The AI guy
United Kingdom
Joined September 2016
- Tweets1K
- Following435
- Followers25.1K
- Likes4.8K
My AI transcription invented ~100 words nobody said, over crowd noise on a live podcast.
So I tested free models on human checked audio, on a no GPU server:
Whisper large-v3: 14.1% wrong, 29 min per hour
large-v3-turbo: 9.7% wrong, <8 min, made nothing up
Smaller model won.
New site's live. Every frame has one lime point, and the last one becomes our full stop. apolloweb.ai
Sound On 🔊
Quit my 6 figure salary job last year to start my own Local SEO agency here in the UK. Reached my goal of £8k MRR in under a year with only UK clients.
Very stressful, lots of times I wanted to give up and thought I couldn't do it. But I never quit and stuck it out.
Hope this inspires you to keep going!
🚀 Local SEO Win (Under 30 Days)
Took a local contractor from avg map rank 4.5 → 1.2 and secured #1 positions across their entire town.
No spam. No CTR. No PBNs.
Just:
• Proper GBP optimisation
• Real local relevance signals
• Clean NAP citations + service pages
• Local link building
Result: 40 → 156 interactions/month and the phone has not stopped ringing.
If you want results like this, DM me “MAP PACK”.
I now run my own Web/SEO Agency (specialising in Local SEO). Spent months testing and learning and now I can rank most profiles into the top 3 within a month.
Here's a case study below of their rankings going from 10+ to number 1 all over the town in just a month.
Excited to announce the launch of ApolloWeb 🚀– my new AI-driven web design & development agency!
At ApolloWeb, we're transforming ideas into standout digital experiences using cutting-edge AI technology. We deliver full, stunning websites in just 1-3 days and at a fraction of the cost of traditional agencies.
Here's a sneak peek at some real-world projects I've already designed and developed for amazing clients. Our AI-driven approach ensures rapid delivery, beautiful designs, and exceptional user experiences.
Check it out: apolloweb.ai/
I Redesigned an Entire Website in 5 Minutes with AI (V0 dev)
Tired of spending hours designing landing pages? I'll show you how I created a stunning, professional landing page, and redesigned an outdated one in just a few clicks, using AI. No coding, no hassle!
Check it out 👉 youtube.com/watch?v=_YF0Mbr8…
Just built the landing page for my SaaS MVP using v0.dev and I'm blown away!
It only took me 10 minutes to create, using the prompt: "Create a landing page for this project. It should be modern, visually appealing, and use the latest and best UI/UX practices. I want it to look like a million dollars while appealing to my target audience. Start by just creating the hero section/navbar." Also pasted in a summary of my SaaS features/target audience.
I've created a quick video walkthrough below showcasing how clean and professional the landing page turned out in just minutes.
If you're bootstrapping your SaaS, definitely give v0 a try. It's impressive how quickly you can test ideas and get something beautiful live.
Would love your feedback, let me know what you think!
💡The specific AI prompts that helped me launch my MVP in 2 weeks instead of 2 months
As a non-technical founder building in public, I was initially frustrated with AI coding tools like Cursor and WindSurf. The code they generated never quite worked... until I transformed how I prompted them.
Here are real examples that made all the difference:
DEBUGGING:
❌ BAD: "Fix the bug in my checkout"
✅ GOOD: "When users click checkout, the cart empties but no order is created. The console shows 'Payment processed' but nothing appears in my database. Here's what the error message says..."
FEATURE ADDITION:
❌ BAD: "Add a dark mode"
✅ GOOD: "Add a toggle button in the header to switch between light and dark mode. Save the user's preference so it remembers their choice when they return to the site."
INTEGRATION:
❌ BAD: "Connect Stripe"
✅ GOOD: "Add Stripe payment processing to my product page. When users click 'Buy Now', they should see a Stripe checkout form for a one-time payment of $29. After payment, they should receive access to the premium features."
I realized AI coding agents need context similar to what you'd tell a freelancer, not magic commands, turned these tools from toys into MVP-building powerhouses.
Being specific about:
- What you're trying to accomplish
- What's currently happening vs. what should happen
- Any error messages you're seeing
- How the feature should work from the user's perspective
What's your most effective AI prompting technique for your no-code/low-code projects?
Very impressed with Claude sonnet 3.7 in windsurf. Look at the screenshot, it's much better at checking multiple of the correct files to make changes and ensure when changes are made it takes into account every file it needs to.
Windsurf now offers Claude 3.7 Sonnet! 🙌
It provides much better context awareness by analysing multiple files simultaneously, resulting in superior code changes. However, it consumes credits faster than 3.5.
My recommendation: use 3.5 for general development, then switch to 3.7 only when you encounter persistent errors. This approach balances efficiency with performance.
What's your experience with 3.7 been like?
Starting your day with Claude 3.7 to analyze multiple files in your codebase provides Windsurf with comprehensive context that 3.5 simply cannot match. This initial investment of credits creates a foundation of understanding that makes subsequent development more efficient. Think of it as giving Windsurf a thorough orientation before asking it to help with specific tasks—the improved context awareness translates to more accurate code modifications throughout your workflow.
Building SaaS with AI: half the time it’s smooth sailing, half the time it’s a bug-fest.
Lesson learned: human intuition still wins the debugging game.
What’s your unexpected lesson?
Everyone thinks AI coding agents like Cursor and WindSurf will replace developers overnight, but the truth is shockingly different - studies show they automate only ~10% of complex tasks in enterprise businesses. They’re powerful aids, not total replacements. What’s your take?
Have you noticed that AI coding assistants often struggle with large codebases? They miss dependencies and make isolated changes that can break things elsewhere.
That’s why I built a web app that generates smart documentation for your codebase—mapping file relationships and dependencies so tools like @WindsurfAI and @CursorAI get the full picture.
So far, it’s led to:
- Fewer bugs from AI-suggested changes
- Less back-and-forth prompting
- A clearer understanding of cross-file impacts
Reply if you’d like early access 👇
