akritiiii

Building iOS, Android & Web Apps | Building MVPs | UI/UX Design & Branding | Turning ideas into products | DM to build yours

[akritiiii @buildyourapp_]

Me after clicking “accept” for the 100th time without reading a word of what claude is doing

[akritiiii @buildyourapp_]

Dear windows kindly fix this

[akritiiii @buildyourapp_]

LLMs ranked for vibe coding (based on my experience building projects):

>Claude Code
>Cursor
>Copilot
>Qoder
>Codex

What’s your ranking? Drop yours below 👇

[akritiiii @buildyourapp_]

With every passing year, this tweet gets more real.

[akritiiii @buildyourapp_]

With every passing year, this tweet gets more real.

[primordial intelligence @primal_brainer]

you’re not an animal lover if you eat meat - just say you like cats and dogs and go.

[akritiiii @buildyourapp_]

koi team kitna bhi bura khel le 50
to bna hi leti bai

[akritiiii @buildyourapp_]

Good morning ☀️

[akritiiii @buildyourapp_]

The moment you reach Claude code limits

[akritiiii @buildyourapp_]

The moment you reach Claude code limits

[akritiiii @buildyourapp_]

A guy built a tool that sends a design screenshot to Claude, GPT-4o, and Gemini at the same time and then scores which one rebuilt it most accurately.

You upload a screenshot, select the models, and hit generate.
All models run in parallel.

Once they finish, each output is compared to the original using pixel-by-pixel matching.
You can:

compare outputs side by side

use an overlay diff slider to spot differences

test Tailwind or plain CSS outputs

https://t.co/DSx3DK8nC3

[akritiiii @buildyourapp_]

VIBE CODERS: stop overcomplicating your stack

Here’s what to use instead:

Databases: Prisma + Postgres. Manual SQL = silent suffering.

Forms: React Hook Form + Zod. Validation bugs will haunt you.

Payments: Stripe or Polar. Never touch PCI compliance.

Search: Algolia or Typesense. It's harder than it looks.

Backend: Serverless + BaaS first. Scale later, survive now.

Error tracking: Sentry or LogRocket. Console.log isn't observability.

Analytics: PostHog or Plausible. You're flying blind otherwise.

UI: shadcn/ui or Radix. Consistency beats creativity at MVP.

Configs: env + dotenv. Hardcoding = instant regret.

File uploads: UploadThing or Cloudinary. Multipart hell is real.

CI/CD: GitHub Actions + Preview Deploys. Future-you will thank you.

Performance: Lighthouse + Vercel Analytics. Slow apps don't convert.

Onboarding: Add empty states. UX beats features every time.

Folders: Modularize early. Refactors cost 10x later.

Docs: Write your README now. Your memory will betray you.

Bookmark to ship better.

[akritiiii @buildyourapp_]

I tested every feature inside Claude Cowork (desktop app).

Here’s what most people are missing 👇

If you’re still using Claude only in the browser…
you’re missing the biggest shift in AI workflows this year.
This isn’t just a UI change.

It turns Claude from a chatbot → into an actual operator.

What stood out:

1/ It works directly on your machine

reads, edits, creates files locally
no more uploading files one by one

2/ It handles big tasks properly
breaks work into smaller chunks
runs parallel sub-agents
no timeouts on long jobs

3/ It outputs real deliverables

Excel, PDFs, PPTs not just text
and can chain them together
Where it gets powerful:

4/ Global + folder-level instructions

different rules for different directories
acts differently based on context

5/ Connectors (huge unlock)
Slack, Drive, Notion, Gmail, Figma
pulls real data without copy-paste

6/ Plugins + Skills

reusable workflows for marketing, sales, finance, etc.

7/ Scheduled tasks
automate recurring work (reports, summaries, cleanup)

Best use cases I found:

organizing messy folders
turning raw notes → structured docs
analyzing multiple files at once
generating reports across tools
automating weekly workflows
Biggest mistake people make:

They use it for generation.

But the real value is automation + maintenance.

[akritiiii @buildyourapp_]

Someone built an MCP server that connects ChatGPT, Claude, Gemini and Perplexity all inside your IDE at the same time

no tab switching
no copy pasting
no API keys
it’s called Proxima
it connects to your existing accounts and runs all 4 models together locally, then plugs into VS Code, Cursor, Claude Desktop, etc.

what’s interesting:

you can ask all 4 the same question and compare answers
perplexity handles search while claude writes code and gemini reviews it
smart router picks the best model automatically
fallback system switches if one model fails

also comes with:

40+ MCP tools (search, code, debugging, analysis)
local API compatible with OpenAI SDKs
simple Python / JS integration

[akritiiii @buildyourapp_]

github: https://t.co/p003ciimRl

[akritiiii @buildyourapp_]

Someone built an MCP server that connects ChatGPT, Claude, Gemini and Perplexity all inside your IDE at the same time

no tab switching
no copy pasting
no API keys
it’s called Proxima
it connects to your existing accounts and runs all 4 models together locally, then plugs into VS Code, Cursor, Claude Desktop, etc.

what’s interesting:

you can ask all 4 the same question and compare answers
perplexity handles search while claude writes code and gemini reviews it
smart router picks the best model automatically
fallback system switches if one model fails

also comes with:

40+ MCP tools (search, code, debugging, analysis)
local API compatible with OpenAI SDKs
simple Python / JS integration

[akritiiii @buildyourapp_]

This guy went from 0 → 33,003 LinkedIn followers in 6 months

Captured 10,965 leads

And sent 0 cold DMs

Most people post on LinkedIn like this:

random topics
random formats
hope it works
He did the opposite.
Built a system 👇

Step 1: Stop guessing content

Instead of thinking:

“what should I post?���

He analyzed:

real conversations
repeated problems
questions people are already asking

Step 2: Use lead magnets (not just posts)

create something valuable (template, checklist, etc.)
ask people to comment a keyword
send it via DM
This drives:
engagement + leads at the same time

Step 3: Automate everything

When posts get:
300+ comments
1000+ comments
Manual DMs don’t scale.

So he built a system that:

detects keywords
sends instant personalized DMs
stores every lead in a CRM

Results:

33k followers
10k+ leads
5–10 inbound leads every week

[akritiiii @buildyourapp_]

This guy went from 0 → 33,003 LinkedIn followers in 6 months

Captured 10,965 leads

And sent 0 cold DMs

Most people post on LinkedIn like this:

random topics
random formats
hope it works
He did the opposite.
Built a system 👇

Step 1: Stop guessing content

Instead of thinking:

“what should I post?”

He analyzed:

real conversations
repeated problems
questions people are already asking

Step 2: Use lead magnets (not just posts)

create something valuable (template, checklist, etc.)
ask people to comment a keyword
send it via DM
This drives:
engagement + leads at the same time

Step 3: Automate everything

When posts get:
300+ comments
1000+ comments
Manual DMs don’t scale.

So he built a system that:

detects keywords
sends instant personalized DMs
stores every lead in a CRM

Results:

33k followers
10k+ leads
5–10 inbound leads every week

[akritiiii @buildyourapp_]

PSA: Claude is giving free credits to Pro/Max users.

If you’re a paid user, check your email 👇

search for an email from Anthropic / Claude (check spam/promotions too)
click the redemption link inside

Important:

the credit is NOT applied automatically
you must claim it manually
After redeeming:

go to Settings → Usage
turn ON “Additional Usage”

Deadline:
April 17, 2026

Don’t miss free credits.

[akritiiii @buildyourapp_]

Same prompt. Same instructions. Same review role.

But the outputs? completely different.

Claude vs Codex… and at this point Anthropic just wants to lose users.