Think Auto · pronounced tā, 他, “him” in Mandarin
TA is an AI assistant that lives in VS Code and your browser. He reads foreign-language docs for you, reads them aloud, explains errors and walks through code with you, in your own language. No switching windows, no copy and paste, and he learns more about you every day.
Sample content prepared for this demo. Read-aloud uses your browser’s built-in voices.
More than 180 million developers build on GitHub, and roughly 5 in 6 of them are outside the United States.1 They are fluent in code, but every doc, error message, issue thread and meeting adds a small language tax. It never shows up in a sprint report, yet it’s paid every day.
A new library. The README is in English, the best issue thread is in Japanese. Two browser tabs and a translation site later, back to the editor.
A red stack trace. The words are clear one by one, but what it actually means takes ten minutes of searching.
Writing a PR description. Drafted in Chinese, pasted into a translator, fixed by hand so it doesn’t sound machine-made.
Stand-up with the overseas team. Not sure whether it’s “ku-ber-NEET-ees” or “koo-ber-NET-eez”, so the word gets avoided.
The tools exist, but they live in other windows, they don’t understand code, and they forget you every time.
Translation sites mangle code and terminology. AI chat assistants answer when asked, in English first, and start from zero in every conversation. Nothing sits where the reading actually happens, speaks the words out loud, and remembers who you are.
TA isn’t another window to switch to. He sits right where you’re reading, and speaks up when you pause.
Hover over an English comment, a README or a Japanese issue and TA translates it into your language. Terms like map() and Promise stay untouched, and code is never translated.
Any text, read in English, Japanese, Korean, Spanish and more. Listen when your eyes are tired, or rehearse with him before a meeting.
When the terminal turns red, TA tells you in plain words what happened, which line is most likely the cause, and what to try next.
Draft commit messages, PR descriptions and issue replies in your own language. TA rewrites them in natural engineering English and lets you compare before you send.
Taking over someone else’s project? Select a block and ask “what does this do?” TA explains it step by step, in your language.
Your languages, the words you keep looking up, how much detail you like. TA remembers, so you never start from zero.
Most translation tools treat you like a stranger on every request. TA remembers your habits and your projects, so his explanations land at the right level. Memories stay in your account, and you can view or delete them anytime.
Write mostly TypeScript and Python at a startup in Taipei.
Explanations in Traditional Chinese, tech terms kept in English.
Asked about debounce three times this week, so today he leads with an example.
OrderSync is your team’s order service. Never translate it.
Giving a tech talk in English next month. Practice five words a day.
Frontier models like Claude now translate technical text accurately, keep identifiers intact, and explain an error instead of just converting its words.
Speech synthesis now sounds human in dozens of languages, and is cheap enough to read every doc and comment aloud on demand.
Long context and memory let an assistant keep a lasting picture of one developer: their languages, projects and habits. That turns a tool into a companion.
TA does one job the others skip: he sits inside the reading, speaks it, and knows who he’s talking to.
| TA | Translation sites | AI chat assistants | |
|---|---|---|---|
| Works where you read, without asking | ● Yes | ○ No | ◐ When asked |
| Leaves code and tech terms intact | ● Yes | ○ Often not | ◐ Mostly |
| Reads aloud in many languages | ● Yes | ◐ Basic | ○ Rarely |
| Teaches how to say tech terms | ● Yes | ○ No | ○ No |
| Explains errors in your language | ● Yes | ○ No | ◐ English first |
| Remembers you across days | ● Yes | ○ No | ◐ Limited |
● built in · ◐ partly · ○ not available. TA column describes the product we are building.
Our home market, starting in Taiwan. We speak the language, know the communities, and can talk to every early user in person. VS Code is used by 75.9% of developers,2 so one extension reaches most of them.
Large, English-heavy tech industries with the same daily friction. The product is the same; the voices and glossaries change.
From Latin America to Southeast Asia, this is most of the world’s developers, and the share keeps growing as new developers come online.
These are the ones that trip people up in meetings, interviews and screencasts. Tap to listen. In the editor, TA spots words like these in your own code.
Pronunciations follow American English. Some terms have more than one common reading.
Docs and error messages are almost all in English. Remove the language barrier and focus on the concepts.
Read your teammates’ issues and write your own PRs without pasting everything into a translation site.
Before a talk, an interview or a stand-up, hear it once first. Say Kubernetes and nginx with confidence.
A generous free tier lets anyone try TA inside their editor. People who rely on him every day, and teams who want a shared vocabulary, pay a monthly subscription.
For trying TA
For individual developers
For global engineering teams
Plans are planned, not yet on sale. We will set pricing together with our first users.
We would rather have fifty developers who can’t work without TA than five thousand who installed it once.
We onboard testers one by one, watch how they actually read and write, and let what they do, not what they say they want, decide what we build next.
Each release does one thing better. Speed of learning matters more than polish in the early days.
TA sends only the text you point at, never trains models on your code, and lets you see and delete everything he remembers.
Hover translation, error explanations, read-aloud and native-language-to-English writing. The four things developers need every day, done well first.
Paste a docs URL or any text to get a side-by-side bilingual version that keeps code blocks intact, with full read-aloud.
Docs, terminal, code review and team chat, all in your language. TA knows your whole project and becomes a real teammate inside your workspace.
Before TA, we built two VS Code extensions for test-automation engineers. Both are free to install today.
Static analysis, navigation and an AI reviewer for Tcl and Expect test scripts.
tclsh on 3,000 randomized tests@tcl chat expert and a Claude Code subagent that know your own framework procs43 unit tests · 20 end-to-end checks
Use a serial port on another machine as a local COMx, so existing test scripts run unchanged.
18 unit · 8 integration · 7 extension tests
“I build with AI agents every day, in a language that isn’t my first. TA is the friend I wanted next to me in the editor.”
他Austin Chang, founder · Taipei
At launch: Traditional Chinese, Simplified Chinese, Japanese, Korean and Spanish, translating from and into English. Read-aloud works in any language your system or our cloud voices support.
Only the text you hover or select is sent for translation. TA never uploads whole files on its own, never trains models on your code, and you can turn memory off at any time.
VS Code first, including editors built on it. A web app for docs and articles comes next. Other editors depend on what our first users ask for.
Copilot helps you write code. TA helps you understand and communicate around it: reading, listening, explaining and speaking. They work well side by side.
We are opening a private test for a small group first. Join the waitlist and tell us how you work. We reply to every email personally.
Email us the languages you code in and the kind of foreign-language content that slows you down most. You’ll be the first to know when the extension opens for testing.