🚀 Aerospace Eng → AI Automation🤖 Founder @ compassai.io 📈 Building tools that drive revenue 🦾 Helping businesses grow with AI
Joined June 2025
- Tweets 198
- Following 66
- Followers 14
- Likes 102
62 Photos and videos
Jun 10
Interesting pattern:
Direct prompting inside video generators often underperforms compared to using Claude as the “prompt orchestrator” first.
Better structure → better generations.
Tested repeatedly with UGC ad workflows. 👀
#GenerativeAI #AIContent #ClaudeCode
12
May 12
“Premium, modern, clean” is why your AI output looks generic.
That is not direction.
That is decoration.
Better prompts define:
brand,
positioning,
offer,
audience,
visual DNA,
tone.
Specific words create premium output.
7
May 8
Most brands don’t need a redesign.
They need consistency.
Claude’s new Design System feature quietly fixes one of the biggest trust killers in online business.
15
May 1
Stop juggling AI apps like a circus act.
Most people are stuck switching tabs, losing context, and restarting workflows.
Big mistake.
I hooked Higgsfield Claude via MCP — now it’s ONE seamless flow:
• Research & strategy in Claude
• Instant video/assets in Higgsfield
• Zero context loss. Zero tab chaos.
This is how AI was meant to work.
If you’re still jumping between tools… you’re burning time.
Fix your workflow.
#AIWorkflow #AITools #MCP #Automation
31
Apr 28
Higgsfield just dropped MCP support.
One prompt → script, images, video, delivery.
No app switching. No manual steps. Just output.
Solo creators now run the same pipeline as full agencies.
The scary part? It works.
(They already caught heat for celebrating 20 jobs eliminated in Feb.)
The tools aren't coming. They're here.
24
Apr 22
Post:
Most people prompt AI like this:
“Create a cool cinematic ad”
That’s way too vague.
Better framework:
→ Creative Direction
→ Hero Element
→ Motion
→ Visual Event
→ Scale
→ Climax
→ Polish
This shot started with one image one structured prompt.
That’s how AI content starts looking premium.
#AI
#Marketing
#AIVideo
#Prompting
#BuildInPublic
23
Apr 19
I stopped writing captions today. Permanently.
One API key. Google AI built my entire workflow — implementation plan, UI prototype, platform descriptions — in 15 minutes.
Copy. Paste. Done.
The manual era of content creation is over.
1
23
Apr 19
Build a Next.js web app called '[].' Users input a video script. The app uses the Gemini API to generate: 5 viral hooks, and platform-specific titles/descriptions for LinkedIn, YouTube, Instagram, TikTok, and X. Include a 'Copy to Clipboard' button for each. Use shadcn/ui for a clean, dark-mode aesthetic.
Technical Requirements:
Stack: Next.js 14 (App Router), TypeScript, Tailwind CSS, Lucide Icons.
API: Implement a POST route /api/generateusing @google/generative-ai.
Prompt Engineering: Force Gemini to return a strict JSON schema: { "hooks": [], "platforms": [{ "name": "", "title": "" }] }.
UI Components: * Textarea for script input with a "Generate" button using Loader2 for loading states.
Card containers for the 5 Hooks.
Tabs component to toggle between LinkedIn, YouTube, Instagram, TikTok, and X.
UX: Add sonner or toast notifications for "Copied to Clipboard" actions.
Styling: [YOUR STYLE]
89
Apr 18
Free API token let’s go
The Claude Code hackathon is back for Opus 4.7.
Join builders from around the world for a week with the Claude Code team in the room, with a prize pool of $100K in API credits.
Apply by Sunday: cerebralvalley.ai/e/built-wi…
21
Apr 17
Most people blame their prompts.
The real problem is the model.
I ran the same prompt through 3 top AI models. Only Seedance 2.0 actually understood it.
This comes down to semantic fidelity — how precisely a model translates what you mean, not just what you typed.
Prompt in the replies 👇
1
34
Apr 17
Cinematic high-speed power slide (drift) originating from the [Uploaded Keyframe]. The vehicle maintains a precise lateral slide toward the camera lens. Thick, photorealistic volumetric tire smoke billows from the rear wheel arches, realistically swirling and colliding with the car's bodywork and the studio floor. High-velocity motion blur is applied to the spinning rims.
Camera Movement:
Dynamic "Snorricam" low-angle tracking shot. Precise, robotic 180-degree orbit starting 1 meter from the front bumper, sweeping smoothly around the side to the rear. Zero handheld shake; execution mimics a Technocrane or Bolt Cinebot arm.
Lighting & Surface Rendering:
Commercial studio aesthetic with dark industrial surroundings. Overhead "light-painting" LED strips create high-contrast linear highlights. Ray-traced reflections on the metallic paint and glass must remain spatially consistent and "stick" to the geometry during the rotation. Sharp, horizontal anamorphic blue lens flares.
Technical Specifications:
8K resolution, 60fps fluid motion, hyper-realistic textures, clean commercial grade, no geometric warping or flickering. Consistent light transport and shadows.
21
Apr 15
Most people are failing at AI video in 2026 for one simple reason:
They’re using the wrong model for the wrong goal. 🧵
1
6
Apr 15
ByteDance is pushing multimodal content generation to another level
Next level control
11