<?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 Eric Baer on Medium]]></title>
        <description><![CDATA[Stories by Eric Baer on Medium]]></description>
        <link>https://medium.com/@baerbaerbaer?source=rss-4077cea51e29------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*G2f35Urn5DE3EXXgsaHOgQ.jpeg</url>
            <title>Stories by Eric Baer on Medium</title>
            <link>https://medium.com/@baerbaerbaer?source=rss-4077cea51e29------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 08:51:31 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@baerbaerbaer/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[Storage Is the Missing Piece for Agent-Built Software]]></title>
            <link>https://engineering.gusto.com/storage-is-the-missing-piece-for-agent-built-software-ede96ac7ac27?source=rss-4077cea51e29------2</link>
            <guid isPermaLink="false">https://medium.com/p/ede96ac7ac27</guid>
            <category><![CDATA[vibe-coding]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[database]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Eric Baer]]></dc:creator>
            <pubDate>Tue, 16 Jun 2026 15:59:25 GMT</pubDate>
            <atom:updated>2026-06-16T15:59:25.971Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Three glowing isometric icons representing cloud infrastructure component: a Compute cube with a lightning bolt, a Storage platform with a floating bucket, and an API Tokens cube, connected by dotted lines on a dark navy background." src="https://cdn-images-1.medium.com/max/1024/1*rtCnDMz0vQyXnpZdjLzlDA.png" /></figure><p>Gusto, like most companies building with LLMs, is amassing a <a href="https://x.com/kdaigle/status/2040164759836778878?lang=en">LOT of software</a>: one-off projects, Claude Artifacts, HTML visualizations, dashboards, and my own growing fleet of personal tools.</p><p>The agent loop made building 100 times easier, but deployment and operations still assume the old world. Your choices are weirdly binary:</p><ol><li>Static sites with data <em>in the HTML</em></li><li>A whole app with Dockerfile, Postgres, secrets, CVE mitigation, deploy pipeline, PagerDuty, and sharp edges.</li></ol><p>When software was hard, most things were #2. Now, most things sit in the middle: stuff that’s real enough to need a database, but not real enough to deserve an app stack.</p><p>Neither fits, and this isn’t about engineers vs. non-engineers. Everyone’s a builder now, and everyone from marketers to top engineering teams is <a href="https://openai.com/index/harness-engineering/">using the same tools</a>. The agent loop is producing heaps of legitimate software, and IMO, if your platform can’t survive this authoring mode, it’s the wrong thing for the moment we are in.</p><p>We need things that make “casual software” easy to create <em>and</em> easy to operate.</p><h3>Saying “yes” to builders, without compromising engineering</h3><p>There is an AI catechism: automate supply, summon infinite demand. And, as our platform and security teams’ request queues can attest, this seems to be true.</p><p>We use Claude heavily in all roles, both inside and outside our engineering org, and our platform help channels were filling up with folks asking essentially: “Where do I put this thing?” So we built a simple static-site hosting tool, and for a minute, everyone was happy. They went off analyzing customer engagement data, code quality metrics, and the 100 other things our MCP tools have access to, and would share this all over Slack.</p><p>Inevitably, some of these sites became load-bearing, which exposed the harder question: “If I need to update the data regularly, how can I store it?” For most of these projects, neither an HTML page nor “Congratulations, you own/operate a service” is a good answer.</p><p>There were two problems to solve:</p><ul><li>How do we make it easy for a non-engineer builder to own/operate?</li><li>How do we support this in our infrastructure?</li></ul><p><strong>On Supporting Non-Engineer Builders</strong></p><p>The agent loop works for Engineers because they quietly absorb its failures. But a workflow that depends on taste, scar tissue, and vigilance does not scale to <a href="https://github.blog/news-insights/octoverse/octoverse-a-new-developer-joins-github-every-second-as-ai-leads-typescript-to-1/">everyone else</a>.</p><p>When the operator may not be technical enough to understand a crash, tools should make it easy to claw their way out with minimal fanfare. The operational burden can’t ask them to look in Buildkite logs to understand that there is an issue with secrets rotation.</p><p><strong>Platform Problems</strong></p><p>Second, we should be seriously considering fleet economics. For “serious software,” a little extra machinery is fine. For a large number of agent-built internal tools, that machinery becomes really important. Every idle database and forgotten secret is operational overhead someone eventually pays.</p><p>Our platform teams are already busy supporting our load-bearing <a href="https://gusto.com/company-news/gusto-reaches-1b-in-revenue">apps</a>. They don’t need a hundred semi-serious tools that keep secrets, generate alerts, appear in service catalogs, and page people after the original author has moved on, if there even was a formal owner in the first place.</p><h3>Safety = Guardrails + AI-native Building Blocks</h3><p>There are two answers: Guardrails to reduce the blast radius, and building blocks that make the safe path the easy path.</p><p>Guardrails matter, and we’re building them: quarantined runtimes, safer agent execution, separate AWS accounts, review bots, and other blast-radius reducers. But guardrails only make dangerous paths less dangerous. They do not create a better default than “HTML blob” or “real service.”</p><p><strong>What about the building blocks?</strong></p><p>The standard engineering toolkit has a lot of sharp edges because it’s fundamentally built to handle really big workloads. For semi-serious apps, here’s my wish list for tools</p><ul><li><strong>Zero Idle Burden.</strong> It should not accumulate rent for existing, and regardless of who is operating it, fleets of small software can’t tolerate CVE mitigation, patching, kernel panics, inactivity emails, or a $5 monthly floor.</li><li><strong>Success has an exit.</strong> If the app becomes important, graduation should be mechanical, not a deep technical migration</li><li><strong>Agent-shaped API.</strong> The entire surface needs to be small enough to fit in context, typed enough to reject hallucinated methods, and low-ceremony enough that the model is not asked to invent and preserve a schema before it has earned one. This is not cosmetic.</li></ul><p>In 2026, you need three basic parts to build an app: compute, storage, and tokens. Compute and tokens both have answers that fit the bill. Compute has scale-to-zero primitives like FaaS that are <a href="https://arxiv.org/abs/2307.03958">hard to get wrong</a>. Tokens are just API calls.</p><p>But what about storage? I am a card-carrying member of the <a href="https://boringtechnology.club/">Boring Technology Club</a>, and Postgres is the <a href="https://www.amazingcto.com/postgres-for-everything/">right answer</a> to many questions, but it’s wrong here for two reasons: Operations and LLM Usability.</p><p>Real databases entail real obligations, and none of that becomes free because the app has four users. Provisioning, secrets, backups, VULNs, migrations, alarms when the disk fills, alarms when the pool is exhausted, etc.</p><p>Worse, “Real” tools invite the agent to generate a real-service ceremony: migrations, schemas, RLS, seed files, whatever. Complexity creep degrades the Agent’s loop speed and reliability as the app grows and can leave the operator unsure why.</p><h3>Gusto’s New AI Building Block</h3><p>We’ve started trying out something new. It’s called <a href="https://github.com/Gusto/baerly-storage">baerly-storage</a>, and we’ve open-sourced an early version of it today.</p><p>It’s an S3-backed datastore with <strong>no runtime</strong>. None. Everything runs to completion inside the HTTP request that triggered it, and the whole public API fits in a single markdown file small enough that an LLM can hold it in context.</p><p><strong>For Platform Teams: It’s just a bucket</strong></p><p>You know what is even more boring than Postgres? Object storage like S3. It’s the only modern infrastructure primitive that is pre-cleared by security, politically boring, and where every major cloud has a compatible endpoint. It’s safe and portable by definition. Your bytes in your bucket with your security.</p><p>And it turns out that with a little work, you can turn an S3-compatible bucket into a document database with a 100 KB TypeScript library. Yes, library. A database without a runtime. No daemon, no service bill, no on-call. The only persistent component is your bucket. The entire thing runs inside the HTTP request that triggered it.</p><p>The trick is not to make S3 behave like a traditional database. It is to use it like other modern storage systems: write immutable data, then publish a new version with a small, atomic metadata update. S3 now <a href="https://aws.amazon.com/blogs/aws/amazon-s3-update-strong-read-after-write-consistency/">gives us the consistency</a> and conditional writes we need for that final “publish” step, so the rest is mostly borrowing from a well-known storage playbook. Iceberg, Delta Lake, Turbopuffer, and Litestream all landed on this shape.</p><p>And you can export the change log as a Debezium-style JSON change event, so upgrading to Postgres is easy.</p><p><strong>For Builders: This is LLM-native storage</strong></p><p>Most APIs are built for humans, then adapted for LLMs. We wanted the opposite: an API where the model is a first-class user.</p><p>The entire public surface is a 12k token doc file, which is tight enough that even small models can include it as context. And, the whole thing is constrained by a type system. The API boundary is small and closed, and if the LLM hallucinates, it doesn’t even compile.</p><p>We don’t want to have to say “you should have used better prompts,” we need APIs whose affordances are small, clear, and typed enough for the model to use natively.</p><h3>How its Goin’</h3><p>We’ve rolled this out across a few internal applications, partnered with a couple of non-engineer builders to make sure it’s as easy as we think it is, and started building an internal SaaS for lightweight apps. In short, we’re testing this in a few constrained ways before making it more widely available. But, so far, so good!</p><p>We are all increasingly authoring software through the agent loop, and downstream of that are many small, semi-serious software artifacts, for which compute and tokens have a primitive, and storage does not.</p><p>So we made one shape of an answer and started trying it. Not because every prototype should live forever on object storage, but because the success path should be graduation, not regret. The alternative for many of these tools is not “use a better database.” It is that the experiment never happens, or quietly becomes another spreadsheet or shadow operational burden.</p><p>Give <a href="https://github.com/Gusto/baerly-storage">baerly-storage</a>, a try. Maybe it will help you build fast, too. Or, join us! Gusto is hiring. Check out our careers page at <a href="https://gusto.com/careers">gusto.com/careers</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ede96ac7ac27" width="1" height="1" alt=""><hr><p><a href="https://engineering.gusto.com/storage-is-the-missing-piece-for-agent-built-software-ede96ac7ac27">Storage Is the Missing Piece for Agent-Built Software</a> was originally published in <a href="https://engineering.gusto.com/">Gusto Engineering</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[First impressions with the web’s latest presentation tool: Spectacle]]></title>
            <link>https://medium.com/@baerbaerbaer/first-impressions-with-the-web-s-latest-presentation-tool-spectacle-7fab1f057051?source=rss-4077cea51e29------2</link>
            <guid isPermaLink="false">https://medium.com/p/7fab1f057051</guid>
            <category><![CDATA[react]]></category>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[web-development]]></category>
            <dc:creator><![CDATA[Eric Baer]]></dc:creator>
            <pubDate>Wed, 27 Jan 2016 15:42:21 GMT</pubDate>
            <atom:updated>2016-01-29T17:17:57.844Z</atom:updated>
            <content:encoded><![CDATA[<p><a href="https://github.com/FormidableLabs/spectacle">Spectacle</a>, if you haven’t seen it, is another in a <a href="https://github.com/search?l=JavaScript&amp;o=desc&amp;q=presentation&amp;s=stars&amp;type=Repositories&amp;utf8=%E2%9C%93">long line of HTML presentation tools</a>, but this one has rocketed into popularity (which isn’t exactly common for something like presentation tooling). Since I had heard it has a few neat tricks up its sleeve, and as a born and bred skeptic, I thought I’d get into the code and see what makes it interesting. Plus, it was written by my friend and co-worker <a href="https://medium.com/u/e50c25098d37">Ken Wheeler</a>, so if I find anything ridiculous I get to give him a hard time about it.</p><p>First, let’s let him make his case — it’s only fair. He recorded a really good screencast introducing the library so here it is:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FvvgtgnIhJ1g%3Ffeature%3Doembed&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DvvgtgnIhJ1g&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FvvgtgnIhJ1g%2Fhqdefault.jpg&amp;key=d04bfffea46d4aeda930ec88cc64b87c&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/c6f371ce4e470590941bb88a0dfec18b/href">https://medium.com/media/c6f371ce4e470590941bb88a0dfec18b/href</a></iframe><h4>Authoring Slides</h4><p>First, Spectacle is 100% powered by <a href="https://facebook.github.io/react/">React.js</a>, which in and of itself is rather uninteresting. What <em>is</em> neat, though, is that you can build your presentation using React components. The fine folks at <a href="http://formidable.com/">Formidable</a>, and most notably <a href="https://medium.com/u/e50c25098d37">Ken Wheeler</a>, have provided a few out-of-the-box components, like &lt;Slide&gt; and &lt;BlockQuote&gt;, that make building presentations as composable UI actually pretty awesome! To make writing slides even better, if you’re used to using React, you’ll find that authoring for Spectacle comes with all of the affordances that you’re used to with source code, like ES-next syntax via Babel (yes, it’s Babel 6.x).</p><p>Some parts of Spectacle are par for the course, like theming and transitions, but there are also a few smaller details that make authoring really nice. For one, the whole project is set up with <a href="https://github.com/gaearon/babel-plugin-react-transform">babel-plugin-react-transform</a>, which means that when you start it up, you’ll get hot-reloading in your browser while authoring slides. The other niceties are mostly the really well-put-together components that get you going quickly. The &lt;Layout&gt; component, for example, uses Flexbox, so you can &lt;Fit&gt; and &lt;Fill&gt; the text in a section on a slide.</p><p>At this point I was already somewhat convinced about writing presentations in React. But then, I realized that I could demo my React components live as a part of the presentation! Not only that, but I could use another <a href="https://medium.com/u/900ee8ad3db9">Formidable</a> creation, <a href="https://github.com/FormidableLabs/component-playground">component playground</a>, to live-edit the components I’m demoing and have side-by-side code + component. This means that I’m editing React components live inside of the Spectacle React components. Much meta!</p><p>The final thing I’ll mention here is that writing UI as React is sometimes a bit heavy handed, and for that, Spectacle provided &lt;Markdown&gt;. I’ll just show you the code on this one:</p><pre>&lt;Markdown&gt;{`<br>![Markdown Logo](${images.markdown.replace(“/”, “”)})</pre><pre>You can write:</pre><pre>* inline images, [Markdown Links](http://commonmark.org)<br>* paragraph text<br>* ~~strikethrough~~<br>* **bold**<br>* _italic_</pre><pre>Most other markdown syntax just works.<br>`}&lt;/Markdown&gt;</pre><h4>Presenting</h4><p>Even if you just use the Markdown components with hot-reloading, the presenter tools still make Spectacle pretty compelling. When you run the start command, you kick off a small (under 30 lines) Express server that hosts and renders the content at 2 URLs: one for content and one with a presenter view. (You know, it always amazed me that Google Presentations never quite got this right.) The presenter view shows you your slide, the next slide, the time left, and where you are in the deck.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Fg853u52edGxiU36N5LU-g.png" /></figure><p>From your presenter screen, you can use keyboard navigation to advance the slides. Oh, and yes — you can do full screen. If you’re curious about how it works, it’s React (and friends) alllll the way down. Spectacle uses the <a href="https://github.com/rackt/history">history module</a> that powers react-router and <a href="https://github.com/rackt/redux">redux</a> to manage URLs and state.</p><p>Now that all the content is available at localhost, you can use <a href="https://ngrok.com/">ngrok</a> to send out a URL and let all of the audience members follow along. If you haven’t heard of ngrok — go read about it! The basic idea is that ngrok creates a secure public URL (https://yourapp.ngrok.io) to a webserver running locally; in this case the aforementioned Express app running the presentation. Even if you think Spectacle is terrible and you hate React and fun and pizza and pool floaties and you just wish we all used Keynote or whatever, you should still go read about ngrok. You’ll never have to share another webex link to do a demo again!</p><p>The last thing worth mentioning is the obligatory post-talk PDF, which you can generate by going to a different URL that stacks all the slides into a single page with a bit of styling to make sure that it separates properly when printing.</p><p>In summary, if you’re one of the thousands of us currently addicted to React then this will probably be preferable to messing around in Keynote/PowerPoint/Impress.js/etc. Even if you’re not — I think there are some really compelling features in Spectacle that I hope will raise the bar for all of the presentation tools out there!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*Rt7ftk3_Ht23593P2K_fBA.gif" /></figure><p>Demo: <a href="http://stack.formidable.com/spectacle">http://stack.formidable.com/spectacle</a><br>Code: <a href="https://github.com/FormidableLabs/spectacle">https://github.com/FormidableLabs/spectacle</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7fab1f057051" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Using React is a Business Decision, Not a Technology Choice]]></title>
            <link>https://medium.com/@baerbaerbaer/using-react-is-a-business-decision-not-a-technology-choice-63c4641c5f7?source=rss-4077cea51e29------2</link>
            <guid isPermaLink="false">https://medium.com/p/63c4641c5f7</guid>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[react]]></category>
            <dc:creator><![CDATA[Eric Baer]]></dc:creator>
            <pubDate>Fri, 04 Dec 2015 23:38:56 GMT</pubDate>
            <atom:updated>2016-01-06T22:48:58.891Z</atom:updated>
            <content:encoded><![CDATA[<p><strong><em>Preamble</em></strong>:</p><p>React has become very popular among developers and there are lots of resources that speak to its technical merits. However, migrating to (or choosing) a new framework ultimately comes down to selling it to everybody at the table — including non-developers. There are very few engineering managers or PMs would agree to a re-write just because it’s the shiniest new thing and worse still, many organizations have been burned by the high churn in JavaScript tooling, which sadly moves projects backwards as a part of moving the web forward. This post is not an attempt to teach you something new about React. It is an attempt at the Executive Summary; a starting point for your pitch to try to sell not just developers but everybody on the wonders of React.</p><p><strong><em>Summary</em></strong>: React is a library for building composable user interfaces. There are many other tools like Angular, Backbone, Knockout and Ember that do similar things but in comparison, React grew out of solving business problems rather than technical ones. Below are a few of the reasons that React is important and how it might benefit any engineering organization.</p><h3>Reduced Risk</h3><p><em>React Stability</em> — Facebook is heavily invested in React (Newsfeed, Instagram, Messenger, Ads Marketplace etc.) and has a number of dedicated engineering resources working on the project. This dog-fooding and investment is not present in any of the competing frameworks. In addition to Facebook engineering, there is an enormous groundswell of excitement around React. There are currently 571 contributors (as of Dec 2015) to the project along with a conference series and a regular release cadence.</p><p><a href="https://github.com/facebook/react/wiki/Sites-Using-React"><em>React In Production:</em></a><em> </em>AirBnB, Asana, Atlassian, BBC, Chegg, CloudFlare, CNN.com, Codecademy, Coursera, Craftsy, Dailymotion, Dropbox, Expedia, Facebook, Feedly, Flipboard, HipChat, IMDb, Imgur, Instagram, Khan Academy, KISSmetrics, Mattermark, Minerva Project, Netflix, OkCupid, Rackspace, Rally Software, Ralph Lauren, Reddit, Redfin, Salesforce, Squarespace, The New York Times, Trunk Club, Twitter, Uber, University of Cincinnati, Venmo, WhatsApp, Wired, Wix, WordPress, Yahoo, Zendesk</p><h3>Development Efficiencies</h3><p><em>Strong Migration Path</em> — React can be dropped into a small area on any existing page regardless of how it’s already built which allows developers to migrate as quickly or slowly as they would like. It <em>is</em> worth noting that React requires a runtime library (39.4kb gziped for React v0.14.0) so piecemeal migration would negatively impact page weight until old libraries are removed.</p><p><em>Performant by Default</em> — React uses patterns that makes it very difficult to write poorly performing code. Additionally, since it eliminates direct interaction with the DOM it not only replaces an existing view layer (Angular/Backbone/Ember) it also obviates the need for a number of dependencies like jQuery which can result in an overall reduction in deployed code size.</p><p><em>SEO</em> — SEO is achieved by sending a completely rendered page from the server to the browser. React is designed with SEO in mind, rendering on either the client or the server, using Node. Other tools do allow for server side rendering but require a number of unstable hacks and a substantial amount of developer support to maintain. React has the potential to simplify build tooling and reduce maintenance costs.</p><p><em>Improved Code Reuse</em> — React is uniquely able to provide performance as well as a fully managed rendering lifecycle for components; the result is dramatically improved developer ergonomics. By making the creation, distribution and consumption of isolated reusable components more straightforward, developers are better able to save time by using and creating common abstractions. This is true of both low level elements like buttons and high level elements such as accordions.</p><h3>Improved Developer Effectiveness</h3><p><em>Reduced Complexity from Collocated Resources</em> — React mixes together HTML and JavaScript under the principle that they are inevitably tightly coupled and that separating them is separating technologies and not concerns. This concept can be extended further to include CSS, which removes a litany of issues involved with CSS development including global namespace, and scope/variable isolation. See: <a href="http://projects.formidablelabs.com/radium/">Radium</a> and <a href="https://speakerdeck.com/vjeux/react-css-in-js">React: CSS in js</a> for details.</p><p><em>Faster Error Isolation</em> — Facebook has provided <a href="https://chrome.google.com/webstore/detail/react-developer-tools/fmkadmapgofadopljbjfkapdkoienihi?hl=en">React specific developer tools</a> for the browser that provides information about which component and what data was used to generate a piece of UI. See <a href="https://www.youtube.com/watch?v=D-ioDiacTm8">this talk</a> for more information.</p><p><em>Easily Understandable Code</em> — In most current tools, <em>Models</em> representing data and <em>Views</em> displaying data are wired together to produce rich UI interactions. Often a change in one model or view (a shopping cart model for example) might trigger “cascading changes” across other views that also depend on the same data. In a sufficiently large application the dependency graph becomes complex and can be difficult to modify without introducing bugs from unexpected side effects. In React, data only flows in one direction which makes views much more understandable. Below are diagrams demonstrating both flows of information.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/984/1*pHvDgaslF8EClCehi6AiMA.png" /></figure><p><em>Improved Testability</em> — A component, React’s fundamental abstraction takes in data an argument and outputs a representation of the DOM with no additional side effects. By removing reliance on consuming and creating state stored in the DOM these components are both more atomic and more testable.</p><h3>Organizational Benefits</h3><p><em>Faster Onboarding</em> — React’s API is very small. Combined with it’s declarative syntax and componentization of UI elements makes it much faster to onboard new developers — specifically developers right out of college or those not as familiar with the frontend environment.</p><h3>Case Studies</h3><h3><a href="http://5by5.tv/changelog/149">Facebook Ads</a></h3><p>“It was extremely difficult to change without causing some side effect or bug somewhere else in the application … When the team rebuilt it in React they found that their rate of new bugs being introduced had gone through the floor”</p><p><em>-Spencer Ahrens</em></p><p>“When a bug did come in it was much easier to figure out what was going wrong and make a targeted fix”</p><p><em>-Spencer Ahrens</em></p><h3><a href="http://techblog.netflix.com/2015/01/netflix-likes-react.html">Netflix</a></h3><p>“React has exceeded our requirements and enabled us to build a tremendous foundation on which to innovate the Netflix experience”</p><h3><a href="https://developer.atlassian.com/blog/2015/02/rebuilding-hipchat-with-react/">Hip Chat</a></h3><p>“The dev speed we’ve gained…proves that we can release new client features faster and with more confidence on this platform than on any native client.”</p><p><strong><em>Correction</em></strong>: A number of people have pointed out that Google uses Angular heavily internally. While this could arguably be called dogfooding it was not my intention to imply that Google doesn’t use Angular at all, rather that I am not aware of Google using Angular for any of it’s large customer facing applications. This distinction is important since the requirements for customer facing applications is both more a more meaningful measure of commitment/usability/developer ergonomics and is a much better proxy for those who are interested in tooling options to use for their own applications.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=63c4641c5f7" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>