application as a group of environments. Between those you can, for example, share deployed versions of your software. We decided to have two dedicated applications, one for production and one for staging. This is an additional safe-guard that separates production from staging. Then all the feature-branches get deployed as an environment in the staging app. This looks like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*iaCgrpxMRqMIlxY4.png" /></figure><p>From a technical standpoint we decided to have two separate CloudFormation templates: one that spins up the application (that’s rarely executed) and one to spin up an environment within an application. We started our work based on Caktus’s <a href="https://github.com/caktus/aws-web-stacks">AWS Webstack</a> but soon deviated so heavily from it that it was not forkable anymore. For example we reworked it to have most parameters not being CloudFormation Parameters in the template but that we set them at template generation time in code. This gave us a better feeling of control as well as reducing complexity at stack creation time. Also we could then create inheritance of settings and selectively overwriting parameters. This ensures environments are as similar as possible and differences are clearly visible.</p><h3>CicleCI to drive deploys</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*rT3hC4vgm64B_55-.png" /></figure><p>We tweaked the deploy step in CircleCI to generate the CloudFormation templates, afterwards the template is executed and this brings up the stack. The stack provisions not only our application on Beanstalks, but also the RDS databases, S3 buckets and alike. We also create a DNS entry so the application can be reached by the branch’s name (that’s what all the work has been for!):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1014/0*JhlYSRCFiFkNTVxj.png" /></figure><h3>Saving some $$$</h3><p>Over time we upgraded out setup to install test data on newly created stack as to make testing even easier. We also applied some tweaks to save costs. First, we modified the branch deploys to use cheaper hardware (think smaller instances) as well as reduced redundancy (e.g. only one instance behind a load balancer). The latter also speed up our deployment times, as we didn’t needed rolling updates on multiple instances. The next improvement was to introduce skipped provisioning on branches that ended on -no-infra. Some things just don’t need an infrastructure to demo (like a modified source code formatter).</p><p>The final step was to introduce a cleanup script, that removes infrastructure once it’s not needed anymore. We decided to remove infrastructure once the respective branch is deleted. We therefore delete branches once we merged the pull requests.</p><p>For this we created a script, that loops over the existing branches and compares them to the provisioned infrastructure (we tag the stacks with the corresponding branch). Upon deleting, it’s important to not only delete the stack itself but also handle S3 buckets. Unfortunately S3 buckets which are non-empty cannot be deleted. Therefore our script empties buckets before triggering the stack delete. Here’s our script for your convenience:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/3627228aadd26290823ada229de804c3/href">https://medium.com/media/3627228aadd26290823ada229de804c3/href</a></iframe><p>As this script needs both Github access as well as AWS credentials, we decided to execute it as a scheduled build on CircleCI (currently on a hourly schedule). So we can provide credentials via environment variables in a secure way. This then looks like this in the CircleCI config:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/11cb036d67b764b69bd40c0142fee356/href">https://medium.com/media/11cb036d67b764b69bd40c0142fee356/href</a></iframe><h3>Today, PO demos are easy as pie</h3><p>While taking some days to get it running, this setup has greatly reduced the time to checkout and discuss new features. It’s also a major force in user testing, as new features are available with minimum delay to our UX researcher to validate them with our users.</p><p>Interested in having this setup too? Reach out to us if you have any questions or do join us: we’re looking for great team members, check out our open positions at <a href="https://www.alasco.de/en/">https://www.alasco.de/en/</a>!</p><p><em>Originally published at </em><a href="https://alasco-tech.github.io/2019/02/14/engineer-to-pm.html"><em>alasco-tech.github.io</em></a><em> on February 14, 2019.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cf65536aace9" width="1" height="1" alt=""><hr><p><a href="https://medium.com/alasco-tech/when-an-engineer-becomes-product-manager-cf65536aace9">When an engineer becomes product manager…</a> was originally published in <a href="https://medium.com/alasco-tech">Alasco Tech</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Where AI can be used and where not]]></title>
            <link>https://medium.com/@sebastian.schuon/where-ai-can-be-used-and-where-not-706c3991e153?source=rss-a8d6117e36a9------2</link>
            <guid isPermaLink="false">https://medium.com/p/706c3991e153</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[venture-building]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[business-models]]></category>
            <dc:creator><![CDATA[Sebastian Schuon]]></dc:creator>
            <pubDate>Sun, 09 Jul 2017 13:49:11 GMT</pubDate>
            <atom:updated>2017-07-09T13:49:11.397Z</atom:updated>
            <content:encoded><![CDATA[<blockquote><em>“Not conscious machines, omnipotent machines or even reasoning machines (yet), but statistical machines that automate and increasingly can outperform humans at certain pattern-recognition tasks. Computer vision, language understanding, anomaly detection and other fields have made immense advances in the past few years.” — </em>Derrick Harris <a href="https://www.wired.com/insights/2015/02/the-upside-of-artificial-intelligence-development/"><em>(Source)</em></a></blockquote><h4><strong>Types of AI</strong></h4><ul><li><strong>Supervised Learning:</strong> Given a large set of inputs and corresponding outputs learn to come up with outputs for any set of inputs</li><li><strong>Reinforcement Learning:</strong> Learn to control a task by executing it over and over in an environment and getting better at it each time (like a baby learns to walk)</li><li><strong>Unsupervised Learning:</strong> try to make sense of unlabeled data — sort of find an explanation or pattern to something observed. The issue here is that the outcome can’t be evaluated</li><li><strong>Expert Systems:</strong> Based on interviews of human experts be able to apply and combine that knowledge to reason better than human experts</li></ul><h4><strong>What is AI good at:</strong></h4><ul><li><strong>Optimize:</strong> solve tasks more optimal than humans do, especially if many inputs need to be considered</li><li><strong>Automize:</strong> Execute tasks that take little effort for a human (rule of thumb: takes human 1 second to do), but do it on a massive, parallel scale</li><li><strong>Predict:</strong> Be able to extrapolate the near term future based on historic data and other signals</li><li><strong>Reason:</strong> Given a lot of constraints and relations find the optimal solution</li><li><strong>Search:</strong> not finding a solution by reasoning about it but being able to search vast amounts of data and find a solution that someone had found before / existed long before</li></ul><p><strong>Where AI is not good at:</strong></p><ul><li>Work on problems where only few samples exist</li><li>Problems where it is hard or impossible to formulate a reasonable cost function</li><li>Systems that need to generalize to a wider range of tasks</li><li>Systems where memory and computing power are a constraint</li></ul><p>This is joint work with <a href="https://medium.com/u/b1775a972e70">Sebastian Schaal</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=706c3991e153" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[An approach to start my next company]]></title>
            <link>https://medium.com/@sebastian.schuon/an-approach-to-start-my-next-company-54afaae01100?source=rss-a8d6117e36a9------2</link>
            <guid isPermaLink="false">https://medium.com/p/54afaae01100</guid>
            <category><![CDATA[ideation]]></category>
            <category><![CDATA[ideas]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <dc:creator><![CDATA[Sebastian Schuon]]></dc:creator>
            <pubDate>Sun, 21 May 2017 16:05:13 GMT</pubDate>
            <atom:updated>2017-05-21T16:05:13.570Z</atom:updated>
            <content:encoded><![CDATA[<p>I recently sold my company. What’s next? Being an entrepreneur I’m set on starting a new venture but the next steps aren’t very clear to me. Here is the result of a month of research how one or specifically me will go about. I’d be curious to hear if anyone as followed a similar approach or can share their experiences in finding your new calling</p><h3>Phase I: Vision</h3><p>I try to determine what I want to achieve with my life. What is important for me in the long-run. I will dedicate a considerable time of my life on the next venture, thus I should ensure it pays into the vision of my life.</p><p>Helpful to me was reflection on the following scenario: “Imagine you are a grandfather and one of your grandchildren asks you: what did you as a job? And why did you work?” Especially the later question helped me to decide what’s my vision and why I want to pursue it.</p><h3>Phase II: Boundary Conditions</h3><p>The amazing part of being able to freely choose what’s next is that there are so many opportunities. This brings along the downside that I won’t be able to evaluate them, so I need a heuristic to keep the process manageable.</p><p>I employ the following big picture strategy: I first determine boundary conditions with a mixture of a white &amp; black list to be able to judge within seconds if an opportunity is worth further consideration. Then then plan is to drill down on the remaining opportunities, learn about them and evaluate if there is a possible business opportunity. Here are the categories I made criteria lists for:</p><ol><li>My goals for this venture and what I’m willing to give to achieve them</li><li>My strengths and weaknesses</li><li>What knowledge / assets in general do I have that I can leverage</li><li>Whom I want to found the company with</li><li>Fields that interest me</li><li>Business models that I favour</li></ol><h3>Phase III: Research</h3><p>Based on the list of fields that sparked my interest I started to learn more about these very fields. The goal is to drill down per field and eventually settle on one combination of user need + business model + proposed solution that can be tested in the next phase as a prototype.</p><p>Per field of interest I keep a document that I continuously append and update. It contains the following sections</p><ul><li>Definitions in the field used</li><li>What visions do people have in the field</li><li>Who’s an expert in the field</li><li>What are current challenges</li><li>What are interesting / hot companies &amp; projects in the space</li><li>Which user needs did I come across</li><li>What did people try and failed in</li><li>What business models seem to flourish</li></ul><p>Then I created an initial list of people known in the field (rather randomly). I make a point to read most of their published work and will feed all learnings in the document above. Next I try to interview them asking questions on the categories above. At the end of each interview I ask whom they consider leading in their field and add these recommendations to the list of people I shall interview. <strong>Tim Ferris Question Catalogue</strong></p><p>Given the myriad of options staying focused is a challenge. I try to force myself into that by creating a sounding board. It’s sort of an accountability group where I ask a small set of people who’s opinion I value and invite them together to dinner some date in the future. I set a milestone in my process I want to have attained until that date and promise them to show them the results of that milestone. Not only creates this the pressure to get it done, but also I gather most valuable feedback on my next steps from the group.</p><h3>Phase IV: Experimentation</h3><p>An important question is: when do I have found a business opportunity that’s could potentially be the <em>one</em>. As of now, I’m looking to two criteria:</p><ul><li>Diving into it feels right, I start working on it until late night and the flow experience kicks in</li><li>It matches my criteria from phase II</li></ul><p>Then it’s time to verify. I plan on building then a quick prototype to test it on real users. This should yield me again quite some insights, based upon which I then can go forward and start a business on this idea or go back to the previous phase and update my research with the new insights.</p><h3>Phase V: Implementation</h3><p>Once the business is started, I want to focus on using the levers I defined in phase II. Also after some time into running the company, I plan to circle back to this process and see if with the current company the set out goals are achievable or if I need to tweak something</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=54afaae01100" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Count your Inbox ToDos — Chrome Extension]]></title>
            <link>https://medium.com/small-business-forum/count-your-inbox-todos-chrome-extension-4c418e684fbf?source=rss-a8d6117e36a9------2</link>
            <guid isPermaLink="false">https://medium.com/p/4c418e684fbf</guid>
            <category><![CDATA[goals]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[chrome-extension]]></category>
            <category><![CDATA[chrome]]></category>
            <category><![CDATA[okr]]></category>
            <dc:creator><![CDATA[Sebastian Schuon]]></dc:creator>
            <pubDate>Tue, 07 Mar 2017 17:07:35 GMT</pubDate>
            <atom:updated>2017-03-07T17:07:35.524Z</atom:updated>
            <cc:license>http://creativecommons.org/licenses/by/4.0/</cc:license>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*e6_zwWWNMZqAPPQkoYWj2w.png" /></figure><p>Ever since the beta I’m a fan of <a href="http://inbox.google.com/">Google Inbox</a>. I use Inbox to manage all my tasks, as most of them come in an email anyways. Furthermore I’m a big fan of goal setting and as such set myself goals to manage my task list down to a certain number (ideally aiming for the sunshine of an empty Inbox).</p><p>I got tired of counting the number of items in my Inbox, thus I created an Chrome extension that displays the number of todos right behind the inbox label.</p><p>I then use this number to track it in my personal <a href="https://en.wikipedia.org/wiki/OKR">OKRs</a> each morning:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Y6X3k6Wpc5jJCgCw95YeCA.png" /></figure><p>The extension is available in the <a href="https://chrome.google.com/webstore/detail/inbox-todo-count/ceaelmbokkdabmijonojglkmkebimhfk">Chrome Web Store</a> as well as the source is on <a href="https://github.com/schuon/inbox-item-count">github</a>.</p><p>I’m happy for feedback and bug reports, send me a <a href="http://twitter.com/schuon">tweet</a> or <a href="/service/mailto:sebastian.schuon@gmail.com">mail me</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4c418e684fbf" width="1" height="1" alt=""><hr><p><a href="https://medium.com/small-business-forum/count-your-inbox-todos-chrome-extension-4c418e684fbf">Count your Inbox ToDos — Chrome Extension</a> was originally published in <a href="https://medium.com/small-business-forum">Small Business Forum</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>