For the last seven months I’ve been building Lumarka, a platform that sits between AI video generators and actually making a movie with AI. It’s ambitious. I probably bit off more than I can chew. But the journey taught me something I didn’t expect. I started with n8n. Built two or three dozen workflows. Got pretty good at it. But somewhere along the way I noticed something: most of my workflows were basically code with a visual wrapper. I was writing logic, handling edge cases, debugging errors… just through little boxes instead of text. And at some point I had this very simple thought: why am I adding this extra layer at all? With Claude Code, I can work directly with the source. No translation layer. No dragging nodes around to represent logic I could just write. When something breaks, I’m not debugging a workflow and the code inside it, I’m just debugging code. When I want to change something, I change it. The feedback loop is tighter. The power is more direct. Don’t get me wrong, n8n still has its place. Modular automations, swapping tools in and out, quick experiments… it’s great for that. I still use it. But for building an actual product? Working directly in code with Claude as my partner is just on another level. What really blows my mind is that I’m even able to say this. For most of my career, “code” might as well have been hieroglyphics. I avoided it. It felt like a different species of thinking. I never imagined I’d understand what was going on under the hood, let alone enjoy it. I’m still lost sometimes. But now when I hit a wall, I have a partner who explains what’s happening in plain English, and I push through instead of bouncing off. That alone feels like a small miracle. So what does this have to do with AI filmmaking? At first glance this probably sounds like a nerdy tooling story. But for filmmakers, this shift is actually about something much simpler: creative control and speed. Right now most AI filmmaking workflows are fragile. You bounce between tools, UIs, prompts, formats, and half the time you’re fighting the software instead of shaping the story. Every extra layer adds friction. Every abstraction hides what’s really happening.

Youtube video tutorial: https://youtu.be/Wc09u6ZwPAM?si=hJwn82ihFiOUra69 Short: https://youtube.com/shorts/t8-Wxso-XAc It's straightforward! You create the "earn" account (https://higgsfield.ai/earn) , which gives you a month ULTIMATE free on Higgsfield.ai ($49 normally) You create the video according to their request, like this video I made. You get paid between $50-$3500 depending on how big your influence is. It's open to anyone with a Youtube or Instagram account Their AI Influencer system is very simple: - select customization features from menus or prompt to generate the influencer cover image - (optionally edit the influencer image) - add motion - (optionally edit) - publish! There are no catches here, I'm an affiliate and not even putting the link up - this just seems like a crazy awesome deal (Higgsfield Utlimate for a month includes unlimited generations!!!)

Adobe just rolled out Premiere 26, (They've removed "Pro" from the name for some reason?) and this one feels like more than incremental polish. A few highlights that stood out to me as someone who’s been using Premiere since version 3: One-click object masking & tracking: Hover, click, isolate. AI-driven masks that actually track moving subjects, without frame-by-frame pain or immediate round-trips to After Effects. Massively faster shape masks: Ellipse, rectangle, and pen masks rebuilt to track up to 20× faster, with live previews and much better refinement controls. Frame.io built directly into Premiere (beta): Review notes, comments, versions, and media ingest without leaving the timeline. Less context switching, tighter collaboration. Built-in Adobe Stock access: Browse, license, and drop clips straight into your edit. Not flashy, but very practical. Firefly Boards import: Early ideation and visual development flowing more directly into the edit. Now, a fair question some people might ask: What does this have to do with AI? Isn’t this just a Premiere update? Here’s my take: If you’re making films with AI, you still need to edit them. There are tools out there calling themselves “AI editors,” but that term is often misleading. Editing isn’t just cutting silences or removing flubs from a talking-head video. Film editors create stories. They shape pacing, emotion, clarity, and meaning. That requires both technical skill and creative judgment, whether the footage came from a camera or a prompt. Generating AI video clips is only the first step. Someone still has to assemble those pieces with intention and care to entertain, educate, or move an audience. That’s why improvements to real editing tools still matter... A lot. One last thought, as someone who’s also seen Premiere grow over decades: As it’s evolved into the Swiss Army Knife of NLEs, it’s also gotten heavier. Not crashy for me, but slower. That’s the tradeoff of being able to do almost everything. Compared to something like Avid, which excels at a few mission-critical things for long-form work, it’s a different philosophy.

It's not just answering questions — it's showing you. "Hey, show me the best coffee shops near the Eiffel Tower" And it just... does it. Pans the map. Zooms in. Highlights spots. Talks back to you about what it found. This isn't search. This is conversation. --- What is this thing? Voice AI Conversation with Google Maps — a voice-powered AI agent that actually controls Google Maps while you talk to it. Ask it anything: - "What's the fastest route avoiding highways?" - "Find me beachfront hotels under $200 in Portugal" - "Show me where all the national parks are in Utah" - "Zoom out and show me the whole country" It doesn't just answer. It shows you. In real-time. While talking back. --- Why this hits different We've all typed into Google Maps. But talking to a map that responds, moves, and explores WITH you? That's a completely different experience. It's like having a knowledgeable local guide who also happens to control a giant interactive map on your wall. The Open Source Repo: https://github.com/moshehbenavraham/chat_with_google_maps --- The nerdy part (for my fellow builders) 🔧 Started with Google's AI Studio demo — cool proof of concept, but basically a toy. Nearly FULLY AUTOMATICALLY built: - Full authentication system - Secure backend (no exposed API keys 🔐) - Database for saving your stuff - AI monitoring dashboard and logging - Local/Dev/Production deployment on Vercel - 18,500+ lines of code Built the entire thing using the Apex Spec System — an open-source spec-driven Plugin/Skill for Claude Code. Every feature was spec'd out first, then built systematically. Complex project, zero chaos. 🔗 github.com/moshehbenavraham/apex-spec-system --- The future is conversational We're moving from: Click → Type → Talk This is just the beginning 🚀 Voice + AI + Maps is one combination. What about Voice + AI + your industry? The patterns we built here — security, monitoring, auth, database — they're reusable building blocks.

So I got a bunch of Tavily API credits for completing their course AND I wanted to show how to use a Boilerplate template to start apps. Combined this with a system (Claude Code Plugin) I've been developing the past couple of months I'm calling 'Apex Spec System' and I made a pretty awesome and good looking app. Complete open-source here: https://github.com/moshehbenavraham/tavily-app How it works: - Phases → major feature groups - Sessions → focused implementation units - Specs → detailed requirements per session - Task checklists → 15-30 items to complete - Validation gates → quality checks before moving on The result: - 15 sessions across 3 phases - FastAPI backend + React frontend + PostgreSQL - Auth, CRUD, 4 Tavily operations, save results with metadata - ~15K lines of production-ready code - 2 afternoons The key insight: AI doesn't drift when it has clear scope, explicit constraints, and traceable progress. It's not magic—it's just structured prompting at the project level. Video below! Curious if anyone else is experimenting with structured AI dev workflows such as BMAD, Github Spec Kit, etc. What's working for you?