I built a macOS app for a market that doesn't know it exists yet: developer English

The embarrassing moment that started everything

Two years ago I gave a lightning talk at a tech meetup. During Q&A, someone asked about my Kubernetes deployment setup.

I said "koo-BER-nets."

The room didn't laugh. Worse — there was a brief, polite silence. Then the moderator gently repeated back "Kubernetes" with the correct pronunciation: KOO-ber-NET-eez.

200 people heard that. I wanted to disappear.

That night I went down a rabbit hole. How do you actually pronounce nginx? (engine-X.) Is "cache" one syllable? (Yes, like "cash.") Does anyone else say "sudo" wrong? (Everyone.)

And then a bigger realization hit me: pronunciation was just the tip of the iceberg. I'd been writing "You should change this" in code reviews for years — not knowing that in English, "should" sounds prescriptive, almost like an order. The native way is "Consider changing this" or just "Nit: rename to X?"

Nobody tells you these things. Native speakers don't correct you because it feels awkward. Non-native speakers don't ask because they don't know they're doing it wrong. It's a silent tax on every developer working in English as a second language.

---

The market nobody is serving

Here's the thing: there are ~18 million professional developers worldwide who work primarily in English but aren't native speakers. That's developers in China, Japan, Korea, Germany, Brazil, India — basically everywhere except the US, UK, Canada, and Australia.

These developers can read docs and write code. Their English is "good enough." But "good enough" has a ceiling:

• You avoid speaking up in meetings because you're mentally translating
• Your PR descriptions sound robotic or overly formal
• You write "Please kindly review" when everyone else writes "PTAL"
• You dodge certain words in standups because you're not sure how to pronounce them

What tools exist today?

| Tool | Why it doesn't work |
|------|-------------------|
| Duolingo | Teaches "the cat is on the table," not "how to push back on a code review diplomatically" |
| Grammarly | Fixes grammar, not tone or register. Doesn't know PR conventions |
| Google Translate | Translation ≠ learning. Zero retention |
| ChatGPT | Works, but breaks your flow: new tab → write prompt → wait → copy. 30 seconds per lookup |
| Dictionary apps | Don't know that "You should fix this" sounds aggressive in a code review |

There's a gap between "English learning tools" (too generic) and "developer productivity tools" (assumes native English). DevGlish sits in that gap.

---

What I built

DevGlish is a macOS menu bar app. The interaction is dead simple:

1. Select text anywhere (Slack, GitHub, VS Code, email, browser)
2. Press ⌘⇧D
3. See an AI-powered learning card in 2 seconds

That's it. No app switching, no prompts to write, no new tab.

But the magic is in what the card shows:

For a word like "cache":

• Pronunciation: /kæʃ/ (rhymes with "cash")
• Common mistakes: "cash-ay" (it's not French), "catch"
• Etymology: from French "cacher" (to hide), but anglicized
• Chinese speaker tip: the /æ/ vowel doesn't exist in Mandarin — it's between 啊 and 诶

For a phrase like "You should change this":

• Tone warning: sounds prescriptive/commanding in English
• Alternatives by severity: "Nit: consider X" → "I'd suggest X" → "This needs to change: [reason]"
• Context: detected you're in GitHub → code review phrasing

For Express mode (Chinese → English):

• Input: 这次事故的根因是数据库连接池耗尽
• Output 1 (Slack): "Traced it to connection pool exhaustion"
• Output 2 (Email): "The root cause was database connection pool exhaustion, resulting in API timeouts"
• Output 3 (Incident report): "Root cause: database connection pool exhaustion. Impact: ..."

The app detects which application you're in and adjusts tone. Slack gets casual suggestions. Email gets professional alternatives. GitHub gets code review conventions.

---

The "aha" — L1 interference patterns

This is the feature that makes DevGlish hard to replicate.

Every native language creates specific interference patterns when speaking English. These patterns are predictable, consistent, and teachable:

Chinese speakers:

• Drop articles: "I found bug" → "I found a bug"
• Over-use "of": "the speed of loading of page" → "the page loading speed"
• Direct tone: "You should fix" → "Consider fixing"
• False formality: "Please kindly" → just don't

Japanese speakers:

• Over-hedge: "I think maybe we should perhaps consider..." → "I'd suggest..."
• False confidence killer: "I will try my best" → "I'll get it done" (try my best implies failure)
• L/R confusion in speech
• 和製英語 (wasei-eigo): "key person" → "key stakeholder"

German speakers:

• False friends: "aktuell" ≠ "actually" (it means "current")
• V/W confusion: "vine" → "wine"
• Word order: "I have yesterday the code reviewed" → "I reviewed the code yesterday"

These interference patterns are baked into every AI response. When a Chinese speaker looks up a word, they get Chinese-specific warnings. Japanese speakers get different warnings. Same word, different guidance.

No generic English tool does this because it requires knowing both the target language AND the source language's interference patterns simultaneously.

---

Tech decisions

For the IH crowd who cares about this stuff:

macOS native (SwiftUI + AppKit) — Not Electron. The app needs to float above other windows without stealing focus (NSPanel), read selected text from any app (Accessibility API), and respond to a global hotkey. These are OS-level capabilities that only work natively.

Cloudflare Workers backend — Edge-deployed, TypeScript, dirt cheap at scale. D1 (SQLite) for user data, KV for sessions and rate limiting. No servers to manage, no cold starts worth worrying about.

Claude API for AI — Powers the explanations, Express mode, and context-aware suggestions. Cost: ~$0.001 per lookup. At 1,000 lookups/day per Pro user, that's $0.03/user/day, well within margin at $7.99/month.

Local tech term database — 85+ terms with IPA pronunciation, etymology, common mistakes, and L1-specific tips. This runs entirely on-device with zero latency. No API call needed for the most common queries.

Landing page: Astro on Cloudflare Pages — Static, fast, simple. 13 language versions for international SEO.

---

Pricing and the free tier strategy

| Tier | Price | What you get |
|------|-------|-------------|
| Free | $0 | 10 lookups/day + full tech term pronunciation database (unlimited) |
| Pro | $7.99/month or $69/year | Unlimited lookups, Express mode, cloud sync, spaced repetition, L1 interference tips |
| Team | $12/person/month | Pro + team vocabulary analytics + admin |

The key insight: the free tier includes the entire tech pronunciation database with no limits.

Why give that away? Because "how do you pronounce Kubernetes" is the #1 Google search that leads developers to discover they need this tool. They come for the pronunciation, stay for Express mode.

The pronunciation database is local (zero marginal cost), but it's the hook that gets the app installed. Once someone uses Express mode to draft a PR description in 2 seconds instead of 2 minutes, that's when they convert to Pro.

---

Market sizing — is this real?

Here's my napkin math:

• ~18M professional developers working in English as L2
• Assume 10% have "good enough but not native" English (the sweet spot) = 1.8M
• Assume 5% are on macOS (conservative for developers) = 90K addressable
• Assume 2% conversion to try = 1,800 users
• Assume 10% convert to Pro = 180 paying users
• 180 × $7.99/month = $1,438 MRR as a baseline

That's just macOS. A Chrome extension or Windows version opens up the other 95%.

The Team tier is where it gets interesting. Companies with international engineering teams (basically every tech company) could expense this. One engineering manager buying it for a team of 10 = $120/month, equivalent to 15 individual Pro users.

I'm not trying to build a unicorn. I'm trying to build a profitable tool that solves a real problem for a specific group of people. $5-10K MRR would be life-changing for me.

---

What's working and what's not

Working:

• The Kubernetes pronunciation hook lands every time. I've tested it in conversations with 20+ developers — universal "oh shit, I've been saying that wrong" reaction
• Express mode is the stickiest feature. People who try it stop using ChatGPT for phrasing
• Context-awareness (detecting Slack vs GitHub) feels magical the first time

Not working yet:

• Spaced repetition engagement is low. People save words but don't review them. Might need to rethink the review UX
• Free-to-Pro conversion needs work. The 10 lookup limit might be too generous (or not generous enough?)
• Marketing is my weakest area. I'm a developer, not a marketer. That's partly why I'm writing this post

What I'm figuring out:

• Should I launch on Product Hunt now or wait until I have more social proof?
• Is the Chrome extension the right next platform, or should I do iOS first?
• How do I reach Japanese and Korean developers? The L1 interference data is ready, but I don't have distribution in those markets

---

The deeper thesis

English proficiency is a career multiplier for developers. Not just for getting hired at international companies — for being perceived as senior once you're there.

A developer who writes "I found a bug in the auth flow — looks like the session token isn't being refreshed on 401s" sounds more senior than one who writes "I found bug. Session token has problem."

Same skill. Same finding. Different perceived competence.

That's unfair. DevGlish is my attempt to level that playing field — not by teaching grammar (boring, and Grammarly exists), but by teaching the unwritten communication norms of English-speaking dev teams.

The patterns are finite. The interference is predictable. The improvement is concrete and measurable. You just need someone to show you what "native" sounds like in context.

---

Try it: https://devglish.com

I'd love your feedback:

1. Does this market make sense to you, or am I chasing a niche that's too small?
2. Pricing: is $7.99/month right for a developer tool?
3. What channel would you prioritize for early growth?

Happy to answer any questions about the tech, the business, or the embarrassing pronunciation stories.