Piyush

@VanpeltVentures Congrats Van 🤝🏻 Keep pushing 💪🏻

A few months ago I had 0 follower, 0 project and now I sit at almost 500 followers and 3 projects out (soon 4 !) YouTube, instagram and TikTok are picking up with a combined following of around 200 ! What an incredible journey so far ! All thanks to the amazing support from X 🫶

here's what nobody tells you about AI outputs... the model isn't your problem your inputs are people write "AI, write about [topic]" result is always generic wikipedia summary what actually works: "here are 3 examples of good, generate 50 variations" result: 50 testable options built on proven patterns the difference: feeding in the context for what you want as output is the differentiator generic context alone = 1 output quality context + examples = 10 quality outputs context + examples + your angle = 50 outputs that sound like you show it what good looks like

most people write prompts like instructions... prompts are systems design here's how to build one: > define your input layer - what are you feeding it? (examples, observations, data) define your processing layer - what transformation happens? (extraction, generation, refinement) > define your output layer - what format do you need? (post drafts, building blocks, analysis) then build separate prompts for each layer

you're making one BIG mistake when writing prompts... you're trying to remove yourself from the equation great prompts don't replace you, they multiply your leverage AI can generate prompts but it can't generate prompts that contain what only you know about your domain here's how to write prompts AI can't replicate: build your domain expertise into the prompt structure not placeholders, but actual dependency on what you know identify where your knowledge transforms output from good to exceptional integrate your context so deep that removing you breaks the system you are the center of the equation that's what makes your prompts valuable

here's how to read a 300-page book in 30 minutes without missing anything important... > upload the book PDF to notebooklm > add 2-3 youtube videos on the same topic > add a few articles and generate a podcast what you get is: → core concepts synthesized → cross-referenced insights → and actionable takeaways you can process 3-4 books a week now this isn't about reading faster it's about extracting what actually matters and skipping what doesn't

@EXM7777 It's really a good tool for reading books and research papers Just feed in the PDFs, docs, links of videos and upload them and make a podcast to listen to as brief summary I use this feature all the time to learn anything in detail under 30-40 mins

everyone is sleeping on NotebookLM... this thing is literally the best option for: - building a custom AI trained on whatever topic you need (content creation, copywriting, automations...) - learning complex topics, studying at any level - understanding RAG, working with large files from any source plus it's powered by the second best model itw and the #1 platform for video education

ai rewards preparation + context more than prompting... i spent 2 hours building a swipe file now every prompt takes 30 seconds and outputs are 10x better the system: → save 50+ examples (tweets, emails, landing pages that performed) → feed to AI: "analyze patterns here" → use those patterns as context for every generation result: one high-performing tweet → 50 angle variations one email hook that got replies → 20 subject lines one landing page that converted → 10 structure tests preparation takes 10 minutes and generation takes 30 seconds people skip prep, then complain outputs are generic do the prep work and feed the proper context first

here's why your prompts aren't delivering... you're missing the step that comes BEFORE writing anything it's called context engineering: loading the model with data before you ask it to create most people skip this because it requires an extra step do the extra work and you'll completely change your experience with LLMs: you want great UI? feed gemini actual landing pages from stripe, linear, notion you want cinematic images? feed nano banana frames from blade runner, interstellar, whatever you want emails that convert? feed kimi the best hooks from recent campaigns the model doesn't magically know what "good" looks like in YOUR niche you have to show it that's the 10x improvement everyone misses (don't be that person, do the prep work)

@Mho_23 MENTOR

i built a tool that allows me to clone the brain of any youtuber… i used it on alex hormozi so i can ask him business questions at 3am and get answers word-for-word like he'd give them works with anyone: gary vee, naval, graham stephan, any creator in your niche i put together: - the tool that uploads entire channels automatically - my 5-minute setup process - prompt framework for answers that match their exact style RT + reply 'MENTOR' and i'll send you everything (must follow so i can dm)

Ship more than you study. Build more than you research. Close more than you perfect. Deliver more than you hesitate. Start more than you plan. Ask more than you assume.

i think i finally figured out why ai never helped me with post ideas. i was burnt out from manually brainstorming for ideas, so i tried letting ai do it for me. it turned out to be a big mistake. so what i did wrong was: i asked, “give me 20 ai automation post ideas based on my previous posts.” the output was: → generic lists → felt fake → zero saves, zero replies then i dug deeper and realized i should provide it a knowledge base with full context and examples. so i created 4 knowledge base docs: > summarizing top 10 videos on writing high-quality tweets > giving context about my business, myself, and topics i want to cover > viral posts from the biggest accounts in my niche > examples of my posts; i asked it to break down my tone, hooks, and structure (saved as context) then i gave it my real experiences and told it to write in that exact style, plus a meta prompt to reference them all with instructions and rules. suddenly, the ideas felt like mine. rough, honest, and value-first. lesson learnt: ai can’t invent my experience; it can only amplify what i already ship. using ai is like giving it as many puzzle pieces as possible for the output you want you’ll always get golden results.

@ralstoncrasta absolutely lfg 🚀 it's crush it now or never moment

@piyushorb bros locked in and cooking hard!!! lets go G

just vibecoded AgentOS a dashboard in ~12hrs straight built for my inbound voice agent service clients can track calls, performance, and progress and message support directly — it pings my slack instantly

most founders aren’t broke because of bad luck. they’re broke because they stay comfortable. yesterday i proved that to myself. i walked into one of the richest neighborhoods in the city and started talking to the people who lived there. not rich. FUCK‑YOU money rich. every house was a mansion. BMWs and AUDIs parked like normal cars. then i saw it. a ridiculous mansion. outside it...a LAMBORGHINI HURACÁN TECNICA. in navi mumbai. most people would take a photo, fantasize, and leave. i didn’t. a thought hit me: nothing changes if nothing changes. so i did something uncomfortable. i cold‑approached an older guy standing outside. jacked. calm. overwhelming presence. (i was in a sleeveless tee. good physique. you can’t fake that kind of discipline.) “is this your house?” “yeah.” pause. “what do you do for a living?” he said: “nothing.” then stayed silent. for 4 seconds. in that silence, i felt the real gap. not money. time. failures. reps. perspective. then the woman with him asked me: “do you know ___?” (one of the biggest restaurant chains here) i said yes. “he owns that.” my brain shut off. before leaving i said, half nervous, half honest: “sir, i run a business too. i’ll get there one day.” he didn’t motivate me. didn’t doubt me. just said, unfazed: “yeah yeah, you can do it.” later, i asked security at another insane mansion. the lambo one. “what does he do?” “builder.” that walk made a few things painfully clear. people say fitness gives confidence. that’s not true. fitness is visible discipline. it’s social proof you can’t lie about. it doesn’t make you better... it just gives people a reason to take you seriously before you speak. business works the same way. founders say: “people ignore me.” “clients don’t reply.” “nobody takes me seriously.” it’s usually not your age. it’s not the algorithm. it’s not bad luck. it’s simpler than that. you haven’t built enough proof yet. proof isn’t followers. proof isn’t fancy words. proof isn’t ai slop. proof is: showing up consistently when nobody’s watching documenting real work, not highlights letting people spend time with your thinking doing it long enough that trust compounds that’s literally why i’m building Tiji AI. we help founders turn their work, thinking, and journey into content that builds trust before the first conversation. not hype. not virality. just enough proof that when you speak, people listen. nobody gave me advice yesterday. nobody taught me a lesson. i just stood close to what’s possible. and once you do that, your internal ceiling quietly disappears.

@almoststarted @ayroids recently launched this product to help validate ideas

@piyushorb Thank you for sharing . Did it help your product ?

I don't know why Every time I build something No one uses that 😭

Google AI Studio is crazy good: - Free - Unlimited usage of Gemini Pro, I guess - Great for vibecoding

my simple prompting process - stop trying to write the “perfect” prompt. start vague, let it hallucinate, then add one change at a time. add constraints, examples, system tweaks. break it till it’s minimal but still works. it feels less like prompting, more like debugging.

planning things out beforehand only leaves you with execution