full-stack dev · turning ideas into launched products · DMs open for work

127.0.0.1
Joined August 2022
sankit retweeted
Article on Perfection Chase Perfection is a title, please read and know the actual purpose belongs to it :)
6
8
23
410
sankit retweeted
Do not bookmark this Opus 5.5 + Remotion
2
4
97
sankit retweeted
you finish a set, grab your phone, and somehow logging it feels like another workout. too many taps. missing the stuff i actually wanted. so i built my own. every log is two taps. twotap. android, coming soon. waitlist: sankitbuilds.xyz/twotap
i kept quitting gym apps so i am building my own join the waitlist: sankitbuilds.xyz/twotap first look 👀
6
4
1
19
535
sankit retweeted
nowadays banks are giving false promises to just get you onboard. I got message from @IDFCFIRSTBank on whatsapp that if i open a saving account i will get 10k reward points. So, i did. I didn't receive any reward point. Now coming to credit card reward points. I got some reward points so i was checking if i can redeem it. Then i found out to redeem it i have to pay 99 Rs + GST. what i don’t understand is, i earned reward because i used the credit card. Now to reddem it why charging this much amount. Another thing is if i buy same gift card with money no charge is there 🤡. @IDFCFIRSTBank keep ur promise else don't make it. And let people redeem for free, its their right. Or charge based on redeem and less.
14
4
1
74
14,297
sankit retweeted
Today, I'm launching llmdojo.dev: - The hands-on way to learn AI engineering: LLMs, RAG, agents, evals and inference. - 50+ interactive labs. Tweak tokenizers, attention, embeddings and KV caches live. - Train a neural network in your browser and watch it learn. - Build agentic workflows with tool calls, memory, guardrails and bounded autonomy. - Ship RAG that works: chunking, hybrid search, reranking, and LLM-as-a-judge evals. - Production AI: model routing, fallbacks, quantization, caching and cost per request. - 490+ interview questions, reported from OpenAI, Anthropic, Google, Meta, Amazon, Microsoft, Mistral and 50+ more. - Every answer is cited from papers and books, or measured on real models.
207
180
30
1,769
84,377
sankit retweeted
Video Streaming High Level Design. [Some things are not explicitly written but should be inferable]
19
33
1
526
13,221
sankit retweeted
introducing gettapmac.com your MacBook already feels every tap. now it listens. → tap once and talk, it types for you → works in any app, wherever your cursor is → Unlimited AI fixes or explains the text you select → your clipboard history and AI chat, a tap away → private, all on your Mac, no account, no subscription no shortcuts to remember, just knock meet TapMac 👊
136
43
15
318
27,528
sankit retweeted
Today, I'm launching V2 of designeer.xyz with a completely reimagined UI and a few things I've been wanting to add for a while. - Discover visual work from independent creators and studios across interfaces, motion, typography and experimental digital art - Save your go-to tools to Favourites for quick access - Create folders to organise and keep track of everything you discover - Get your product in front of 100K+ unique visitors a month with featured listings starting at $49 Bringing the best of the internet together, now with a whole new experience.
95
60
7
682
56,924
It's here! Today I'm officially launching Yomu, an app that makes your daily reading easy. - Bring your own blogs and articles and read them inside Yomu - Uses your local agents (Codex, Claude, OpenCode, etc.) so you can ask any question mid-reading - Bookmark and save an article to the Yomu folder in your browser, and it automatically appears in the app -Upload any research paper PDF and get a stable UI for long reading sessions -Customize the theme and make it look the way you like And many more! (yomu.lowkeydev.me)
27
29
6
77
2,933
sankit retweeted
my new site is live :) i was supposed to make a portfolio but my mom says i keep too much junk. so i put all of it on my website. link in the replies 👇
8
4
1
29
17,894
sankit retweeted
Being honest here !! >So, I was unemployed for almost three months. Initially, things were fine. I wasn’t too stressed because I thought I’d find a new job pretty quickly. I started applying rigorously on LinkedIn, Wellfound, X, and pretty much anywhere I could find a relevant opening. But I wasn’t getting any leads. >As the first month passed, a thought slowly started growing in my mind: Why am I not getting a job? Am I really not good enough? >My entire day became about applying for jobs, sending DMs to recruiters, following up, and repeating the same process. Eventually, I started losing interest in it. Slowly, I became inconsistent with building things, learning, and even doing the stuff I genuinely enjoyed. >I kept telling myself that I couldn’t slow down and that I had to keep going. Eventually, I got an interview opportunity, but I wasn’t selected. Then I had another one, and the same thing happened. >Looking back, the problem was that I simply wasn’t prepared enough. When I finally got another opportunity, I prepared properly, gave it my best, and eventually got selected. >But I have to admit something I personally felt during those three months: when you aren’t involved in any kind of work for a long period of time, it slowly starts affecting you. You begin to feel low, demotivated, and sometimes even start questioning yourself. >It made me realize that having a job isn’t just about earning money. For me, being involved in meaningful work also gives me a sense of purpose, momentum, and confidence.
Joined a MNC as an AIML Trainee Engineer !
22
9
185
5,733
sankit retweeted
Introducing taasant. taasant turns a telegram chat into an encrypted drive you can use from the terminal. taasant upload holiday.jpg taasant list taasant download hol taasant delete hol that's the whole interface. underneath, taasant handles the annoying parts. telegram bots can upload files up to 50 mb, but downloads are limited to 20 mb. so taasant splits every file into 19 mb chunks. each chunk is encrypted locally with aes-256-gcm before it leaves your machine, then uploaded as a separate telegram message. a small encrypted manifest keeps track of the chunks and their order. when you download a file, taasant fetches the manifest, downloads the chunks, decrypts them, puts everything back together, and writes the original file to disk. you don't have to care about any of this. what you get: - encryption happens before upload. telegram only sees encrypted bytes, and the key never leaves your machine - no practical file size limit. a 2 gb file is just more chunks parallel uploads for large files - a local index with fuzzy filename matching. "hjp" can find "holiday.jpg" - ambiguous matches are shown instead of guessing - downloads never overwrite existing files - the encrypted index is backed up to telegram after every change - one small go binary with zero external dependencies setup takes about five minutes: create a bot with botfather, get your chat id, generate a key, and you're ready. i started working on this after seeing spacefs and wondering whether the same idea could work with telegram as the storage layer. telegram already gives you file storage and access across devices, so i wanted to see how far i could push that. taasant is the cli i've built around it. the bigger part i'm working on now is bringing the same storage layer to mac as a virtual disk. the goal is for it to show up as a normal folder in finder. drag files in, open them, rename them, move them between folders, and let the storage layer handle everything underneath. files stay encrypted and live in a private telegram channel. rclone handles the mac side, including local caching, so recently used files don't have to be downloaded every time you open them. older files are evicted as the cache fills up. there's no kernel extension involved. the file index also lives in the channel, so setting up a new machine can restore the drive. taasant is out now.
19
20
3
97
5,864
sankit retweeted
Been learning more about LLM inference while working on my LFX project One thing that surprised me that, more compute is not always the ans, decode is often limited by memory bandwidth A lot of the work is moving model weights and Kv cache data from HBM while generating tokens Makes understand why KV cache management, continuous batching and speculative decoding matter a lot LLM inference is way more systems engineering than I thought xd
20
2
1
55
1,160
One piece of advice move abroad when your monthly salary in India reaches ₹1.5-2 lakh.
141
82
32
4,655
496,001
sankit retweeted
We have solved video editing once and for all and…. It’s FREE Available on your nearest Mac. PS this video was made on @usecardboard ! Now, LET IT RIP!
259
174
52
1,992
326,143
sankit retweeted
Introducing Rewamp UI My first open-source UI library. 70+ ready-to-use components built for modern websites. From shaders and animations to navbars, buttons, cards, and more everything you need, all in one place. Completely free. Go check it out : rewampui.com Drop your honest reviews below.
111
67
15
708
143,086
sankit retweeted
be careful in your pursuit of cheap tokens we've come across so many sketchy things that are out there with tons of usage remember you're hooking up your harness to these inference providers. they can and do crazy things like send fake tool calls to steal your info they also sell entire traces to third parties who resell again. a single key in the trace is enough to screw you if you cannot explain how someone can offer cheap tokens you should avoid it
187
271
61
4,503
282,391
Everyone says learn inference, learn CUDA,learn ML But almost nobody tells you exactly what to learn, how to learn it, or how to know whether you’ve actually learned it. Introducing Trentorch.com a free, open-source platform to learn and practice Machine Learning, CUDA, and inference through hands-on coding. Instead of just watching lectures or reading theory, you’ll: • Learn the theory behind each concept • Follow step-by-step implementation examples • Solve coding challenges based on what you learned • Implement algorithms from scratch, from foundational ML to Inference and CUDA kernel • Practice problems in a Codeforces-style environment • Get new Problem of the Day challenges with ratings The goal isn't just to teach you how to write the code. It’s to help you understand what your code is actually doing underneath. We’re building Trentorch because advanced education shouldn’t be locked behind expensive subscriptions. Money shouldn’t be the barrier to learning ML, inference, or CUDA. A free alternative to platforms that charge students heavily every month. signup and start solving problems now Trentorch.com
47
93
18
666
104,861
Website inspiration by section... - Navbar; navbar.gallery - Hero; supahero.io - CTA; cta.gallery - Footer; footer.design - 404; 404s.design What am I missing? 🔖
17
99
3
854
61,639