Computers stopped being fun. Here's a fix from Sri Lanka.

Computers stopped being fun for a lot of people — and a recent Ask HN thread put words to it. The asker started writing software 25 years ago, moved into project and service management, and says the spark is gone. Eighteen points, twelve comments. Not a viral post. Just a quiet one that landed hard, because this slug — 2026-05-17-ask-hn-when-did-computers-stop-being-fun — describes something I see in friends here too: developers who are exhausted by Jira and unimpressed by another LLM demo.

---

🔍 The thing that killed it was scope, not the keyboard

Building something used to mean 200 lines, one machine, one user — you. Building something now means a monorepo, CI/CD pipelines, observability, generated types, three SaaS subscriptions, a feature flag for the rollout, and a meeting to discuss the meeting about the deployment.

Writing tools for induwara.lk in Next.js with TypeScript, even that "small" stack has more moving parts than the entire computer most of us learned on. The experience of assembling a working development environment is no longer a small task.

Worked example: The Sri Lanka tax calculator is roughly 400 lines of actual math and UI. The surrounding scaffolding — sitemap, schema markup, theming, analytics, accessibility passes, type generation — is another 4,000 lines. The math was the fun part. The fun is still in there. It is buried under infrastructure that most personal projects do not need but feel required to include.

The fun did not leave. It got buried. Recognising that is half the fix.

---

🚫 "Everything already exists" is a product framing, not a truth

The HN asker says he will not build new things because "everything I may need either already exists or is too complex." That is accurate if you are building products with a market. It is not accurate if you are building things for yourself.

• There is no NIC decoder that also tells you which historical Sri Lankan king was born in the same week.
• There is no fixed deposit calculator that prints a warning haiku when the rate dips below 8%.
• There is no election-prediction bot for your office WhatsApp group that scores everyone's calls against actual results.
• There is no tool that calculates whether you can afford to quit and freelance based on your current EPF balance and expected gig income.

Nobody needs these. That is the point. The moment a project needs to justify its existence to an audience, it has already drifted out of the fun zone and into product mode.

Test: Would you still write it if the only user was your 14-year-old self? If yes — that is the spark. If no — you have drifted into building for an imaginary product manager without noticing.

---

🛠️ Constraints are the trick, and Sri Lanka supplies them free

The reason coding felt magical in the early days was not the language or the IDE. It was the constraints. Dial-up modems, 64MB of RAM, an IDE that ran in 12MB. You had to be clever because you could not afford to be lazy. The constraint was the teacher.

If you are coding in Sri Lanka, useful constraints come pre-installed and do not require any effort to install:

Constraint	What it actually teaches
A $5/month Hetzner box or Oracle free tier	How to host things instead of paying SaaS to host them for you
LKR-priced salaries making subscription stacks sting	Why to skip the Copilot subscription for one project and see what your brain still does
Patchy upcountry internet at peak hours	Local-first thinking — build the offline PDF tool you actually need
No access to Apple Developer Program (costs $99/year)	Web-first discipline — make it work in the browser before thinking native
Power cuts during evening coding sessions	UX empathy — design for intermittent connectivity from the start

I built a working-days calculator in an afternoon between a power cut and a deadline. It does one thing. It accounts for Sri Lankan public holidays. It has no analytics. It will never be monetised. It was the most fun I had had with code in months — because it had one user and one job.

---

📊 The fun curve by project type

Project type	Fun at start	Fun at month 3	Fun at month 12
Toy for yourself, no users	High	High	Either done or dead — both fine
Open-source tool for a specific community	High	Medium (maintenance creep)	Low (issue queue)
SaaS product with paying users	Medium	Medium	Depends entirely on traction
Work software at a large company	Low	Lower	"It is just a job"

The toy for yourself is the only project type where the fun curve does not trend downward. That is not a coincidence. It is structurally incapable of disappointing you because it has no expectations to fail.

---

💡 What this means for you

If you recognised the HN post in your gut, three concrete moves that actually work:

1. Pick a project with no users except yourself. Not even friends who said they would use it. No GitHub stars to check in the morning. If it dies in ~/sketches/ next week, that was always the deal. The value was in the building.

2. Pick a constraint that hurts a little. One file. No external dependencies. CLI only. Plain HTML with no framework. Sinhala-only output. Whatever makes you think instead of letting the framework think for you. The tighter the constraint, the more your own problem-solving skills get exercised.

3. Cap it at one weekend. Real fun has an ending. If the project spills past Monday morning, you have slid back into product mode. Kill it without guilt and start something smaller and stranger.

The spark does not come from reading more HN threads about how coding used to feel. It comes from shipping something nobody asked for, that solves a problem only you have, on a stack that costs almost nothing and impresses nobody. The corporate version of computing got tiring. The other version is still around, and it costs about an afternoon.

🔗 Useful Tools

• Sri Lanka Working Days Calculator — the kind of single-purpose tool that is fun to build and useful to have
• Pomodoro Timer — cap your weekend project sessions properly; time constraints help more than unlimited hours
• Reading Time Estimator — useful when you are building a blog or documentation tool and need a simple content metric