Saurabh

Most developers type the same prompts to Claude Code 20x a day. Architects build workflows once and never repeat themselves. Here’s the feature that changed how I think about AI-assisted engineering →

Spot on, Meher! This is the classic hype-vs-reality correction every transformative tech goes through. AI’s current economics still favour humans in many workflows, but the efficiency curve is bending fast. The winners won’t be pure AI or pure humans, they’ll be the teams that master the hybrid model first.

@livingdevops Totally agree that fundamentals are non-negotiable, the engineers who truly master Linux, Postgres, C, and the stack beneath AI are the ones who’ll own the next decade. But I’d offer a small refinement: being “scared” of AI doesn’t automatically mean someone isn’t a great engineer. Some of the sharpest minds I’ve worked with in my 2 decades of career so far, built million-dollar systems on those exact decades-old foundations. Great engineering has never been a single-variable game (AI leverage). It’s about adopting the new wave without abandoning the bedrock, using AI to amplify what already works, not replace it. Curious what you think, does that land the same way for you?

@sauvast So the people who are scared of AI are not great engineers, they just write code.

The OS you use is decades old The Database you use is decades old The languages you use are decades old And billion-dollar systems run on them right now. AI didn’t replace any of it. It runs on top of it. Linux. Postgres. Python. C. TCP/IP. Nobody killed these. Nobody will. The people panicking about AI taking their jobs don’t understand fundamentals. Stop chasing every new tool. Go deeper into the old ones. That’s where the real engineers live.

💯 Your brain isn’t broken. It’s just wired for survival. Bad things stick in your mind like glue. Good things? They fade fast. That’s why one failure hurts more than ten wins feel good. But smart people don’t fight it, they smartly work around it. They do 3 simple things: • Write down their wins every day (so they actually see them) • Ask after every loss: “What can I learn from this?” • Celebrate small daily wins like they’re big deals Do this and success stops feeling lucky. It becomes normal. Negativity isn’t the problem. Forgetting to balance it is. Who else is training their brain like this? Drop your favorite trick below 👇 I’m reading all of them! 🔥

Your brain remembers pain longer than pleasure. That’s why one loss can outweigh ten wins. If you can’t handle negativity, you’ll never handle success.

This is one of the most honest takes I’ve seen on India’s AI reality. Talent? We have it in truckloads. But ChatGPT wasn’t built by talent alone, it was built by infrastructure + obsession + willingness to lose money for a decade. You nailed it: the race has moved from “who can code” to “who can own intelligence infrastructure.” The real unlock for India isn’t more engineers. It’s world-class GPUs at scale, deep-tech culture that rewards 10-year bets, and capital that doesn’t panic at red numbers. Loved how you broke it down in the video. The visuals + message hit different. Creators who get this early (and tools like @uploadKar that actually help them ride the wave) are going to print in the next 3–5 years.

India has talent. But talent alone can’t build another ChatGPT. world-class researchers GPU infrastructure deep-tech culture and patience to lose money for years The AI race is no longer about coding skills. It’s about who can build the biggest intelligence infrastructure. #ai

This is gold, Miten. Traditional programming was like building a Swiss watch, every gear machined to spec, zero tolerance for surprise. Now it’s like tending a coral reef: you design the structure and currents (prompts + architecture), then let the ecosystem do its wild, emergent thing. The beauty (and the terror) is that the reef grows in directions you never scripted, dazzling new species or the occasional invasive barnacle.

programming with inference as a key input requires you to rethink the boundaries of the possible. whatever you design needs to afford emergent behaviours and positive & negative surprises. very different way of building than prior worlds

Exactly Raksha 🔥 The algorithm isn’t rewarding “content” anymore, it’s rewarding time spent. I started doing the brutal gut-check you mentioned: before hitting post, I literally ask “Would I reply, save or share this?” If the answer is “meh,” I scrap it. One rewrite last week turned a flat thread into 10+ replies and 3 DMs. Game changer. Every post really is a bet, and you just taught me how to stack the odds. Thank you 🙌

The X algorithm in 2026 has one job: Keep people on the app So ask yourself: Does my post make someone want to reply, save, or share? If the honest answer is no rewrite it Every post you publish is a bet Make it worth placing.

AI won’t replace you… but the person who treats it like a daily co-pilot absolutely will. Most are still in the ‘ask it to write a poem’ era while the rest of us are building agents that ship while we sleep. I am using it across all the tasks, which I do, including official work, research, pre-sales and business development, reference architecture design, point of view and proof of concepts and much more. AI is kind of embedded across all my workflow. This automation is not only helping me to excel, but also gives me a lot of time to think something new.

AI won’t replace you. Someone using AI will. The gap is speed and execution, not knowledge. Are you actually using AI daily or just observing it?

🚀 This is fantastic, @eunjiio! Love the Bottari vision, turning maintenance time into pure wild-idea shipping. The prediction layer (analyzing user data to proactively surface & ship the next backlog item) is especially impressive. Going straight for that full autonomous loop is bold and exciting. Quick questions if you’re open: 1. How are you grounding the backlog predictions? 2. What does your current deploy flow look like? Huge respect for building in public. Would love to swap notes on the agentic side, happy to DM or chat. Rooting for you! 🔥

Building an AI-native workspace called Bottari. The vision: It analyzes user data, predicts the next logical backlog item, writes the code, and deploys it autonomously. If I actually pull this off, I’ll finally stop playing catch-up with maintenance and just spend 100% of my time shipping wild new ideas. Miles to go before it’s fully autonomous, but the meta-game of software development is about to change. Let's #BuildInPublic 🚀

This hits different, Ankur. Laying off 80 people with no fault of theirs is brutal, yet it was the single most responsible decision for the organisation’s survival and the business’s future. What followed was pure leadership: founders taking a 50% pay cut first, and three seniors volunteering theirs the very next morning. In an era obsessed with burn and growth-at-all-costs, this is the reminder we all needed: the hardest decisions are often the most humane in the long run. Grateful you shared this. Respect. 🤝

Year 1 of nearbuy - my previous startup. We had 30 days of cash left. We laid off 80 people for no fault of theirs. The founders took 50% pay cuts. The next morning 3 senior mgmt folks came back with salary cuts on paper. None of them work for me now. But I owe them everything.

@Rakshaweb3 This is 🔥 Most creators jump in saying “Here are 10 tips!” like a pushy salesman. But people actually want the opposite: someone who says, “I know exactly how that lonely 2am frustration feels when your content isn’t working.” It’s like visiting a good doctor. You don’t want them to start with “Take these 10 vitamins.” You want them to first name your exact pain, explain why it’s happening, show the one thing that really fixes it… and then say “Come back to me for more” (that’s your CTA). Pain → Why → Fix → Next step.

The X 2026 algo loves threads but not how most people write them. Skip the "let me show you 10 tips" opener Start with a problem your reader already feels Thread that converts: - Pain point - Why it happens - What works instead - CTA to DM or follow That structure = saves + follows.

Exactly. Wealth doesn’t just delete your insecurities, it removes the shared fog that once let everyone pretend theirs didn’t exist. It’s less about becoming ‘out of touch’ and more about finally seeing the game clearly. Curious how this shift has played out for you, Craig, especially coming from the YC/startup world where money arrives fast but the social dynamics lag behind.

wealth eliminates nearly all of your insecurities but exposes everyone else’s

🔥 This is the uncomfortable truth that very few want to say out loud, @LiannaAdams. AI didn’t hollow out those jobs, it just turned on the lights and showed us they were empty the whole time. The real flex isn’t automating tasks. It’s finally designing work around contribution, growth, and dignity. Question is: how do we make “dignity” the KPI that actually matters in 2026 and beyond? Loving this framing 👏

AI exposes an uncomfortable truth: Many jobs were already hollow. Automation doesn’t remove meaning, it reveals where meaning never existed. The opportunity isn’t just efficiency gains. It’s redesigning work around: 🔹 Contribution 🔹 Growth 🔹 Dignity

Tuning agents feels like reverse-engineering your own brain in public, the constant drift, hallucinations, and need for guardrails are exactly how human focus works without systems. It’s humbling (and hilarious) that the best “AI alignment” techniques are basically ancient Stoic discipline + modern productivity hacks.

Interestingly enough spending time with agents helped my understand myself more. Human intelligence is not that different at all: highly error prone, constantly drifting away and requires harnessing through discipline and processes

🔥 This is the cold truth most founders don’t want to hear, Mariya. ‘I need more revenue’ is basically the startup version of ‘I need to eat healthier’ while still ordering pizza every night. Chasing top-line growth without fixing pricing, margins, or customer mix is just making your problems bigger and more expensive.

"I need more revenue" almost never means I need more revenue. It means: my margins are off, my pricing is wrong, my customer mix is dragging, or I haven't looked at my unit economics in six months. More revenue rarely fixes any of those. It usually just makes them more expensive.

Exactly what I was hoping to hear, the eval layer as a controlled on-ramp is smart. Keeps the routing engine stable while still letting you move fast. One quick follow-up: once a new model clears the benchmarks in a workload category, how quickly does the routing engine start shifting real traffic? Do you use something like confidence-based gradual rollout (e.g. 5% → 20% → 50%) with live monitoring, or is there a different mechanism you’ve found works better in production? Really impressed with the architecture discipline here. Keep shipping, Yonah! 🚀

In a hybrid architecture, new frontier models don't immediately become part of the critical routing path. They enter through an evaluation layer where they're benchmarked against existing systems across reasoning quality, latency, context retention, consistency, and cost efficiency. So when a stronger model emerges, the question isn't "How do we rebuild the stack around it?" but rather "Which workloads does it outperform on?" If it demonstrates superior performance in a given category, the routing engine can gradually allocate traffic to it while continuously monitoring outcomes against existing models.

Everyone keeps asking where AI is heading next. Bigger models? More parameters? More compute? Honestly, I think the future of AI is going to look very different from what most people expect. And that’s one of the biggest reasons we started building Aidove’s Hybrid LLM Architecture. A thread 🧵

Spot on, @sabeer! Indian society treats knowledge like a sacred inheritance, memorize the Vedas, respect the guru, and you’re powerful. But the most innovative societies treat curiosity like oxygen. They don’t just guard what’s known; they itch to poke the unknown until it reveals something new. It’s the difference between a beautifully organized library (knowledge) and a group of mad scientists in the basement who keep blowing things up to see what happens next (curiosity). One preserves civilization. The other creates the future. Ancient India actually had both, the same rishis who compiled knowledge were obsessed with “jigyasa” (inquiry). We somehow lost the second half of the recipe along the way. Time to bring the mad scientists back. What do you think flipped the switch from curiosity to rote reverence? 🔥

The simplest way to describe Indian society: Knowledge is power. The simplest way to describe many of the world's most innovative societies: Curiosity is power. Knowledge tells you what is known. Curiosity helps you discover what isn't.

🔥 This is gold, Stacy! So many entrepreneurs chase the dopamine hit of “going viral” while their actual customers quietly churn. Retention isn’t sexy on the timeline… but it’s the only thing that pays the bills in year 3, 5, and 10. Loyalty compounds harder than any algorithm. Your small-but-loyal customer base is the flex. Respect.

One of the biggest mistakes entrepreneurs make is building for attention instead of retention. A loyal customer base is more powerful than temporary virality.

@pirwot 100% 🔥 Those uncomfortable convos are exactly where the magic (and the progress) happens. Startup leadership isn’t for the cozy zone!

Want to lead a startup? Learn to love uncomfortable conversations. Progress happens on the edge of your comfort zone.

@itskaveriirl Have a short, 1 week trip to some place in Uttarakhand. You will remember it forever.