<?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 Chris Drouin on Medium]]></title>
        <description><![CDATA[Stories by Chris Drouin on Medium]]></description>
        <link>https://medium.com/@chris.drouin_32477?source=rss-04d01de8eae4------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*mwWKYGaPzWMIh5kQ4XdorA.jpeg</url>
            <title>Stories by Chris Drouin on Medium</title>
            <link>https://medium.com/@chris.drouin_32477?source=rss-04d01de8eae4------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 01:26:15 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@chris.drouin_32477/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[Moving Faster without Breaking Things (with AI)]]></title>
            <link>https://medium.com/runkeeper-everyone-every-run/moving-faster-without-breaking-things-with-ai-71758204771f?source=rss-04d01de8eae4------2</link>
            <guid isPermaLink="false">https://medium.com/p/71758204771f</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[asics]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Chris Drouin]]></dc:creator>
            <pubDate>Mon, 24 Feb 2025 19:31:26 GMT</pubDate>
            <atom:updated>2025-02-24T19:31:26.394Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Two runners jog along a scenic road with a mountain in the background, wearing bright athletic gear and matching teal running shoes." src="https://cdn-images-1.medium.com/max/1024/1*KOBaoNaRyYq1Dtn9d8VnBw.jpeg" /></figure><p>The mantra of Silicon Valley has been, for decades, “Move fast and break things.”</p><p>To be honest, I’m not a huge fan of the second part! Repeatedly and publicly breaking things in search of progress eventually garners your company a certain reputation with customers. How many of you reading this loved one of the products in the <a href="https://killedbygoogle.com/">Google Graveyard</a> that was set aside after a shift in corporate goals? How many of you have woken up to a completely new interface on a website or app that you use daily that completely breaks your workflow?</p><p>ASICS’ philosophy, by contrast, has long been “kaizen” — continuous improvement. You can trace the evolution of our products over generations. We’re not afraid to make bigger changes <em>once they have been properly validated</em> — see how the design innovations of the GEL-KAYANO® Lite series got folded back into the mainline GEL-KAYANO® series once we confirmed the success of the light-weight design. (For the software engineers here, it’s a little bit like we made a branch and then merged the changes back into main after passing the QA tests).</p><p>All of this applies to the current boom in artificial intelligence.</p><p>Over the last two years, many companies have tried to disrupt the market (and the world) with generative AI powered products and services. Some of them are succeeding, but most are overpromising and underdelivering. Take, for instance, Cognition’s “Devin” software engineer agent. It was pitched as an automated full stack software engineer — an app that you could practically slot right into your org charts. Except: in practice, <a href="https://futurism.com/first-ai-software-engineer-devin-bungling-tasks">researchers have found that it’s slow and usually fails to produce correct results</a>.</p><p>As a data scientist working in the AI field, one thing that I fear is that my corner of the industry is going to burn through its supply of good will before the hype catches up with reality of what these tools are capable of. And that’s a shame, because AI tools can already be put to practical use in so many areas of our business.</p><p>For a concrete example of this: last year, I spent a fair bit of time writing a program that could analyze pictures of our shoes and tell us what colors are visible. You’d think this would be something we already know — we categorize all our shoes by color, right? — but I was prompted to act when I discovered that this shoe:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/624/1*7_ZJqgFubh4YceqYy7-Q3w.png" /><figcaption>A women’s GEL-EXCITE™ 9 shoe</figcaption></figure><p>… is classified as a “White/Black” colorway in our system. Whoops. Technically that’s correct, given the way ASICS specifies primary and secondary colors, but I think most of our customers would expect us to call out the hot pink midsole. A few hundred lines of Python later, I had a system in place that could slice up an image into pieces and tell me the dominant colors visible in each chunk. Progress!</p><p>This year, while experimenting with Meta’s LLaMa-3.2 multimodal model, I was able to replicate that work in about 30 minutes with a natural-language prompt. Feed the model a picture of a shoe, give it a few examples of the format of the output you’re looking for, and you are (mostly) off to the races. The results aren’t 100% perfect, but they are on par or better than what you’d get from sitting a human down and asking them to classify hundreds of shoe pictures.</p><p>Better color classification for our shoes is not going to revolutionize the world by itself. <em>Most</em> successful applications of generative AI are not going to revolutionize the world by themselves. But the sum of many <em>kaizen</em>-style iterative improvements can be revolutionary taken together. Better color classification can mean better search results when you search for “pink running shoes” on asics.com, better demand forecasting to inform which shoe colorways we should produce more of, or any of dozens of other small improvements that touch upon ASICS product metadata.</p><p>Maybe the big agentic models are going to change how we work over the coming years; in the meantime, there are plenty of places where simple applications of AI can reduce friction and make our lives easier.</p><p><em>What unexpected practical applications have you found for generative AI? Sound off in the comments!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=71758204771f" width="1" height="1" alt=""><hr><p><a href="https://medium.com/runkeeper-everyone-every-run/moving-faster-without-breaking-things-with-ai-71758204771f">Moving Faster without Breaking Things (with AI)</a> was originally published in <a href="https://medium.com/runkeeper-everyone-every-run">ASICS Digital</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>