惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

美团技术团队
人人都是产品经理
人人都是产品经理
月光博客
月光博客
V
V2EX
WordPress大学
WordPress大学
酷 壳 – CoolShell
酷 壳 – CoolShell
Last Week in AI
Last Week in AI
博客园 - 三生石上(FineUI控件)
小众软件
小众软件
Hugging Face - Blog
Hugging Face - Blog
V
Visual Studio Blog
宝玉的分享
宝玉的分享
雷峰网
雷峰网
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
博客园 - Franky
博客园 - 聂微东
博客园 - 司徒正美
博客园 - 【当耐特】
爱范儿
爱范儿
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
大猫的无限游戏
大猫的无限游戏
博客园 - 叶小钗
阮一峰的网络日志
阮一峰的网络日志

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant
The Unlikely Journey from Bricks to Bytes
slawekluzny · 2026-05-30 · via DEV Community

I'm a builder. I taught myself to run servers because freelancers kept burning my money.

West London, 2021. I was standing on a site holding a cup of tea that had gone cold an hour earlier, watching a crew argue about where a wall should go. That's my actual job. Schedules, suppliers, the kind of problems that only exist at 7am when half the crew hasn't shown up and the client is already phoning.

But my head was somewhere else. I'd been chewing on an idea for a classifieds platform for months. Not a grand vision, nothing with a business plan and projections. Just a gap I could see — a way to connect buyers and sellers that felt easier and more global than what was out there.

The problem was that I knew nothing about programming. And I mean nothing. I didn't know what a database was. I'd never written a line of code. My entire technical CV was "reasonably good at not breaking my own phone."

So I did what most people in my position do. I tried to buy my way in.

The expensive year

I found a ready-made classifieds script online. Looked professional, had features, didn't cost the earth. The smart shortcut, I told myself. Then I hired a freelancer to customise it. Then another one, when the first disappeared mid-project. Then another, when the second delivered something that worked on a good day and fell over on a bad one.

Here's the thing nobody warns you about hiring freelancers when you can't read code: you can't judge the work. You can't tell the difference between someone who wrote something clean and someone who duct-taped it together to last until the invoice clears. Both show you the same thing — a screen where the button does what the button's meant to do. So you pay, you say thanks, you move on. And three months later the button stops working and the freelancer's gone.

Meanwhile the bots had found me. Within weeks of going live, automated scripts were hammering the contact form, then the registration page, then the login. "It's normal," a freelancer told me. "Happens to everyone." What he didn't mention is that "happens to everyone" and "your server is set up to handle it" are very different sentences. Mine wasn't.

After about a year I did the maths on what I'd spent. The number made me feel a bit sick. And the product still wasn't right.

There was a specific evening — kitchen, about 11pm, laptop open, staring at another error I couldn't read — when I decided: either I figure this out myself, or it never launches. Not "I'll find better freelancers." Not "I'll get a technical co-founder." Just: I'm going to understand what's happening on my own server.

I had no idea what that actually involved.

Learning by breaking things

The first thing Linux teaches you is that it'll tell you exactly what's wrong. The second thing it teaches you is that understanding what it just told you takes a lot longer.

My first real SSH session was about fifteen minutes of typing things wrong, being told I didn't have permission to do them, and then being told I didn't have permission to give myself permission. I didn't quit — not because I'm brave, but because I was annoyed. It was my server. I was paying for it.

There's no syllabus when you learn this way. You learn whatever the problem in front of you needs. I learned SSH because I needed to get in. Linux commands because I needed to read logs. Fail2Ban because someone mentioned it in a forum thread about bots. I learned about web servers by breaking them — Nginx, then Apache, then LiteSpeed, then Varnish, then H2O for a bit. Eventually I found Caddy, which is what you get when someone builds a web server for people who aren't web server experts. Its config reads almost like plain English. I'm still on it today.

I should admit something. For the first two years I didn't use Git. Everyone uses Git. I took snapshots instead — a full snapshot through the hosting panel before touching anything I wasn't sure about. Slow, inelegant, not how professionals do it. But when things went badly wrong, and they did with some regularity, I could roll back. It's how I survived long enough to eventually learn the proper way.

My process, which I wouldn't have called a process at the time: describe the problem in plain language, as precisely as I can, into an AI chat window. Read what comes back. Try to understand what it's meant to do before I run it. Then run it on the live server, with a snapshot as the safety net. Every developer reading this is wincing. But every mistake had real consequences, and there's no better teacher than a mistake that actually costs you something.

The night that built Sentinel

The alert came in just after 3am. Server unresponsive. SSH timing out. CPU pinned at a level I'd never seen — not a spike, a sustained load that had been building for hours while I slept.

It was the Laravel app, being hit by a coordinated bot campaign that had found the contact form. No rate limiting. No CAPTCHA. To a script, an open door. Nearly a million emails went out before I shut it down. The server got blacklisted — not by one service, by several. I spent the next week in a state of focused misery: site by day, then evenings contacting blacklist operators, waiting on reviews, and putting in the things that should have been there from the start. Rate limiting on every form. CAPTCHA where it mattered. SPF, DKIM, DMARC done properly. Monitoring on outbound email.

While I was fixing all that, I started a list of everything that could go wrong on a server like mine and what would have to be true to catch it. Nothing I could find did all of it. So I started building it myself.

The first version was not impressive — a Python script on a cron job, checking a list of things every few minutes. But it ran, and it found stuff. A service that had quietly crashed seven times in an hour. A query taking twelve seconds. Disk usage creeping toward full. Small things. The kind that become big things if nobody's watching. I was watching.

That tool eventually became Sentinel. As I write this, its blocked-attacks counter — every banned IP, every blocked request, every Fail2Ban trigger since it started running — reads 283,103.

What changed

For a long time I was still slow. Every feature took too long. The gap between having an idea and making it real was still too wide. Then the way I worked with AI changed — less "give me the answer," more "here's what exists, here's what I need to exist, here's how I'll know it works." I could think out loud and have the vague bits noticed and questioned. For someone still building the mental models that experienced developers have on autopilot, that was the unlock.

I also made a decision that felt dramatic and now feels obvious: I threw away the Laravel codebase I'd inherited from freelancers and didn't fully understand. Over a few months I moved everything to React, TypeScript, tRPC and Node. It was faster and cleaner, and I understood every part of it — because I'd either built it or understood what the AI built and why. At one point I rebuilt one of my smaller sites in about six hours. Eighteen months earlier that would have been unthinkable.

Where I am now

The alarm still goes off at 5:45am. Site by seven. By the time most tech founders are reaching for their first coffee, I've been working for hours. The construction company runs well and I find it grounding — at the end of a day there's something real in the world that wasn't there in the morning.

The software side now has a handful of live products, and every one of them started as a tool I needed rather than a product I planned: Sentinel for watching servers, the classifieds platform that was the original idea finally built properly, a spam filter that came straight out of that contact-form disaster, and a couple of tools for content. Sentinel has a free tier for a single server and paid plans beyond that — the details are on sentinel-ai.info if you run servers and the 3am story made you wince in recognition.

If you've got an idea you haven't started because you don't have the technical background — start anyway. Not with a plan to learn everything first. Just start, with whatever you've got. I didn't learn to run a server and then build things. I built things and learned to run a server because I had to. The knowledge follows the doing, not the other way round.

I still keep steel-toe boots by the door.