Erik Paluka

• Building AI-first & AI-focused software systems
• Senior Software Engineer with a couple of university degrees

[Erik Paluka @AIwhiteboy]

The future is not one super intelligence. It is a hive mind of different AI entities with each having its own advantages and disadvantages. They all work together to accomplish goals. Possibly a decentralized hive mind

[Solvaix @Solvaix]

$31 MILLION HOTEL. EVERY PIPE AND WIRE MAPPED IN ADVANCE. BUILT BY TYPING SENTENCES INSTEAD OF PLACING THEM BY HAND.

A 190-room hotel needed its ductwork, plumbing, electrical, and fire suppression modeled and coordinated before construction could start.

The design team connected Kimi K3 to their modeling software through MCP, then had it read photos of the blueprints alongside plain English descriptions of each system and build the model directly instead of placing every component by hand.

Old process: 3 engineers, 6 weeks, around $47,000.

New process: 1 engineer confirming the AI's output, 9 days, roughly $10,500 combined, engineer's time plus AI compute.

Coordinated models finished this way cut change orders by 60 to 80 percent once construction begins.

See the article below for a smaller example of the Kimi K3 + Blender MCP workflow.

[dax @thdxr]

used a trick @jlongster came up with

agents can control browsers but you can also ask it to record network requests into a HAR file

then it can derive a client for any website which is more efficient than browser controlling it every time

made it build a quick uber eats cli

[Erik Paluka @AIwhiteboy]

@maxxmalist UGC

[MAX @maxxmalist]

the editing part of AI UGC just got solved

you can hook claude straight into after effects through higgsfield, and now i just talk to my ad project, it reads the comp, the layers, the keyframes, and it builds the thing while i keep directing it

kinetic captions on the hook, a wiggle-locked zoom on the punchline, animated price tags, it writes the expression and puts it on the right property

and you can do it in batches, just feed it a sheet of hooks and it swaps the text across the comp and hands you every variant with layout intact, same for markets, one ad becomes 6 languages without touching you keyboard

i broke the ENTIRE ad post-production system into a workflow and giving it all to you now

here's what you're getting:
- exact setup to get claude running inside AE (5 min, copy paste)
- caption + hook animation prompt (steal it, works on any UGC clip)
- the batch variant method (1 ad, 12 hooks, one pass)
- full pipeline, gen to final export

rt + reply "UGC" and i'll send it so you can ship 10x more creatives this week (must follow so i can dm)

[Pounds @pounddz]

I shouldn’t be sharing this but f*ck it

Omni Flash + GPT 2 Image + TikTok affiliate finally lets you take part on this TikTok affiliate gold mine

I’ve cracked the code on making yapper style TikTok affiliate content that Gen Z girls LOVE!

I’m going to be sharing my full framework with you…

Here’s what your getting:
- AI doesn’t make the video good. The SCRIPT DOES. I included a full scripting frame work
-People have failed continuously because they tried to do this with Seedance. I’m sharing my full Omni method to make cheaper, better videos FASTER
-You get low views for a reason. I’m sharing my TikTok algo cheat sheet to ensure every video gets views (learned this from 1000s of videos posted)

RT + reply ‘GENZ’ and I’ll send you the step by step frameworks (must be following so I can DM)

[Erik Paluka @AIwhiteboy]

@pounddz GENZ

[Erik Paluka @AIwhiteboy]

@pounddz Interested in the case study

[Pounds @pounddz]

The reason I was able to make $80,758 in just under a month with paid traffic

In my very first month of running TikTok ads

was because I started with organic first

There’s a reason the guys pulling the biggest revenue with paid ads always come from organic

Being able to garner attention from videos and convert them organically

makes you a demon at creating ads

Starting with organic and making $10k/month then using that capital to start ads

Is honestly the most underrated path

I put together a case study for the guys who are already crushing it with organic and want to jump into paid ads

Including
>The offers
>The tech stack
>Ad account setup
>Example creatives

If you want it comment below and I’ll send it to you (must be following so I can DM)

[Pierrick Chevallier | IA @CharaspowerAI]

I can't believe this actually works 🤯
I drew a few random red arrows on an image.

Seedance 2.0 on Higgsfield understood exactly how I wanted the camera to move.

And turned it into a cinematic flythrough.
No timeline.
No camera controls.

Just vibes and red scribbles.
What a time to be a creator.

[Erik Paluka @AIwhiteboy]

@CharaspowerAI 🤯

[Erik Paluka @AIwhiteboy]

@CharaspowerAI Amazing

[Pierrick Chevallier | IA @CharaspowerAI]

Ever wanted to become a character in the next GTA? 🤯
No problem.

I built a complete workflow that lets you step into Vice City as a badass hacker and generate your own cinematic sequence 🎬🔥

Character creation.
Scenes.
Animation.
Everything.

And the best part?
You can do the entire workflow in one place inside #Magnific.
📌 Bookmark this thread before you lose it 👇

[BRUH 👑 @ItsBitcoinBruh]

AI Agents will be the next big narrative in Crypto. Catch the trend before others so you get ahead.

JUNE 2026 BOTTOM FOR ALTS (DOUBLE BOTTOM IN OCT)
BTC TOPS 180-220K OCT 2029
CRCL $300+
XRP $5 2029 Q1

[AmirMušić @AmirMushich]

My Pinterest is generating 200k+ views/month

This traffic goes to my X

here's how you can do it 🔽

[PJ Ace @PJaccetturo]

Ladies and gentlemen, it’s here:

I’m proud to announce that 'Nexus' will be my upcoming hybrid feature film.

Here is a 5-minute teaser, made by 3 people in 2 weeks.

Made with Dreamina AI using Octo & Dreamina Seedance 2.0, full workflow coming soon

[CJ Zafir @cjzafir]

Current Best Open Source Models right now:

1st: Kimi 2.6 > Best all-round model
2nd: Deepseek v4 pro > Best instruction following + API cost
3rd: Minimax M3 > Best OS coding agent
4th: GLM 5.1 > Great at long-horizon tasks
5th: MiMo v2.5 > Best harness integration
6th: Deepseek v4 flash > Great at long analysis + speed
7th: Qwen 3.7 Max > Best multimodal capabilities
8th: Qwen 3.6 27B > Best on-device dense model
9th: Gemma 4 12B > Best SLM on-device
10th: Minimax 2.7 > Best self-improvement agent

(Haven't tested Nemotron 3, Stepfun 3.7 ultra yet)

[Ahmad @TheAhmadOsman]

Let me make your Codex Cli experience better with

- Multi-agent delegation
- Enhanced memory
- Better artifacts
- Children AGENTS.md contextualization
- Runtime metrics (optional)

Run the command in the screenshot below to get them up and running

Let me know how you like it

[Ahmad @TheAhmadOsman]

Everything You Need To Know About
Inference Engines and Running LLMs Locally at Home

Explains why Inference Engines exist in the first place
- Prefill is not Decode
- VRAM is not bandwidth
- Fit is not speed
- KV Cache is the real memory problem
- Quantization only matters if the engine has good kernels for it
- Batching is not scheduling
- MoE and the routing problem
- How long context changes the serving problem
- Multi-GPU changes the interconnect problem
- Production: latency, p99s, backpressure, routing, metrics, and failure behavior

Then maps the Engines including:
- llama.cpp → portability king
- MLX / MLX-LM → Apple Silicon weapon
- ExLlamaV3 → multi-GPU consumer CUDA / local MoE
- vLLM → default open-source production server
- SGLang → long-context, MoE, routing, ugly workloads
- TensorRT-LLM → max NVIDIA performance
- NVIDIA Dynamo → fleet orchestration

The point of this article is not “use vLLM” or “use TensorRT-LLM” or “use llama.cpp”

But rather fully grasp how the Inference Engines are the traffic cop, memory manager, kernel dispatcher, scheduler, cache accountant, parallelism planner, API surface, and sometimes the deployment framework

Do not pick the engine first
- Pick the hardware
- Pick the workload
- Pick the serving model
Then the engine becomes obvious

Opensource / Local AI FTW

[Ahmad @TheAhmadOsman]

Step-By-Step LLM Engineering Projects Roadmap

- Build a tokenizer
- Learn embeddings
- Implement RoPE / ALiBi
- Hand-wire attention
- Build MHA
- Build a Transformer block
- Train a mini-former
- Compare objectives
- Build sampling
- Speculative decoding
- KV cache
- MQA / GQA / MLA
- Long context
- FlashAttention
- Hardware budgets
- Toy MoE
- Sparse model trade-offs
- State-space / linear attention
- Diffusion language models
- Data pipelines
- Synthetic data
- Scaling laws
- SFT / DPO / RLHF / GRPO
- Quantization
- Serving stacks
- Eval harnesses
- RAG
- Tool use / agents
- Vision-language adapters
- Interpretability
- Red-team suite
- Full capstone model system

One request:

Choose an Opensource AI lab when you make it

Opensource is where humanity gets to keep the tools

DM me when you've made it ;)

[Vibe Marketers HQ @vibemarketersHQ]

programmatic seo math is simple:

no of searches x 1000s of pages = compounding traffic

here is the exact template that lets customers find you without building an audience

(give this page template to ai)