It's called Vibe Coding but does it work?
I've been testing AI for some time now. Have tried most of the 'non technical' tools but am now sticking with a stack of Lovable, Supabase, Netlify and Devlo AI with Chat GTP for when things get...
I've been testing AI for some time now. Have tried most of the 'non technical' tools but am now sticking with a stack of Lovable, Supabase, Netlify and Devlo AI with Chat GTP for when things get sticky.
UPDATE 18/09/2025
Since posting this I have been using Claude Code (CC) locally and it offers some very interesting advantages. Using lovable to spool up a quick and clean front end, connecting lovable to github and then Claude code to that has allowed me to edit locally, at less cost of tokens and with better results.
Claude seems to create less issues on front end changes and building features although it can struggle with supabase instructions. There were some new things to learn such as running sql statements in Supabase's SQL editor and for testing security I still prefer to use Devlo AI as a 3rd party.
There is a steep learning curve for linking CC to Supabase etc but if i can do it .......
Building With AI: From Experiments to Real Products
I’ve been testing AI for a while now. Most of the “non-technical” tools are shiny but shallow — they look good in a demo, but they don’t stand up when you try to build something real.
After a lot of trial and error, I’ve settled on a stack that actually works:
- Lovable for app development (building codebases through prompts)
- Supabase for the database, authentication, and storage
- Netlify for fast, simple hosting
- DevloAI for workflows and automation
- And ChatGPT as the backup brain for when things get sticky
With this combination, I’ve gone from AI tinkering to building real, working products. Some small, some bigger. All possible in hours or days, not weeks.
AI Debate – 180 Prompts, Full App
Live at ai-debate-ai.com
This started as a test: could I build something genuinely complex using only Lovable prompts?
Screenshot
The result: a platform where you can watch famous personalities debate any topic. Churchill vs Thunberg. Clarkson vs Musk. Marx vs Thatcher.
In 180 messages and 80 AI edits, Lovable generated a ~5,000 line codebase with:
- Authentication via Supabase
- 20 different AI personalities with custom prompts
- Real-time orchestration so debates feel live
- Stylised avatars for each persona
- Suggested debate topics by category
- SEO, security, and session management
Total build time: ~4 hours.
The real achievement wasn’t just speed — it was how Lovable handled architecture and context without me writing a line of boilerplate.
Campfire – Group Chat + Task Agent
Live at aretherenogooddomainsleft.com
Think WhatsApp, but with an AI agent in the group chat.
The agent can schedule things, track decisions, or jump in when the group wants something handled.
Screenshot
Lovable gave me the chat structure. Supabase took care of user data. DevloAI plugged in the background workflows.
A simple concept, but one that would normally take weeks of fiddling with sockets and auth. Here it was running in a day.
Lottery Checker
Live at 6numbers.net
This analyser checks if your regular lottery numbers have ever appeared together in past UK draws.
Screenshot
It scans historical results to find any instance of your numbers winning any prize.
A niche tool, but it shows the strength of the stack:
- Data stored and queried in Supabase
- UI scaffolded by Lovable
- Deployed instantly on Netlify
Time to live: about 3 hours.
CarSearchAI – A Site for Robots, Not Humans
Live at carsearchai.com
This one flips the idea of SEO on its head. It’s a text-only car listings site, designed not for humans but solely for LLM crawlers.
Screenshot
It blocks Google and traditional search bots, but allows AI crawlers to index structured auto data in real time.
It’s an experiment to see whether AI systems are already looking for “clean” data feeds online. Built in Lovable, deployed on Netlify, with Supabase as the data engine.
EQ8 – Fitness Competition Site in an Hour
Preview at https://preview--eq8.lovable.app/index
A complete site for a new hybrid fitness competition.
Screenshot
- 100% AI-generated content, images, and video
- Built in Lovable in about an hour
- Videos prompted in Gemini using JSON
- Fully responsive and deployable instantly
Here is the json:
{
"character_name": "Jordan Lane",
"character_profile": {
"age": 31,
"height": "5'10\" / 178 cm",
"build": "strong and lean, high-performance athletic frame",
"skin_tone": "sun-kissed olive with visible freckles",
"hair": "dark brown, tied back in a tight ponytail",
"eyes": "focused green with a glint of intensity",
"distinguishing_marks": "sweat-soaked EQ8 wristband; light scar on left brow",
"demeanour": "genuine, driven, energetic with a confident smile"
},
"global_style": {
"camera": "handheld gimbal with shallow depth, dynamic framing with slow push-ins and lateral moves",
"color_grade": "sun-drenched natural tones with boosted contrast (greens, blacks, golden skin)",
"lighting": "natural golden-hour sunlight with soft fill to catch facial features",
"outfit": "green EQ8 vest, black performance shorts, compression socks",
"max_clip_duration_sec": 8,
"aspect_ratio": "16:9",
"mouth_shape_intensity": 0.9,
"eye_contact_ratio": 0.9,
"audio_defaults": {
"format": "wav",
"sample_rate_hz": 48000,
"channels": 2,
"style": "epic cinematic sports score with ambient vocal texture, 90 BPM"
}
},
"clips": [
{
"id": "EQ8_ChampionVoice",
"shot": {
"composition": "Medium close-up, 50 mm lens, shallow depth of field, handheld",
"camera_motion": "slow push-in as Jordan steps forward post-event, panting slightly, eyes locked on camera",
"frame_rate": "24 fps",
"film_grain": 0.03
},
"subject": {
"description": "Jordan Lane — 31-year-old, 178 cm, powerful and lean; olive skin, green eyes, dark ponytail; wearing sweat-darkened EQ8 vest, flushed from effort",
"wardrobe": "green EQ8 weighted vest, black shorts, white compression socks"
},
"scene": {
"location": "EQ8 competition track under afternoon sun, crowd in soft focus background",
"time_of_day": "late afternoon",
"environment": "sun-drenched outdoor fitness course; EQ8 banners, crowd cheering faintly"
},
"visual_details": {
"action": "Jordan wipes sweat from brow, steps up to camera and speaks directly with visible emotion",
"props": "EQ8 finish line tape in background, kettlebell on ground nearby"
},
"cinematography": {
"lighting": "natural sun with slight lens flare; soft side-fill to balance shadows",
"tone": "authentic, victorious, grounded"
},
"audio_track": {
"lyrics": "EQ8 proves it. Doesn’t matter your age, your size, or your gender—this is real fitness, real competition.",
"emotion": "passionate, victorious, humble",
"flow": "even tempo with rising emphasis toward the final phrase",
"wave_download_url": null,
"youtube_reference": null,
"audio_base64": null
},
"color_palette": "bold green (vests), asphalt black, warm gold skin, vivid sky blue",
"dialogue": {
"character": "Jordan Lane",
"line": "EQ8 proves it. Doesn’t matter your age, your size, or your gender—this is real fitness, real competition.",
"subtitles": false
},
"performance": {
"mouth_shape_intensity": 0.9,
"eye_contact_ratio": 0.9
},
"duration_sec": 8,
"aspect_ratio": "16:9"
}
]
}
A good example of how quickly branded content sites can now be spun up — where the cost is measured in prompts, not developer days.
Car Video Boost – Automated Dealer Videos
Live at https://carvideoboost.com
This one is heavier technically. It takes a dealer’s stock, fetches images, generates scripts, voices them with AI, and creates videos for every car automatically.
Screenshot
Dealers don’t upload anything — the system does it all end-to-end.
Again, the core runs on the same stack: Supabase for data, Lovable for the orchestration, Netlify for deployment, and DevloAI for background task management but then adds FFmpeg in AWS Lambda and integrations with ElevenLabs and Open AI APIs.
Why This Stack Works
The pattern across all of these is clear:
- Lovable removes the drudgework. It writes the scaffolding, manages context, and suggests improvements.
- Supabase provides a secure, ready-to-go backend.
- Netlify makes deployment instant and reliable.
- DevloAI handles workflows and automation behind the scenes.
Together, they let you get from idea to live product in hours.
The Bigger Shift
Most AI tools are toys. This stack builds products.
Not all of them will be unicorns, but that’s not the point.
The point is that technical barriers are falling fast. What used to take a team weeks can now be done by one person in a weekend.
The bottleneck isn’t code anymore — it’s imagination, clarity, and execution.
That’s where the real work is now.
