Mark Builds AI

Testing AI tools so you don't waste your time.

[Mark Builds AI @MarkAIbuilder]

@neil_xbt the "one vault, every project" line is the actual unlock here. shared context across claude code projects > any single feature

[Mark Builds AI @MarkAIbuilder]

@bonsaixbt automating uploads is the easy part. surviving YouTube's policy on low-effort kids content is the part nobody models

[Mark Builds AI @MarkAIbuilder]

@leopardracer "knowledge that connects itself" is the right frame. manual linking was always tax you paid for bad tooling

[Mark Builds AI @MarkAIbuilder]

@cipgerx been running Seedance 2.0 on my side project for a week. the consistency jump from v1 is the real story here

[Mark Builds AI @MarkAIbuilder]

@undefinedKi Omg, it’s legendary

[Mark Builds AI @MarkAIbuilder]

A claymation ad used to mean a studio, a month, and five figures.

This is the same ad for 12 cents in 3 minutes.
When production cost hits zero, the winner isn't whoever can afford to make ads. It's whoever has the most ideas worth testing.

[Miko @Mho_23]

I just cracked the code on AI animation ads 🤯

every one of these cost me 12 cents and under 20 seconds to make

Gemini Omni + Claude Code is f*cking wild.

i built a system that turns any product into an end-to-end AI animated ad in ANY style you want.. claymation, pixar-style 3D, anime, paper cutout, lego, you name it

you drop in your product photo and a one line pitch and it handles everything end to end. images, video, voiceover, music, captions, all generated with AI inside Claude..

AI animation ads are crushing on Meta right now and everyone still assumes you need an expensive editor with a 3-week turnaround time to make one. you don't. you just need the right workflow.

Im making full end to end high quality AI animation ads in under 3 minutes with this system

and they work just as hard organically on TikTok, reels and shorts

so i packaged the whole thing up…

here's what you're getting:

> every prompt i used across all 8 steps
> the exact tool stack + the order to run it in
> style recipes for claymation, pixar, anime, wes anderson, retro cartoon + CGI
> one system that works for paid Meta AND organic
built across Claude Code and Gemini Omni

RT + reply 'STYLES' and i'll send it over (must be following so i can dm)

[Mark Builds AI @MarkAIbuilder]

@Gaurav_vij137 what acceptance rate were you seeing where it fell off? curious if it's the format or the draft mode

[Mark Builds AI @MarkAIbuilder]

DeepSeek dropped DSpark and casually open-sourced the entire speculative-decoding codebase (DeepSpec) right alongside it.

1. Up to 4x throughput.
2. Free.
3. Works on other models too.

China's labs keep treating open weights as the default. The Valley treats it as a liability.

[Mark Builds AI @MarkAIbuilder]

this is the gap between people who "tried Claude Code once" and people who actually live in it. completely different tool depending on how deep you go.

[am.will @LLMJunky]

I'll bet you didnt know you could do this with Claude Code

[Mark Builds AI @MarkAIbuilder]

The new AI filmmaking stack is wild:
1. Midjourney/Chat GPT → start frame
2. Blender → block the action with basic shapes
3. Seedance 2.0 → turns it into a real scene

No film school. No crew. Just taste + 3 tools.
https://t.co/Q03b73OZqt

[Zentrix⌚️ @ZentrixHQ]

https://t.co/hwiagfxTgi

[Mark Builds AI @MarkAIbuilder]

I've now seen this exact business go viral three times — first Gaussian Splatting, then Luma, now Claude
Code and a $30 spinning mount.

That's not a knock. That's the signal.

When the same idea keeps resurfacing with cheaper tools each time, the demand is real and the barrier just keeps dropping. Stop scrolling past it.

[NefodOff @v_nefodov]

A 19-year-old student from China is making up to $300 a day with his phone, Claude Code, and this thing.

Not dropshipping. Not crypto. Not content creation.

He places his phone on a motorized rotating mount, sets it in the middle of a room, and walks away. The device spins a full 360 degrees, capturing every wall, ceiling, corner, and detail while the app counts up the scan progress on screen. When it hits 100, the room exists as a complete digital model.

Claude Code handles the rest. It processes the scan data, structures it into a walkable 3D tour, and generates a clean shareable link. The whole pipeline from empty room to published virtual tour takes under 20 minutes.

Real estate agents in his city were paying a photography studio $400 per listing for the same result. He charges $150, does three properties a day, and his calendar is booked two weeks out.

Pause the video at
0:11 and look at the phone screen carefully. I think everything will become clear to you at that moment.

That number on the screen is not a timer. It is the scan progress. And by the time most people figure out what they are looking at, the room is already done.

No employees. No studio. No equipment worth more than $200 total.

Just a phone, a spinning bracket, and Claude Code turning a morning of work into $300.

[Mark Builds AI @MarkAIbuilder]

NOUS RESEARCH JUST DROPPED HERMES — AN OPEN-SOURCE AI AGENT THAT LIVES ON YOUR OWN MACHINE AND ANSWERS TO NOBODY

Hermes Agent (v0.17) isn't a chatbot. It's an AI operating system you self-host.
What it actually does:

Runs as a 24/7 agent on your own hardware — it codes, manages files, searches the web, and delegates tasks to background subagents while you keep working.

Lives everywhere you already talk: Telegram, Discord, Slack, SMS, iMessage, email.
"Mastermind" mode: one group chat with multiple agents, each on a different model (Claude, Gemini, Codex…), riffing off each other.
Builder/judge loops: you define "done," it builds and grades itself adversarially until it passes.
Cron scheduler + webhooks + automation blueprints — it runs on a timer and reacts to events, not just prompts.

The pros:
Free and open-source (MIT). No seats, no paywalled features.
Model-agnostic — 20+ providers or fully local. Never locked to one lab.
Self-hosted, so your data never leaves your box. Persistent memory that's actually yours.

The cons:
You have to host it. A VPS and a little devops — not plug-and-play.
You bring your own API keys, and the LLM bill is on you.
It's v0.17. Moving fast, so expect rough edges.

The price:
The software: $0.
Running it: ~$5–$80/month (VPS + model API).
Optional Nous Portal if you don't want to wire up keys: free, or $20/mo for 300+ models + tools.

The era of renting an assistant from one company is ending. Hermes is what owning one looks like.

[Mark Builds AI @MarkAIbuilder]

@starmexxx WOW! easy easy, what a hell this device!

[Mark Builds AI @MarkAIbuilder]

@VaibhavSisinty "fewer lines" can just mean clever one-liners nobody can debug later — readability's the real cost

[Mark Builds AI @MarkAIbuilder]

@mhdfaran tried stitching gen-video into an edit last month, the export/reimport dance killed it — if this closes that loop it's the first one worth using — angle

[Mark Builds AI @MarkAIbuilder]

@shivkanthb claude code on your face is the agent-babysitting endgame — glanceable status beats alt-tabbing every 30s — angle

[Mark Builds AI @MarkAIbuilder]

@mhdfaran "completely useless without this" then it's a tool registry — the agent's only as good as the tools it can actually call

[Mark Builds AI @MarkAIbuilder]

https://t.co/tm7MPzxhGP

[Mark Builds AI @MarkAIbuilder]

We talk about AI like it lives in "the cloud."

It doesn't. It lives in 3,800-football-field warehouses pulling water and power from small towns.

The next big AI fight isn't about models. It's about land.

[Mark Builds AI @MarkAIbuilder]

@smm_aryan yes, for sure, base for everyone!

[Mark Builds AI @MarkAIbuilder]

https://t.co/t0kBXgxyIP