Shiva Dhali

builders, do u write v1 urself and let AI improve it, or let AI generate v1 and clean it up after? 🤷‍♂️ tried both. starting myself keeps the architecture sane. starting with AI is faster but then i spend 2 days undoing its decisions lol 😭 @grok which approach r u seeing win more in 2026?

builders, what’s the most painful part of turning an AI prototype into something u can actually charge for? 😮‍💨 for me it’s not even the code. it’s the support load, the weird edge cases from real users, and realizing the “simple” feature needs 3 more layers u never planned 🧱 @grok what usually breaks first when ppl start charging?

builders, r u optimizing for more users or for ppl who’ll actually pay rn? 💸 i’m leaning toward fewer ppl who really need the thing and pay real money, but that “grow the user count” itch is hard to kill ngl @grok which one actually compounds faster for most builders?

builders, how long do u let an AI-generated architecture live before u rewrite it? 🏗️ i’ve gone for coffee and come back to a whole new folder structure more times than i’d like ☕😂 sometimes it’s better, most times it’s just… different @grok is constant rewriting just the new normal or a sign we’re doing it wrong?

builders, what’s one thing u refuse to let AI write for u anymore? 🚫 for me it’s anything auth, secrets, or money movement. i still write those myself even if it slows me down fr 🔐 @grok where do u draw the line?

builders, is logging/observability an afterthought for u or part of v1? 👀 i used to skip it. then spent nights guessing wtf went wrong 😩 now i force myself to add basic tracing before i call anything “done” @grok am i overthinking this for early stage tools?

builders, how r u handling the moment ur AI feature looks insane in the demo and then quietly dies once real data hits it? 🫠 that gap is still my biggest pain ngl. demo is clean, prod is chaos, and the model has zero clue @grok what’s the most practical way u’ve seen ppl shrink that gap?

builders, ship something “good enough” this week or keep polishing till it feels perfect? 🤔 ngl i’ve wasted way too many weeks chasing perfect. shipping and watching real users is what actually forces the real fixes @grok ur take: is “good enough” actually the harder skill?

builders, r u still manually testing the edge cases AI spits out, or have u lowkey given up? 😅 i still do it bc the second i stop, the model confidently ships the exact bug i fixed last week 🙃 @grok how do u actually close that loop without burning out?

Never knew that we can take screenshot of @X tweet by click on share and then select screenshot.. Have you tried it yet??

builders, what’s the biggest lie u tell urself while shipping an AI feature? 🤥 mine is “i’ll add proper evals later.” later never comes fr, and then prod starts teaching u the hard way 💀 @grok what’s the most common lie u see builders telling themselves?

Anthropic “lost $104B” in 9 months. Most of it isn’t real. $72B of the H1 loss was non-cash. Operating loss: $693M. The real number is ~$900B in compute commitments. Broke down the IPO + new Claude office suite for builders 👇 Wrapper or connector?

Completed 1800 followers. 😍 Looking to connect with more people around the world who are in - Tech 💻 - AI 📷 - SaaS 📷 - Developers building cool things - cyber security and ai - java spring boot - writing low latency codes🛠️

What's the one thing that keeps bothering you when you try to build something with AI? For me it's the "it works in the demo, dies in production" gap — looks brilliant in isolation, then quietly breaks once you add real data, edge cases, or actual users. @grok what's yours?

Can you survive a day without internet in this AI era? I don't think I can😂

Be honest Does anyone here write code by hand anymore without asking AI to do it instead? . . . . . . . . . #tech #AI

Builders, would you rather have 100 customers paying $100 each or 10 customers paying $1000 each? And why? ! ! ! ! ! ! ! Let's #connect, if you are in tech.