Part 10 of 10

Ship It

What I got right, what I got wrong, and what comes next.

For everyone·12 min read

For everyone.

The Full Arc

FIG 10.1

Seven weeks from dark mode quest to shipping 1.0.0-alpha.1 on three platforms

On February 7th, a player named AnszaKalltiern filed a crash report. EVEMon — the character planner that EVE Online pilots had relied on for twenty years — was crashing at sixty characters.

Eighty-four days later, EveLens shipped: 1,800 tests, seven assemblies, a priority-queue scheduler, forty-six ViewModels, six dark themes, a GPU-accelerated skill constellation, and native builds for Windows, Linux, and macOS.

Every thread I pulled revealed the next. Crash → scheduler → god object → monolith split → event system → ViewModels → framework replacement → cross-platform → community feedback → stable release.

I didn't plan to rebuild an entire application. The codebase left me no choice.

What I Am

I'm not a software engineer. I don't have a computer science degree. I don't have engineering training of any kind.

I built EveLens using Claude Code — Anthropic's AI coding tool. Every commit in the repository was assisted by AI. I used multiple agents in parallel: architecture analysis, implementation, validation, testing. I cross-validated their findings. I made the architectural decisions. I directed the implementation order. I reviewed the output.

The AI turned my vision into working software at a speed that shouldn't be possible. But AI couldn't decide what to build. It couldn't decide when the architecture was clean enough. It couldn't decide to credit every contributor in the About window. Those decisions were mine.

This wouldn't be possible without AI. The idea that a single person can restructure a twenty-year-old codebase, build a new UI framework, ship cross-platform, and write 1,800 tests — it's not possible alone. AI didn't replace the need for a developer. It replaced the need for a team.

What I Got Wrong

Scope Creep

What started as dark mode became a full rebuild. Should have shipped incremental value earlier.

Testing Too Late

Wrote most tests after the fact. Test-first would have caught more regressions during the migration.

Solo Development

One person, one vision. Faster decisions, but no code review, no pair debugging, no second opinion.

I should have shipped sooner. I spent weeks building a perfect foundation before giving users anything. Every individual decision was technically justified. The aggregate effect was nothing shipped for two months. The community was patient, but they shouldn't have had to be.

Features that didn't make it. I'm being transparent:

Booster and accelerator simulation was attempted but didn't work. Not ported.
Mass re-authentication was worked on but never shipped.
Three data tabs had code behind them but weren't wired into the UI at launch.
That's six features that didn't ship out of thirty-eight total. I could have hidden this. I'm choosing not to, because the moment someone finds an overclaim, the whole thing loses trust.

What I Got Right

Ship it. The perfect is the enemy of the good. EveLens isn’t done — but it’s real, it’s useful, and it’s in people’s hands.

Architecture First

Fixing the foundation before adding features. Every subsequent change was easier because the boundaries were clean.

Test Everything

1,554 tests. 14 architectural laws enforced by code. No regression survives.

Ship Imperfect

The perfect is the enemy of the good. EveLens shipped with 6 features unported — and that’s fine.

Cross-Platform

One codebase, three platforms. The Avalonia bet paid off. Linux and macOS support for free.

Architecture first. The fourteen laws and assembly boundaries are what made the Avalonia port possible. Without clean separation, I'd still be fighting circular dependencies. The ViewModel layer built for Windows Forms carried over to Avalonia unchanged. The skill calculations still work exactly as they did when Jimi Charalampidis refined them. Peter Han's ESI handling still works.

Community-driven development. After shipping alpha, AnszaKalltiern — the same person whose crash report started this whole thing — became the most active tester. Seventy characters. Every build tested at scale. BritishDragonAdmin pushed UX improvements. NotmoGit caught data bugs. SmokeMeAKipper-eve did a security review. Every feature in 1.1.0 traces back to someone who asked for it.

Crediting everyone. The About window names every contributor from the project's twenty-year history. Twenty retired developers. Five consultants. Five community contributors. The original creator. The copyright line reads: Copyright 2006-2021 EVEMon Development Team, 2025-2026 Alia Collins.

Transparency. Claude Code is mentioned in the About window. Stated openly in this blog. Being upfront lets me focus on what matters — does the product work?

What's Not Done

Honest accounting:

122 remaining references to the old god object — being migrated
637 references to the static settings system
No true dependency injection container — AppServices is a static facade
The "Monitored" character concept is vestigial and needs removal
Update dialog doesn't show release notes yet

What This Means

EVEMon was created in August 2006. For twenty years, through three generations of maintainers, it was the tool every serious EVE player opened alongside the game client. Jimi Charalampidis contributed 5,886 commits. Richard Slater added 1,494. BradStone started it with 526. Peter Han saved it during the ESI migration with 328 commits.

When a tool that the community relied on for two decades stops evolving, the community doesn't just lose a utility. They lose institutional knowledge, proven patterns, and a reference implementation that other developers learned from.

EveLens preserves that knowledge while rebuilding the foundation. The skill calculations still work. The plan optimizer still works. The ESI integration still handles the same edge cases. But now it runs on Linux. And Mac. And has 1,800 tests making sure none of it breaks.

What Comes Next

1.2.0 is the Skill Farm Dashboard. ISK tracking per character. Extraction readiness. Monthly economics. Whether your accounts are paying for themselves.

Every feature request, every bug report, every "hey this thing is broken" shapes what gets built next. That's not a platitude — you can trace every feature in 1.1.0 to a specific person who asked for it.

The code is open source under GPL v2. The git history is public. The architectural laws are enforced by tests. Anyone can contribute.

The Lineage

EveLens was built on the foundation created by Six Anari in 2006 and maintained by:

Jimi Charalampidis — 5,886 commits (2010-2016)
Richard Slater — 1,494 commits (2007-2009)
BradStone — 526 commits (2006-2008)
Peter Han — 328 commits (2018-2021)
53 other contributors — ~600 commits across all eras

Their work is the bedrock. The About window names every one of them.

EveLens is free, open source, and available on Windows, macOS, and Linux.

evelens.dev | github.com/aliacollins/evelens

Eighty-four days. One crash report. An entire application transformed.

Fly safe. o7

— Alia Collins

This is Part 10 of a 10-part series. Start from the beginning →