<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Made by Asking</title>
    <description>The latest articles on DEV Community by Made by Asking (@madebyasking).</description>
    <link>https://dev.to/madebyasking</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4205637%2F69a7f400-9565-4975-9920-4eb017156e70.png</url>
      <title>DEV Community: Made by Asking</title>
      <link>https://dev.to/madebyasking</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/madebyasking"/>
    <language>en</language>
    <item>
      <title>A live call coach with two modes: interviews and meetings</title>
      <dc:creator>Made by Asking</dc:creator>
      <pubDate>Mon, 26 Oct 2026 03:45:56 +0000</pubDate>
      <link>https://dev.to/madebyasking/a-live-call-coach-with-two-modes-interviews-and-meetings-2j42</link>
      <guid>https://dev.to/madebyasking/a-live-call-coach-with-two-modes-interviews-and-meetings-2j42</guid>
      <description>&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fy62iyyhmf7r106p3g5qy.gif" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fy62iyyhmf7r106p3g5qy.gif" alt="The interview view of the UI demo" width="720" height="360"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Try the UI demo in the browser, no install: &lt;a href="https://henrylop.github.io/live-call-coach/ui/" rel="noopener noreferrer"&gt;https://henrylop.github.io/live-call-coach/ui/&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Interview mode
&lt;/h2&gt;

&lt;p&gt;In an interview, the right example is somewhere in your head and it surfaces five minutes late. &lt;code&gt;live-call-coach&lt;/code&gt; is a small package that listens to the other side of a call and shows tiered hints while the question is still fresh.&lt;/p&gt;

&lt;p&gt;The quick tier answers in about 1.5 seconds with an opening line, three key points and a hard-question flag. The deep tier is grounded in your own notes and cites them. When a new question starts before a card is ready, the stale card is dropped. The whole pipeline runs on a simulated clock, so replays are instant and the tests are deterministic.&lt;/p&gt;

&lt;p&gt;There is a UI demo in &lt;code&gt;ui/&lt;/code&gt; that needs no server: open &lt;code&gt;index.html&lt;/code&gt; and watch a scripted interview play through. The demo and the tests use an offline stub, so nothing here calls a paid API.&lt;/p&gt;

&lt;h2&gt;
  
  
  Meeting mode
&lt;/h2&gt;

&lt;p&gt;Work calls ask for facts you own but cannot look up in the moment. &lt;code&gt;live-call-coach&lt;/code&gt; has a meeting mode for that: for each question it builds a card with a line to say and the facts behind it, with their sources, and it keeps a running ledger of commitments (who, what, when).&lt;/p&gt;

&lt;p&gt;Two details carry most of the design. A card that is still being built when the next question starts is dropped rather than shown late. And the reader that follows the call is read-only: it has an allow-list with one entry, turning on captions, and a test scans the source for anything that types or sends.&lt;/p&gt;

&lt;p&gt;The UI demo in &lt;code&gt;ui/&lt;/code&gt; replays an invented call with no server, so you can see the cards arrive, a stale card leave and the ledger fill. The code runs offline against a stub model.&lt;/p&gt;

&lt;p&gt;Repository: &lt;a href="https://github.com/HenryLop/live-call-coach" rel="noopener noreferrer"&gt;https://github.com/HenryLop/live-call-coach&lt;/a&gt;&lt;/p&gt;

</description>
      <category>python</category>
      <category>opensource</category>
      <category>ai</category>
      <category>productivity</category>
    </item>
    <item>
      <title>Four rules for letting an AI agent deploy to Salesforce</title>
      <dc:creator>Made by Asking</dc:creator>
      <pubDate>Sun, 25 Oct 2026 19:34:17 +0000</pubDate>
      <link>https://dev.to/madebyasking/four-rules-for-letting-an-ai-agent-deploy-to-salesforce-bh7</link>
      <guid>https://dev.to/madebyasking/four-rules-for-letting-an-ai-agent-deploy-to-salesforce-bh7</guid>
      <description>&lt;p&gt;In a Salesforce deploy, the slow part is rarely the deploy command. It's working out what goes in the package, which tests to run and which org to hit, and then, when it fails, figuring out what the error means for &lt;em&gt;this&lt;/em&gt; change. Both are language problems, and language is what models are good at.&lt;/p&gt;

&lt;p&gt;Deploying is not a language problem. So when we designed an agent that takes a change from a plain request to a verified deploy, the first question wasn't "what can the model do" but "what must it never do". These are the four rules it runs under.&lt;/p&gt;

&lt;h2&gt;
  
  
  1. The model never touches an org
&lt;/h2&gt;

&lt;p&gt;The model gets no credentials, no shell and no sf CLI. It gets a fixed menu of actions: validate, deploy, run tests, query, retrieve. It picks one and fills in the parameters; code checks them and runs the action with the sf CLI and git. If an action isn't on the menu, it doesn't exist.&lt;/p&gt;

&lt;p&gt;That one rule removes a whole category of "the agent ran something creative" stories.&lt;/p&gt;

&lt;h2&gt;
  
  
  2. Humans approve at the right moments, not every moment
&lt;/h2&gt;

&lt;p&gt;Approving every step is how people learn to click Approve without reading. So there are three gates:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;The plan.&lt;/strong&gt; Components, target org, tests and risk on one card. Approve, edit or reject.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A changed plan.&lt;/strong&gt; If the fix after a failure changes the components or the scope, it comes back to you. A same-plan retry doesn't.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Production.&lt;/strong&gt; A check-only validation runs first, then a human click deploys. Always. Destructive changes too.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Sandbox steps run on their own while everything stays green.&lt;/p&gt;

&lt;h2&gt;
  
  
  3. "Done" means checked, never the model's word
&lt;/h2&gt;

&lt;p&gt;After a deploy, code queries the org: is the Flow active, did the tests pass, does the profile have field access, does the validation rule actually block the case it was written for? The result is an evidence table. The model writes the summary from that table, not the other way round.&lt;/p&gt;

&lt;h2&gt;
  
  
  4. Every run is logged and capped
&lt;/h2&gt;

&lt;p&gt;Plans, actions, approvals and outputs go into one log per run, with a cost cap the run can't go past and a limit of three retries on failures. When it stops, it says why.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the model is actually for
&lt;/h2&gt;

&lt;p&gt;Two jobs. Turning "add a Discount Approved checkbox to Quote, with a validation rule that blocks a discount over 20% unless it's checked, UAT only" into a plan a reviewer can read in 20 seconds. And turning a failed deploy into a sentence:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;An existing test builds a quote with a 25% line discount and doesn't set the new checkbox, so the new rule blocks it. Fix: set the checkbox in that test's setup.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;Those two jobs are its whole role. The rest is ordinary code.&lt;/p&gt;

&lt;h2&gt;
  
  
  Early access
&lt;/h2&gt;

&lt;p&gt;We're opening this to a small group of Salesforce admins, developers and release folks first. If deploys are a weekly pain for you, the page is here: &lt;a href="https://made-by-asking.pages.dev/deploy/?utm_source=devto" rel="noopener noreferrer"&gt;https://made-by-asking.pages.dev/deploy/?utm_source=devto&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;There's one optional question on the form, and the answers decide what gets built first.&lt;/p&gt;

&lt;p&gt;&lt;em&gt;AI-assisted: drafted with Claude Code; reviewed and edited by a person.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>salesforce</category>
      <category>devops</category>
      <category>ai</category>
      <category>agents</category>
    </item>
  </channel>
</rss>
