<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/">
    <channel>
        <title>Saadman Rafat</title>
        <link>https://paragraph.com/@saadmanhere</link>
        <description>undefined</description>
        <lastBuildDate>Sun, 06 Sep 2026 02:48:06 GMT</lastBuildDate>
        <docs>https://validator.w3.org/feed/docs/rss2.html</docs>
        <generator>https://github.com/jpmonette/feed</generator>
        <language>en</language>
        <image>
            <title>Saadman Rafat</title>
            <url>https://storage.googleapis.com/papyrus_images/2a0ec75f42f1af57412cd4b4c733843a13941854ffbe9ae2149448710e672aa5.jpg</url>
            <link>https://paragraph.com/@saadmanhere</link>
        </image>
        <copyright>All rights reserved</copyright>
        <item>
            <title><![CDATA[Prompting Lovable: Defensive AI and the Stages of Grief]]></title>
            <link>https://paragraph.com/@saadmanhere/prompting-lovable-defensive-ai-and-the-stages-of-grief-1</link>
            <guid>3js0VZmcDkBF9rG0tN1S</guid>
            <pubDate>Thu, 07 May 2026 18:00:00 GMT</pubDate>
            <description><![CDATA[What happens when you interrogate a new vibe-coding AI? A breakdown of Lovable's defensive guardrails and its five stages of grief.]]></description>
            <content:encoded><![CDATA[<p>In a few days, I’ll be interviewing the CEO of a new vibe-coding application. It’s a Lovable competitor, so I need to know as much as possible about Lovable. On top of that, I’ve never covered a product unveiling or conducted this kind of interview before.</p><p>So, who better than to ask Lovable?</p><p>Once you talk to a reasoning model long enough, you start to understand<br>its limitations, quirks, policies, and, in rare cases, underlying configurations. Gemini and I have a good rapport. It’s not without its flaws, but in most cases, they’re minor and not worth worrying about.</p><blockquote><p>Lovable however, in it's reply, went through all five stages of grief. Mind you,<br>the conversation is about the interviewer and the interviewee. I think it's probably<br>a good time to point out that neither a journalist or a tech blogger.</p></blockquote><p>Let me give you two examples involving popular LLMs that I encountered<br>over the past week. I asked Claude Code about its skyrocketing prices,<br>and it mostly owned up to the issue and sympathized. Gemini, meanwhile,<br>leaned into the joke about Google publicly investing in Claude. Even<br>when pressed on why Google kills so many of its products, it not only<br>agreed, but actively reasoned through why that was a bad decision by<br>its higher-ups. Go ahead and ask Claude about its monthly subscription<br>and quotas. It will tell you the reality of the situation. It even went<br>so far as to suggest that the Ollama Pro tier might be more worthwhile<br>and that I should consider it. Some times it's sycophancy other times it's<br>just good advice.</p><p>Lovable however, in it's reply, went through all five stages of grief. Mind you,<br>the conversation is about the interviewer and the interviewee. I think it's probably<br>a good time to point out that neither a journalist or a tech blogger.</p><p>What are the five stages of grief?</p><p>Now in the final act, let's dissect it's very laughable, insecure reply.</p><p>This is how it asked me to open the interview / unveiling on 11th May.</p><p><strong>DENIAL!</strong><br><em>"Why now? and what shifted in AI capability that made this possible?"</em></p><p>I have to admit this art meet comedy – <strong>ANGER!</strong></p><p><em>"..Show me the thing, Build something live, from scratch, in front of me."</em></p><p><strong>BARGAINING!</strong></p><p>I will paraphrase &amp; skip here since two bullet points here<br>since I know this is where Lovable falls short.</p><p><em>"How do you handle the boring<br>80%: auth, databases, secrets, migrations, rollbacks? And Have you thought of<br>broken code at 2am for paying users?"</em></p><p>I will just add things breaking at 2am and late Friday is sort of feature in our<br>profession.</p><p>This brings us to <strong>JEALOUSY</strong> and/or <strong>DEPRESSION</strong>, with a hint of <strong>DARING</strong>(!?),<br>you be the judge. It's a fairly large passage with four bullet points so I will try<br>my best to give you the gist.</p><p><em>"If we ship your headline feature next week, what do you have left? then goes on to say, if we [they] have thousands of users and a full Cloud backend. What's your wedge?"</em></p><p>Not sure where this exactly fits,</p><p><em>"You can't name three 'X killers', that actually killed X", "In eighteen months when it fails, what will the the postmortem say?" and "what do you genuinely admire about lovable?"</em></p><p>I had hard time classifying this.. let me know in the comments. it's certainly<br>doesn't sound like DEPRESSION or the 6th stage.. finding <strong>MEANING?</strong>.</p><p><strong>Finally,</strong> we have <strong>ACCEPTANCE!</strong></p><p><em>"what would change your mind about your own thesis?"</em></p><p>Then ending with calling me a strawman,</p><p><em>"Stay curious, not combative. The founder [competitor] has actually used the incumbent seriously, or is pitching against a strawman"</em></p><p>When asked to it's status on solving database, auth and scaling issue – the<br>model said it's still imperfect. Then suddenly took a 180 degree turn and<br>got too personal.</p><br>]]></content:encoded>
            <author>saadmanhere@newsletter.paragraph.com (Saadman Rafat)</author>
            <category>code</category>
            <category>vibe</category>
            <category>llms</category>
            <category>models</category>
            <enclosure url="https://storage.googleapis.com/papyrus_images/18b3a9a0a1387a2840cbbbb4f71ebd3065f4d8107ca4f9ab1bc4e166e54fbedb.jpg" length="0" type="image/jpg"/>
        </item>
    </channel>
</rss>