Devdeep retweeted
Still hand picking UI components?
Now no need to do it .
Check out Avara UI.
@Pranto39
I stopped hand-picking UI components.
Now my AI agent builds the page for me, using redesigned @shadcn /ui-based components, through MCP.
Meet AvaraUi.com : → 189 components → 24 website sections & app blocks → AI agents MCP support → Built on shadcn/ui
100% free. Try it now AvaraUi.com 👈
Devdeep retweeted
Asked at Microsoft, Tesla, Waymo 🤔
Your team says the new model is better because it scored higher on a benchmark.
How do you decide if it's actually better for your product?
I found out what my X account is actually worth.
Up to $109/month. 🤯
Honestly, I still can't believe it 👀
creators.unfair.so/worth/_de…
Need advice from senior backend developers
I'm a Node.js developer and thinking about my next step.
Should I learn NestJS, or focus on mastering Node.js
If you were in my position, what would you recommend learning next?
Would love to hear your advice.
Devdeep retweeted
My X account is worth $29.5K, yet I’m STILL not verified on unfair.so😭
At this point, the only thing truly unfair is Unfair itself 💀
@notmiramurati WHAT ELSE DO I NEED TO DO TO GET VERIFIED? 👀
Someone tag the team until they notice this. Let’s see if unfair.so can finally prove its name wrong. 🥲🔥
who wants to be a millionaire?
i actually want to know if your account on X is worth $1M
go check this out myaccworth.com
Devdeep retweeted
Asked at Scale AI, Stripe 🤔
Your model works great offline but fails on real user inputs in production.
How do you detect and handle distribution shift?
Replying to @nikks_techie
Start by logging feature stats and prediction confidence in production. A sudden drift in those distributions or a confidence drop flags a shift before accuracy tanks.
Devdeep retweeted
Interviewer:
If an LLM can answer questions about the future, does that mean it can predict the future?
Devdeep retweeted
Asked at Microsoft, Databricks, Glean & Palantir 🤔
You’re deploying a 70B-parameter LLM in production.
How much GPU memory do you actually need to serve it?
Replying to @nikks_techie
Roughly 140 GB of VRAM if you keep fp16, but most teams shard across 8 to 12 A100 40 GB cards or use 4-bit quantization to drop it below 80 GB.
Replying to @SakshiSugandhi
AI can speed up the build, but the deep understanding you gained lets you troubleshoot, design, and innovate where the model can’t. Your years still matter.
Devdeep retweeted
bro asked claude for an office and got a whole GTA side quest 💀
I asked Claude to build an office for the @tavus PALs...
I was not expecting it to build THAT kind of office













