The Unlikely Journey from Bricks to Bytes

April 2020: I'm holding a trowel in one hand and my phone in the other, scrolling through yet another failed deployment of 24ad.info. The classifieds platform I'd envisioned was losing £400/month to freelancers who promised "production-ready solutions" that crumbled under traffic.

The Hard Reset

When third Cloudflare bypass attack hit in Q3 2021 (283,103 intrusion attempts logged), I made three decisions:
1. Ditch all pre-built scripts and "experts"
2. Learn infrastructure the hard way on OVH VPS (€5.99/month, 2vCPU)
3. Build only what absolutely needed to function

The stack evolved brutally:
- 2021: LAMP stack with 11s TTFB
- 2022: Caddy replacing Apache/Nginx (43% faster static loads)
- 2023: Sentinel monitoring 22 critical processes with 99.2% attack prevention
- 2024: Full rewrite to React 19 + tRPC (6h migration for breathtime.info)

Sentinel: The Accidental Product

What began as a survival tool became our flagship. Key metrics from last month:
- Blocked 17,942 brute-force attempts
- Automated 831 database optimizations
- Maintained 100% uptime across 3 production apps

Pricing reflects what I wished existed when I started:
- Free Tier: 1 server, basic monitoring
- Pro (£39/mo): 5 servers + Fail2Ban integration
- Enterprise (£149/mo): Unlimited nodes + AI threat prediction

The AI Pivot Point

Game changers in my toolbox:
1. Claude Opus: Architectural decisions debugged conversationally
2. GPT-4 Turbo: 73% faster React component generation vs. hand-coding
3. Self-hosted LLMs: Privacy for client data pipelines

A typical workflow now:

[ACT AS SENIOR SRE]  
Here’s my server load spike pattern during UK rush hours:  
- 08:00-10:00: 5.2 → 8.7 load avg  
- Current setup: 4GB RAM, 2vCPU, MySQL 8.0  
Propose three concrete optimizations sorted by ROI  

Lessons from the Trenches

What Worked

• OVH’s bare metal reliability: 5× less downtime than previous cloud providers
• Caddy’s automated TLS: Saved 14hrs/month on certificate management
• Radical transparency: Publishing server logs built client trust

What Nearly Sank Us

• Early Laravel technical debt (56% slower than current Node.js implementation)
• Underestimating bot traffic (772GB bandwidth overage in one month)
• "Temporary" MySQL patches becoming permanent firefighting

The FixFlex Ethos

1. No "rockstar developers": Our best-performing project was spec’d by a plumber using AI
2. Infrastructure as craft: Every server config documented like medieval masonry techniques
3. Constraint-driven innovation: £0 marketing budget → built tools so useful they market themselves

Tonight’s Reality:
While our stack now serves 17,000 monthly users across products, I still keep steel-toe boots by the door. Some habits—like manually checking server logs at 5:30am—stick harder than mortar.

The blue-collar approach to tech: Measure twice, cut dependencies once, and always keep trowel-sharp tools within reach.

[Current project stats: 5 active builds, 3 live platforms, 99.5% uptime over 724 days]