Youssef • Webflow Designer & Developer

@chalaska I respect the hustle, but try running an auto-repair shop. I've currently got 10 heavy duty machines waiting to be unpacked, assembled and installed. (4 post car jacks, alignment machines, weight is about 3000kg)

Digital business are cool but have you tried physical? Been setting up an AI-enabled and physiotherapy-led rehabilitation practice in Melbourne …and yes my hands are destroyed from setting these up

@fayhecode I now replied to you via DMs, as I wished to send a longer message.

JazakAllahu khayran for taking the time to explain your position. I think we've identified the core disagreement. My concern is reducing errors and grounding answers in authentic sources, while your concern is that inference itself is inappropriate in this context. What I find interesting is that throughout this discussion, I never argued that people should trust a generic LLM. In fact, I have repeatedly explained I implemented safeguards, source grounding, references, and limitations built into the feature. Yet the conversation keeps returning to LLMs in the abstract, rather than the actual implementation being discussed. For what it's worth, this is also the reason why the feature is only available in French. I deliberately limited its scope rather than presenting it as a source of authority. The app already includes Ibn Kathir tafsir, and there are safeguards designed to reject content that is not grounded in Quranic sources. There is also a disclaimer inside the app stating that it is trained on Quranic content to help users find information and references, not to provide religious rulings or scholarly opinions. Most importantly, the user can directly verify the cited verse and source themselves rather than blindly accepting a generated answer. The Quran verses themselves are not generated by the model. Once a verse reference is selected, the verse is displayed directly from a local Quran database without inference. This means the system may retrieve an irrelevant verse, but it cannot invent a Quran verse that does not exist. See it as a Google search trained only for the Quran. I also showed the app to the imam of my local mosque. He understood how the feature works and viewed it as a useful way to help people find references, rather than as a replacement for scholars. I appreciate that you took the time to explain your concerns, and I respect your intention to advise. I simply believe we have reached different conclusions on this matter. May Allah guide us both to what is correct.

Quit my job 6 months ago. Already 4 apps on the App Store, still stuck at 100$MRR and talking to myself on X🫪

This is correct.

I did not see your DM. X's DMs are not very reliable due to the amount of spam there. However, now as we're discussing publicly, perhaps it's of benefit to others to continue publicly. I'll address your points one by one. "you publicly concluded that the app is harmful without asking a single question about how it works, what safeguards are in place, or what limitations are communicated to users." This does not matter. Because there remains the core problem with the app, which is plainly obvious and does not need further questioning. The problem with the app is that it involves tokenization and inference; where each token is generated by probability, not by knowledge. Even if cross-referenced, it's still based upon probability. This alone is enough to dismiss it outright In fact, it might be valid to say that LLMs are a kind of divination, which is outright haram; but I will not go that far for now, as this would require more research. However, the reason I say this is because we don't really know how LLMs reason, even though we know how to make them. (But this is a different conversation for another time). "The app is not a generic chatbot answering from memory. It uses retrieval-based grounding, which is exactly the direction explored in recent research on faithful Islamic question answering: arxiv.org/abs/2601.07528" It would be better if the app was a chatbot pulling answers from memory. This way, the answers in memory could be written and curated by true scholars of Islam. Rather than inference; which is prone to error, hallucination, and agreeableness. As for the paper you linked to; I know nothing of the people who created it or their motives. However, from reading the PDF, they say it should not be used as a replacement for qualified scholarly guidance (See screenshot below). So your own reference refutes you. "Every source of religious knowledge outside the original Arabic text carries a risk of misunderstanding, including translations, tafsir books, websites, lectures and even human explanations." I'm not sure exactly what you mean. Are you saying that only the Quran is correct? Of course not, we know that the three righteous generations were upon the correct understanding and so we can take from them. But still, an error made by a human is not the same as an error made by a machine. This is why I am warning you against this; because you will be responsible for any errors this machine makes. I want good for you. I don't want you to fall into this sin; nor do I want anyone reading the AI generated tokens to misunderstand Islam. This would be a major problem. If you want to ground it in valid sources, then just directly and transparently incorporate Ibn Kathir, As Sadi, and others into your app. May Allah make this easy for you. "Even the companions of the Prophet were known for their caution and hesitation when answering religious questions. Yet you seem completely certain about a system you have not investigated and whose implementation you do not know." We are to be certain about what is certain. This app generates answers using inference and this is a problem due to lack of true knowledge, accountability. "In Islam, new things are not considered forbidden by default. If someone wishes to claim that such a tool is impermissible, the burden is on them to demonstrate why." You're correct. Things are not haram by default. Nor did I declare what you're doing is haram. That is not my point. My point is that AI is not appropriate or sufficient and is prone to misleading people For example, in the limitations section of that study it says: "In addition, our grounding is primarily Quran centric, so questions best supported by authenticated hadith, fiqh sources, or scholarly consensus may be disadvantaged." So this alone makes what you're doing wrong. Brother, I can't be clearer about this. You should remove AI from your app. Focus on adding in true sources of knowledge. Include bookmarking, highlighting, notes, progress indicators, and so on. This will help people. AI will not. Go to Medina, speak with the real scholars about this. May Allah help us all.

Wa alaykoum salam wa rahmatullahi wa barakatouh. I reached out privately first because I thought this discussion deserved a private conversation. You didn't answer. Instead, you publicly concluded that the app is harmful without asking a single question about how it works, what safeguards are in place, or what limitations are communicated to users. Saying that this is "100% wrong" actually proves my point. You show no uncertainty whatsoever, despite not knowing how the system works. The app is not a generic chatbot answering from memory. It uses retrieval-based grounding, which is exactly the direction explored in recent research on faithful Islamic question answering: arxiv.org/abs/2601.07528 Every source of religious knowledge outside the original Arabic text carries a risk of misunderstanding, including translations, tafsir books, websites, lectures and even human explanations. The question is not whether a risk exists, but whether we are taking reasonable steps to reduce it. That is precisely why the app grounds answers in retrieved sources rather than relying purely on model memory. Even the companions of the Prophet were known for their caution and hesitation when answering religious questions. Yet you seem completely certain about a system you have not investigated and whose implementation you do not know. In Islam, new things are not considered forbidden by default. If someone wishes to claim that such a tool is impermissible, the burden is on them to demonstrate why.

As-salamou alaykoum wa rahmatoullahi wa barakatouh ; frère, j’ai revérifié ton application et il s’agit d’un chatbot IA pour le Coran auquel les utilisateurs peuvent poser des questions. Ceci est dangereux. Je suis sûr à 100 % qu’il ne répondra pas correctement à toutes les questions. C’est la religion d’Allah. Ce n’est pas quelque chose pour lequel on peut volontairement permettre la diffusion de mauvaises informations. En tant que musulmans, nous sommes ordonnés d’ordonner le bien et d’interdire le mal. Cette application est une machine, pas un Akhi ni une Ukhti ; ce n’est pas un musulman et elle n’a pas d’âme. Elle n’a aucun sens du bien et du mal et ne peut pas être tenue responsable. Au contraire, tu seras tenu responsable des personnes qui seront égarées au sujet de l’Islam, frère. Retire la fonctionnalité IA. Ce n’est ni bon pour toi ni pour les autres musulmans. Tu as été informé. (Yes, I used AI to translate into french because I hoped it would get your attention and be clearer. If there are errors, then my point is also proven).

@fayhecode And, as I have greeted you with the Salam, it is wajib upon you to return it. As Allah says in the Quran that we must return the Salam with that which is equal or better.

@mason__capital I've been shorting FIG.

@YS Send your portfolio

$FIG As a designer, there is nothing more frustrating than seeing your designs get built in prod with missing details corner radii, line thickness, text placement etc — the details that lead to a crafted product Figma Make <> your codebase is an incredible solution. Bring designers closer to production!

@dariusdan You've made it when you truly don't care.

You know you’ve made it as a designer when someone copies your work and you feel flattered for exactly 3 seconds before the rage kicks in.😆

@DylanCalluy Left is iPhone. You can tell with by the way the details around the drain cover are rendered. It's a little muddy.

One is iPhone. One is Leica Q3. Guess away!

@luckenco After watches, you will discover telescopes. Kinda game over then. The ticking of the celestial bodies is something else to witness.

like wtf

@cedric_design @billow_app Congrats, I hope it goes well for you.

The wind broke our poster yesterday because it couldn't handle the hype. Today we break cover. The Billow site is live. Join the waitlist to be part of cohort 1 which will be opening up soon! (Only 200 spots) 🔗👇

Assalamualaikum - May Allah reward you for your efforts and guide you to what is good. However, you must not mix AI and Quran. It is not correct, and will lead people to misunderstand the Quran; which you will be held accountable for. I spoke with another brother on X about this issue. Please read this thread. x.com/YS/status/2025…

@YS Yes I am

@timsoulo @ahrefs "Adding schema markup had zero meaningful impact on AI citations." I knew it. I've explained this to clients so many times. If schema works, then AI fails. The point of AI, and crawling in general, is to find the best information. Not the prettiest.

In the last 6 months at @Ahrefs, we analyzed over 1 billion data points across 14 studies. Here's what we learned about AI search optimization: 1) "Best X" blog listicles are the single most prominent content format cited by AI chatbots. They make up 43.8% of all page types cited by ChatGPT specifically. 2) 67% of ChatGPT's top 1,000 citations come from sources marketers can't influence: Wikipedia (29.7%), homepages (23.8%), app stores (6.6%). Only 32.3% are influenceable content like educational pages, reviews, news, and blog posts. 3) 28.3% of ChatGPT's most-cited pages have zero Google organic visibility. These pages get cited repeatedly by ChatGPT despite not ranking in Google at all. A completely separate discovery layer. 4) ChatGPT only cites about 50% of the URLs it retrieves. It fetches dozens of pages per query but uses half as background context without attribution. This means that being retrieved and being cited are very different things. 5) Adding schema markup had zero meaningful impact on AI citations. AI Overviews actually dipped −4.6%, while AI Mode (+2.4%) and ChatGPT (+2.2%) showed changes indistinguishable from zero. 6) YouTube mentions have the highest correlation (0.737) with AI brand visibility out of all the factors we studied (including all the conventional SEO metrics like backlinks, page count, DR, etc). This held true for both Google-owned and OpenAI products. 7) AI Overviews reduce clicks to the #1 result by 58%. That’s up from 34.5% just 10 months earlier. The trend is accelerating. 8) 99.9% of AI Overviews appear on informational intent queries. Transactional, navigational, and local searches are almost entirely AIO-free. Shopping triggers AIOs just 3.2% of the time. 9) For a given search query, Google’s AI Mode and AI Overviews reach the same conclusions 86% of the time — but cite almost entirely different sources (only 13.7% citation overlap). 10) AI Overviews change every 2.15 days on average, with 70% of content differing between consecutive observations. But semantic similarity stays at 0.95. The words, sources, and entities constantly shuffle, but the actual meaning barely moves.

"Adding schema markup had zero meaningful impact on AI citations." I knew it. I've explained this to clients so many times. If schema works, then AI fails. The point of AI, and crawling in general, is to find the best information. Not the prettiest.

@cedric_design @framer @TrueformCo This the joy of designing in code. When the design is done, the dev is done. Very rare do you need to redo stuff, you just refine and end up with the right result.

At Trueform, we shipped over 40 @framer sites in 2025 with a team of only two. Here's how we did it 🧵

It is well known that in Indian culture there are entire industries whose purpose is to provide people with fake credentials. Fake degrees, diplomas, even PhDs. These are often attested directly from the source, not copies. Just do a quick search for "Fake Degrees India" and you'll see. It's crazy. As a result, many people from India, compared to the West, do not have a strong and serious appreciation of genuine credentials and skills. They just exaggerate or lie their way into business by promising the sun, moon, and stars, without being able to actually deliver. When exposed, they just do not get it or see why it is an issue in the first place. There is a clear cognitive dissonance. It's crazy.

Irish tech recruiters getting Indian fatigue

@TimoBuilds_ A better question to ask is why he paid in the first place; this is what you want to optimize for.

oh shit.. just realized that my only recurring paying customer just left the building that means my startup is back at €0 MRR 😭 i reached out and asked one simple question: what made you quit? too expensive? not useful enough? missing something? wrong timing? this one answer might teach me more than another week of building in case of he won't reply, we need to dive deeper into @posthog analytics rn