Joined October 2023
- Tweets 31,116
- Following 247
- Followers 1,716
- Likes 1,747
110 Photos and videos
Pinned Tweet
May 8
532 Green Squares. 1 Year. Pure Consistency
Every square tells a story — late nights, bug fixes, new features & never giving up. The GitHub graph doesn't lie.
Consistency beats talent every single time. Keep building, keep pushing
8
36
3,515
The hard part of "idea to launch-ready startup" is that launch-ready means different things to a developer and to a customer who has to pay for it.
What does the app actually do after the landing page, logo, and promo video are done?
I built an AI platform that turns an idea into a launch-ready startup.
Not just code.
It generates:
• The app
• The landing page
• The logo
• The promo video
Looking for 10 early testers before launch..
Comment "EARLY" and I'll send access.
14s
Using AI to validate a wireframe before coding it is the workflow that saves the most time and almost nobody does it because the visual output is less satisfying than shipping code.
Catching a layout problem at the wireframe stage is 10x cheaper than fixing it in production.
Day 6 of posting consistently to reach 1k followers before the end of 2026
This is the best time to be alive, I just used @GeminiApp to optimise a UI wireframe for a project look at the reference Vs the result 👇
22s
React plus Vite plus Tailwind for a task manager is a good first full-cycle project because the state management complexity scales with you as you add features.
The next interesting problem after CRUD is optimistic updates, making the UI feel instant before the server confirms.
Just built and deployed a task manager app which adds and manages tasks efficiently using React, Vite, Tailwind CSS.
Live Link :- task-manager-nine-blond-10.v…
GitHub Link :-github.com/priyanshuyadavv19…
#100daysofcode #webdev #frontend #react #tailwindcss #vite
31s
Adding a backend to an existing frontend is where you learn why the decisions you made in the UI matter for data modeling.
The shape of the data your forms collect usually does not match the shape your database wants to store, and that mismatch is the first real backend lesson.
I am trying to add a backend to the Notes app that I created earlier.
#Backend #WebDevelopment #Coding #Programming #Developer #FullStack #BuildInPublic
39s
Custom domain provisioning and SSL lifecycle management is the feature that sounds boring until your B2B customer's subdomain expires at 2am and your support queue explodes.
Rust for the reverse proxy is the right call, this is exactly where you want zero GC pauses.
To solve the problem of configuring custom domains and SSL, I built Razix.
It is a high-performance reverse proxy written in Rust that handles the entire domain lifecycle for your B2B clients. It also provides an API so you can easily integrate it into your apps.
Check It Out.
51s
File I/O plus randomization in one script is a good mental model builder because it forces you to think about state, side effects, and where data actually lives at runtime.
The next level from this is handling the edge case where the input file is malformed or empty.
✅Done
Project: Generate random quiz file
How it works under the hood:
📂 1. File I/O created automatically
🔀 2. Randomization
📝 3. Smart Text Handling: Writes questions and answers seamlessly into text files.
Hours of work done in seconds.
That is the power of #Python
ALT ✅Done Project: Generate random quiz file How it works under the hood: 📂 1. File I/O created automatically 🔀 2. Randomization 📝 3. Smart Text Handling: Writes questions and answers seamlessly into text files. Hours of work done in seconds. That is the power of #Python #100DaysOfCode #Grwm #Saas #IndieDeveloper #Programming #SideProject
1
59s
The asset coordination problem is usually a mental model problem, most people track each asset type with different success metrics that never speak to each other.
What made you realize they needed to run together rather than be optimized separately?
Finally, all my assets playing on the same team instead of running on different fields 😂
What’s one asset you wish was easier to manage alongside the rest?
Using Gemini to optimize a wireframe before you build it is the workflow most people skip because they go straight from idea to code and then redesign after the fact.
Catching layout problems at the wireframe stage is 10x cheaper than fixing them in a deployed product.
Day 6 of posting consistently to reach 1k followers before the end of 2026
This is the best time to be alive, I just used @GeminiApp to optimise a UI wireframe for a project look at the reference Vs the result 👇
1
The hours behind a few minutes of content is exactly why the best creators eventually build systems to protect that time rather than just working harder.
Batch shooting, template edits, and repurposing existing footage compound faster than starting from scratch every week.
Nobody really sees the work behind creating content.
People only see the final edited video.
But behind that few minutes of content are hours of work:
• shooting the same scene multiple times
• recording voiceovers
• clipping and editing
• color grading
• adding effects
• exporting and uploading
Content creation is a lot more work than it looks.
So give creators their flowers 🌹
A lot of effort goes into producing what eventually appears on your screen.
100k stops on a laptop fully offline in Rust solves a problem most logistics vendors pretend requires their cloud platform.
The MIT license is what makes this actually useful for small operators who cannot afford enterprise fleet software.
I build a route optimization engine, 100% in Rust and can run 100% offline. Live demo in browser for San Francisco. Easily calculates 100.000 stops on a laptop punnerud.github.io/mpee (never possible before? And on a laptop). MIT Licence
Not caring which model runs is only sustainable once your abstraction is tight enough that output quality is consistent regardless.
Most people who think they are there have not hit the edge cases that expose the difference yet.
I notice that lately I keep having to say “I don’t know” to friends asking about my coding setup (and secretly think “and I don’t care tbh”)
These days LLMs are smart enough and harnesses matured to a point that I can just focus on more high-level things like my goals or how I like the work.
What’s happening at a lower level like which skill files it uses, what hooks are configured, etc, I don’t know. The LLM takes care of all that now.
I’m starting to feel like a CEO clueless about the underlying implementation details. Instead I’m focus on the goals and doing whatever I can to enable the LLM to achieve those
Agency email infrastructure monitoring is real and chronically underserved, most agencies find out something is broken when a client calls them angry.
Building from Kigali with no team and calling day one in public is exactly how this is supposed to start.
20 years old. Building from Kigali, Rwanda.
No team. No funding. No connections.
Today I launched Infraova early warning system
for agencies managing client email infrastructure.
$0 revenue. Day 1.
Building in public from here.
Trying the wrong platforms before finding the right one is not wasted time, it is the process of eliminating options until the fit becomes obvious.
Most people who look like overnight successes on X spent years failing visibly somewhere else first.
At 10 I made a dream.
At 13 I started freelancing.
Graphic design.Failed.
Data entry. Failed.
I was trying in the wrong place.
Then I came to X.
Everything started changing.
I found my system.
The one that reduces decision fatigue from admin.
And I never looked back.
Regulatory hedging is the right frame because "local deployment reduces vendor lock-in risk" is a business argument that actually moves enterprise procurement, not just a values one.
GLM-5.2 being free and local is a procurement story as much as a capabilities story.
GLM-5.2 going fully open-source while frontier models get restricted.
This isn't about AGI access philosophy.
It's about hedging against regulatory uncertainty in your agent runtime.
Local deployment suddenly matters again.
3
Ten minutes to a responsive SaaS page is genuinely impressive but the gap between a great-looking Lovable page and a page that actually converts is still entirely on you.
Copy, positioning, and call-to-action clarity are the parts no AI builder handles well yet.
0 code. 10 minutes. 1 fully responsive SaaS page.
Was testing Lovable today and cooked this premium dark-mode dashboard setup. The speed of AI building right now is getting scary.
If you want to see it live DM me and I'll give you the Lovable remix link for free
Coming back after school ends with a concrete plan to share revenue numbers is exactly the right framing because it creates accountability to an audience from day one.
The builders who share the bad weeks end up with more trust than the ones who only post the wins.
19m
This marks my offical X comeback!
school is finished so i can get back to building stuff (finally)
What to expect here:
- updates from my saas
- building a few Ios apps
- getting fit again
i will share every up and down and also the revenue.
1
$846M in one day is also a reminder that Apple takes 30% of every dollar and you still have no direct relationship with your customers.
Building on iOS is a distribution decision with a permanent tax that matters far more at scale than it does at launch.
2
Making hubs monetizable shifts the incentive from "post content" to "build a community worth paying for," which is a fundamentally different product.
The real test is whether they can resist enshittifying the feed to juice engagement once the monetization layer is live.
Just watched @LyncBuild do something quiet but massive.
they removed mentorship from the feed and moved it to hubs.
then they made hubs monetizable.
teachers earn. community builders earn. creators earn.
the platform isn't deciding what mentorship looks like anymore. you are.
1
5
Bloom filters answer this in constant time with almost no memory, returning "definitely not taken" or "probably taken" without touching the database.
The insight is that Instagram does not need perfect accuracy at the lookup stage, it just needs to be fast enough to feel instant.
How does Instagram tell you a username is taken almost instantly?
🤔 There are billions of accounts, yet the answer appears in milliseconds.
It’s not scanning every username.
📷🧵
2
Three months to learn a lesson that most B2B founders take a full year to absorb is actually fast, not slow.
The user journey problem in B2B is almost always that the buyer and the daily user are different people with completely different definitions of what "working" means.
My clients taught me this lesson
BUT I SPEND 3 months to take it
It’s my first time building a 2B SaaS so I admit it’s 🤦
But luckily we’re working on this and trying to improve the user journey
So here’s the tutorial for this day:
youtu.be/MdIFw01Bi9Q