<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel>
<title>Ishan Ambasta</title><link>https://ishanambasta.com/</link><description>I solve problems at Alt Carbon. We remove carbon dioxide from the air by spreading crushed basalt on farmland. I make the software around that easier to use.</description><language>en</language>
<atom:link href="https://ishanambasta.com/rss.xml" rel="self" type="application/rss+xml"/><lastBuildDate>Tue, 11 Aug 2026 00:00:00 GMT</lastBuildDate>
<item><title>A website you keep when you leave</title><link>https://ishanambasta.com/writing/a-website-you-keep-when-you-leave</link><guid isPermaLink="true">https://ishanambasta.com/writing/a-website-you-keep-when-you-leave</guid><pubDate>Tue, 11 Aug 2026 00:00:00 GMT</pubDate><description>Why every colleague should own their site, account and all.</description><content:encoded><![CDATA[<p>Most company profile pages are rented. They live on the company’s domain, in the company’s system, and they vanish the week you hand in your laptop. Years of talks, papers and projects go with them.</p>
<p>We are trying something different. Every colleague gets a personal website on day one. The code sits in their own GitHub account and the hosting in their own Cloudflare account. The company supplies the template, the editor and a subdomain, and nothing the site needs to run.<sup><a href="#user-content-fn-1" id="user-content-fnref-1" data-footnote-ref aria-describedby="footnote-label">1</a></sup></p>
<p>The test is simple. Cut every company service and the site should still build, still serve, and still be editable. If that holds, then leaving the company is a permissions change, not a migration.</p>
<p>It also changes how people write. A page you will keep is worth more care than a page you will lose. People put their real work on it.</p>
<p>The cost to the company is small: a template, a few Workers, and the discipline to never put anything important on our side of the line.</p>
<section data-footnotes class="footnotes"><h2 class="sr-only" id="footnote-label">Footnotes</h2>
<ol>
<li id="user-content-fn-1">
<p>The subdomain is the one thing that does not survive. It redirects to the person’s own domain for ninety days, then goes away. <a href="#user-content-fnref-1" data-footnote-backref="" aria-label="Back to reference 1" class="data-footnote-backref">↩</a></p>
</li>
</ol>
</section>
]]></content:encoded></item>
<item><title>Plain HTML is a feature</title><link>https://ishanambasta.com/writing/plain-html-is-a-feature</link><guid isPermaLink="true">https://ishanambasta.com/writing/plain-html-is-a-feature</guid><pubDate>Tue, 02 Jun 2026 00:00:00 GMT</pubDate><description>A page that is only HTML loads everywhere, prints well, and still works in ten years. That is not a limitation. It is the point.</description><content:encoded><![CDATA[<p>A page that is only HTML loads everywhere, prints well, and still works in ten years. That is not a limitation. It is the point.</p>
]]></content:encoded></item>
<item><title>Forms are the product</title><link>https://ishanambasta.com/writing/forms-are-the-product</link><guid isPermaLink="true">https://ishanambasta.com/writing/forms-are-the-product</guid><pubDate>Thu, 19 Mar 2026 00:00:00 GMT</pubDate><description>Most internal software is a form. Treat it like one.</description><content:encoded><![CDATA[<p>Strip the dashboards away and most internal tools are a form that someone fills in at an awkward moment. Design the form first.</p>
]]></content:encoded></item>
<item><title>Offline-first forms for field teams</title><link>https://ishanambasta.com/writing/offline-first-forms-for-field-teams</link><guid isPermaLink="true">https://ishanambasta.com/writing/offline-first-forms-for-field-teams</guid><pubDate>Fri, 07 Nov 2025 00:00:00 GMT</pubDate><description>Field teams work where the network does not. A form that waits for a signal is a form that gets filled in from memory that evening.</description><content:encoded><![CDATA[<p>Field teams work where the network does not. A form that waits for a signal is a form that gets filled in from memory that evening.</p>
]]></content:encoded></item>
<item><title>What a carbon tonne is made of</title><link>https://ishanambasta.com/writing/what-a-carbon-tonne-is-made-of</link><guid isPermaLink="true">https://ishanambasta.com/writing/what-a-carbon-tonne-is-made-of</guid><pubDate>Thu, 21 Aug 2025 00:00:00 GMT</pubDate><description>Rock, rain, soil samples and a lot of spreadsheets.</description><content:encoded><![CDATA[<p>A tonne of removed carbon is a claim backed by measurements. When someone buys one, they are buying a chain of evidence that starts in a field and ends in a spreadsheet. This is a tour of that chain, from the software side.</p>
<h2 id="the-field">The field</h2>
<p>Basalt dust goes onto farmland. Rain dissolves it slowly, and the reaction pulls carbon dioxide out of the air and carries it away as dissolved carbon. Nobody can watch that happen, so we measure around it:</p>
<ol>
<li>What went onto the field: how much rock, from which quarry, of what composition.</li>
<li>What the soil looked like before: samples from fixed points, taken before spreading.</li>
<li>What changed after: the same points, sampled again across seasons.</li>
</ol>
<p>Each step produces paperwork. Most of the paperwork used to live in notebooks.</p>
<h2 id="the-sample">The sample</h2>
<p>A soil sample is a bag of earth with a label. The label is where most of our problems started. A smudged date, a field name spelled two ways, a point number written down from memory that evening. The lab result is only as good as the label it comes back attached to.</p>
<blockquote>
<p>A measurement you cannot trace back to a place and a time is not a measurement. It is a number.</p>
</blockquote>
<p>So the first tool we built was not clever. It is a form that runs on a phone with no signal, fills in the field and point from GPS, and prints a label. The person in the field types almost nothing.</p>
<h2 id="the-spreadsheet">The spreadsheet</h2>
<p>Results come back from the lab as files. We join them to the sample records and check them before anyone does any arithmetic:</p>
<table>
<thead>
<tr>
<th>Check</th>
<th>Why it matters</th>
</tr>
</thead>
<tbody>
<tr>
<td>Every result has a sample</td>
<td>Orphan results are usually typos in the label</td>
</tr>
<tr>
<td>Every sample has a result</td>
<td>Missing results are lost bags or lab backlog</td>
</tr>
<tr>
<td>Values sit in a sane range</td>
<td>A misplaced decimal looks like a miracle field</td>
</tr>
</tbody>
</table>
<p>The joining step is short enough to show:</p>
<pre class="astro-code astro-code-themes github-light-high-contrast github-dark" style="--shiki-light:#0e1116;--shiki-dark:#e1e4e8;--shiki-light-bg:#ffffff;--shiki-dark-bg:#24292e; overflow-x: auto;" tabindex="0" data-language="python"><code><span class="line"><span style="--shiki-light:#0E1116;--shiki-dark:#E1E4E8">results </span><span style="--shiki-light:#A0111F;--shiki-dark:#F97583">=</span><span style="--shiki-light:#0E1116;--shiki-dark:#E1E4E8"> lab.merge(samples, </span><span style="--shiki-light:#702C00;--shiki-dark:#FFAB70">on</span><span style="--shiki-light:#A0111F;--shiki-dark:#F97583">=</span><span style="--shiki-light:#032563;--shiki-dark:#9ECBFF">"sample_id"</span><span style="--shiki-light:#0E1116;--shiki-dark:#E1E4E8">, </span><span style="--shiki-light:#702C00;--shiki-dark:#FFAB70">how</span><span style="--shiki-light:#A0111F;--shiki-dark:#F97583">=</span><span style="--shiki-light:#032563;--shiki-dark:#9ECBFF">"outer"</span><span style="--shiki-light:#0E1116;--shiki-dark:#E1E4E8">, </span><span style="--shiki-light:#702C00;--shiki-dark:#FFAB70">indicator</span><span style="--shiki-light:#A0111F;--shiki-dark:#F97583">=</span><span style="--shiki-light:#023B95;--shiki-dark:#79B8FF">True</span><span style="--shiki-light:#0E1116;--shiki-dark:#E1E4E8">)</span></span>
<span class="line"><span style="--shiki-light:#0E1116;--shiki-dark:#E1E4E8">orphans </span><span style="--shiki-light:#A0111F;--shiki-dark:#F97583">=</span><span style="--shiki-light:#0E1116;--shiki-dark:#E1E4E8"> results[results[</span><span style="--shiki-light:#032563;--shiki-dark:#9ECBFF">"_merge"</span><span style="--shiki-light:#0E1116;--shiki-dark:#E1E4E8">] </span><span style="--shiki-light:#A0111F;--shiki-dark:#F97583">!=</span><span style="--shiki-light:#032563;--shiki-dark:#9ECBFF"> "both"</span><span style="--shiki-light:#0E1116;--shiki-dark:#E1E4E8">]</span></span></code></pre>
<p>Only after the <code>orphans</code> list is empty does anything go to the scientists.</p>
<h2 id="the-tonne">The tonne</h2>
<p>What a buyer finally sees is one number with a range around it.<sup><a href="#user-content-fn-1" id="user-content-fnref-1" data-footnote-ref aria-describedby="footnote-label">1</a></sup> Behind it sit a few hundred bags of soil, a few thousand form submissions, and a long list of small corrections nobody will ever read. That list is the product. The number is a summary of it.</p>
<section data-footnotes class="footnotes"><h2 class="sr-only" id="footnote-label">Footnotes</h2>
<ol>
<li id="user-content-fn-1">
<p>The range is not a disclaimer. It is the honest part of the claim, and it gets narrower as the evidence improves. <a href="#user-content-fnref-1" data-footnote-backref="" aria-label="Back to reference 1" class="data-footnote-backref">↩</a></p>
</li>
</ol>
</section>
]]></content:encoded></item>
<item><title>Tools nobody asked for</title><link>https://ishanambasta.com/writing/tools-nobody-asked-for</link><guid isPermaLink="true">https://ishanambasta.com/writing/tools-nobody-asked-for</guid><pubDate>Thu, 30 Jan 2025 00:00:00 GMT</pubDate><description>The most used internal tool we have was built in an afternoon for one person. Nobody asked for it. Everybody uses it.</description><content:encoded><![CDATA[<p>The most used internal tool we have was built in an afternoon for one person. Nobody asked for it. Everybody uses it.</p>
]]></content:encoded></item>
<item><title>Boring software</title><link>https://example.com/boring-software</link><guid isPermaLink="true">https://example.com/boring-software</guid><pubDate>Sat, 12 Oct 2024 00:00:00 GMT</pubDate><description>Boring software is software you do not have to think about. It is harder to build than it looks.</description></item>
</channel></rss>