Rebasely Beta: data may be purged

justintypescript-ultra-strict Public

Markdown · 33.4 KiB · at 503e839

TypeScript-Ultra-Strict

This is my proposal for TypeScript-Ultra-Strict: a stricter TypeScript for a world where AI writes most of the code.

This repo is a thought experiment, not a production tool. The compiler works, type-checks, and every example passes, but the macro syntax is lowered by a preprocessor with known edge cases, the bundler plugins skip semantic checking, there are no published packages, and there is no stability promise. It exists to make the proposal concrete enough to argue with. Do not build anything you care about on it yet.

Status: working prototype. This branch adds v2, which replaces the macros and the new extension with plain .ts, a checker and a small type library, and adds security types. The v1 description follows it.

v1 status: packages/cli implements tsus check and tsus build: a tokenizer built on ts.createScanner, a preprocessor that lowers match and the ? operator with a source map back to the .tsus file, and a full ts.createProgram type-check over the project. match is exhaustiveness-checked at compile time, tsus build emits .js.map files that point at the .tsus source, packages/runtime implements Option/Result, the typed match helpers and the boundary layer, packages/plugin provides webpack/Vite/esbuild/Bun adapters, the checker enforces 18 strictness rules with machine-applicable fixes (tsus fix, tsus explain, tsus migrate, tsus init; see For agents), and all 57 projects under examples/ pass, 53 for v1 and 4 for v2 (npm install && npm test). LEARNINGS.md documents what broke along the way, how that shaped the design, and the known limitations.

TypeScript's early design decisions assumed humans would write and maintain it, so a lot of features exist to make a human's life easier. We valued speed over quality. With AI writing most of our code, we can flip that tradeoff and force the language to be declarative and verifiable, like Rust.

v2: plain .ts, a checker, and security types

v2 lives on the v2 branch. It keeps what v1 proved and drops what it cost. Everything is ordinary, erasable TypeScript: no macros, no .tsus extension, no compile step. A file that uses v2 is a .ts file that tsc, esbuild, wrangler and node's own type stripping handle unchanged (example 61 runs the same file through an esbuild bundle and through node main.ts; the root tsconfig.json checks the v2 files with erasableSyntaxOnly). v2 is three pieces:

  • @tsus/v2 (packages/v2), a type library with a small runtime: Option, Result, match, Sql, Permit, Secret, Untrusted.
  • tsus check --v2 [dir], a checker over the project's own TypeScript program (its tsconfig, plus strict and noUncheckedIndexedAccess). Codes TSUS500 to TSUS607, documented with the reason each one exists in docs/v2/rules.md and by tsus explain.
  • tsus audit <file.ts...>, a report-only pass over existing plain TypeScript.

What the experiment taught, and what v2 does about it

On a new feature built from a detailed spec (the search query language, judged on 118 hidden tests and 20,400 fuzz pairs), v1 tied plain strict TypeScript on correctness and cost 28% more tool calls, 65% more code, a 51-line adapter and workarounds such as parts.slice(0, 1).join("") for parts[0]. Its checker reported nothing in 8 runs. In the on-call pilot, the Option and Result wrapper objects cost 2 to 6 times on the hot path, while rewriting existing logic exposed real bugs. And the risks that reviews of AI-written web code kept finding were SQL injection, permission scoping, secrets and untrusted input, for which v1 had nothing.

Lesson v2
Wrappers cost 2 to 6x Option<T> is T | undefined; Result<T, E> is { ok: true, value } | { ok: false, error }; match is one lookup and a call
The adapter was mostly null conversion Comparing with null is allowed, producing it is not (TSUS502); safeDb turns D1's SQL NULL into undefined
Banning undefined and ! banned indexing noUncheckedIndexedAccess and an === undefined check
Friction rules caught nothing Dropped: no class, var, throw, enum, ==, default export, parameter reassignment, unused let, tagged-errors-only. Each kept or dropped rule is listed with its evidence in docs/v2/rules.md
Exhaustive matching was the one prevention reviewers credited Kept: match() and matchLiteral() are typed so a missing arm is a type error, and TSUS501 checks switches
No security model Sql, Permit, Secret, Untrusted, with checker rules for what the types cannot express
Rewriting found bugs tsus audit lists where the bugs hide in code nobody is rewriting

The library

TypeScript
import { match, matchLiteral, matchResult, Ok, Err, sql, safeDb, html, fence, secret, reveal, untrusted, validate } from "@tsus/v2";
import type { Option, Result, Sql, Permit, Secret, Untrusted } from "@tsus/v2";

const area = (s: Shape): number =>
  match(s, "kind", { circle: (c) => Math.PI * c.r ** 2, square: (q) => q.side ** 2 }); // a missing arm is a type error

const rows = await safeDb(env.DB).all<IssueRow>(
  sql`SELECT number, title FROM issues WHERE repo_id = ${permit.resource} AND title LIKE ${pattern}`); // values are bound params

const page = html`<li>${issue.title}</li>`;                     // Untrusted text is escaped
const prompt = `Summarize:\n${fence(issue.body, "issue body")}`; // marked as data for the model
const headers = { authorization: `Bearer ${reveal(apiKey)}` };   // the one explicit way out of a Secret
  • Sql is a branded type that only the sql tag makes; every interpolation becomes a ? and a bound parameter, nested Sql splices in, sql.ident takes only a string literal type. SafeDb accepts only Sql, so a query built with + or an untagged template is a compile error (TS2345, reported as TSUS601). For raw D1 calls the checker flags prepare(text) whenever the text is not a compile-time constant.
  • Permit<"repo.read", RepoId> is an opaque capability. Only issuePermit from @tsus/v2/authority makes one, and only files named *.access.ts may import it (TSUS602). A data-access function that reads private data takes a Permit, so calling it without the check is a missing argument and a hand-made object is a missing brand (TSUS600). It should read the id from permit.resource; a separate id of the same type is flagged (TSUS603), since that is how a check on repo A guards a read of repo B.
  • Secret is not a string. reveal() is the explicit way out. The checker rejects a Secret, or an object containing one, in templates, +, String(), JSON.stringify, console.* and other log calls, URLs, fetch and Response (TSUS604). At runtime it prints as [secret] as a backstop.
  • Untrusted is the raw string at runtime (marking costs nothing) with a type that is not a string. It reaches HTML through html/escapeHtml, SQL through sql, a prompt through fence, and code that needs a plain string through validate(text, /pattern/). Untagged templates and concatenation are rejected by context: HTML (TSUS605), SQL (TSUS601), an LLM prompt (TSUS606, a template assigned to a name like prompt, system or messages, or passed to a model call), anything else (TSUS607). fence wraps the text as <untrusted source="..."> and neutralizes the closing tag and role markers inside it, the markers Rebasely's injection screen looks for.
  • Files named *.boundary.ts may cast, use !, any and null, and call raw drivers: they are where outside data is checked and converted. Their names make them reviewable.

Proof: the same module twice, with planted bugs

examples/60-v2-security-demo is a Workers-style module (D1 queries, a permission check, a webhook secret, user text into HTML and into an LLM prompt) written in plain strict TypeScript (plain/) and with v2 (v2/), with the same six bugs planted in both, and a fixed v2 version (fixed/). The runner asserts all of it: plain passes tsc --strict --noUncheckedIndexedAccess; v2 reports every planted bug with the expected code and nothing on any other line; the fixed version passes, bundles with esbuild, and a harness runs the attacks against both versions on the same in-memory SQLite.

Planted bug plain strict TS tsus audit on the plain file v2 (tsus check --v2) At runtime, plain / fixed v2
1. Search text pasted into the query nothing string-built SQL TSUS601 ' OR 1=1 -- returns the private repo's rows / no rows
2. Sort column pasted into the query nothing string-built SQL TSUS601 a CASE WHEN sort leaks who owns a private repo / ignored
3. List handler skips the permission check nothing unchecked permission TSUS600 (TS2345) private rows listed signed out / 404
4. Webhook key in a log line nothing nothing TSUS604 key in the logs / not
5. Issue titles into HTML unescaped nothing nothing TSUS605 <script> served / escaped
6. Issue body into an LLM prompt unmarked nothing nothing TSUS606 <system> reaches the model / fenced and neutralized

The audit finds the two classes it can see without types; the other four need to know which values are secrets and which are user text, which is what the v2 types add.

Runtime

npm run bench rebuilds the on-call micro benchmark from the pilot (bench/oncall; the pilot's own scripts were not committed, so the event generator is new and absolute numbers differ from the pilot's). The v1 rules are the pilot's compiled rules.js and runtime, unchanged. A differential check first confirms all versions agree on 20,000 events. Node 24.17, darwin arm64, median of 7 runs (raw output in bench/oncall/RESULTS.txt):

Path before (plain .ts) v1 v2
jobFor, per event (page of 1,000) 310 ns 523 ns (1.69x) 348 ns (1.12x); 339 ns (1.09x) with the rank as a switch
dedupe test, per job 8.3 ns 56.9 ns (6.88x) 7.6 ns (0.92x)
absent-or-value, 1,000 reads 775 ns (x ?? 0) 14,368 ns (fromNullable(x).unwrapOr(0)) 1,115 ns (fromNullable(x) ?? 0)
4-way literal match, 1,000 calls 934 ns (switch) 47,445 ns (match macro) 13,813 ns (matchLiteral, arms inline); 6,016 ns (arms hoisted)

Option is free. match() is not: its arms are closures, so a hot loop should use a switch, which TSUS501 checks for exhaustiveness all the same. About half of what is left on jobFor is tryCatch (a closure and an Ok object per event) around JSON.parse: with a plain try/catch there, v2's jobFor measured 326 ns (1.05x) in the same run.

tsus audit

tsus audit <file.ts...> [--json] [--scope dir] [--permission-fn name,...] reads plain TypeScript with the project's own tsconfig and writes nothing. It ranks: string-built SQL (a query whose text is not a compile-time constant, or a template that assembles SQL text from non-constant parts), unchecked permission paths (a database read reachable from an entry point, through the call graph of the package, with no call to something named like a check before it), ignored promises, swallowed errors, non-exhaustive switches over unions, implicit null paths (index reads that noUncheckedIndexedAccess would reject, truthiness tests on optional strings and numbers, !) and casts. Every finding is a lead with the reason it was flagged; --json also lists each database read it considered checked and the check it found. Example 63 pins the findings on a legacy file and asserts the file is unchanged.

Run read-only against Rebasely (origin/main 438e678a, apps/web/src/search/.ts and apps/web/src/triage/.ts), it reported 17 and 65 findings. Of the 82, 7 were true positives, all in triage: per-repository Actions failures in the Inbox are dropped without marking the result partial (inbox.ts:184), and six failures are swallowed without a log while the calls next to them log theirs (inbox.ts:222-224, handlers.ts:54-55, handlers.ts:101). The other 75 were false positives: 3 wrong (two SQL fragments whose column names are literals at every call site, and one data read reached through a dispatch table) and 72 correct about the pattern but harmless on inspection (bounds-checked index reads, documented fallbacks, safe casts). The search module, which the experiment's judge had already fuzzed, had none. LEARNINGS.md has the breakdown.

Limits

  • The security rules are as good as the marking: text is Untrusted because a boundary said so, and a value that never got marked is a plain string. The prompt rule recognizes prompts by name and by call; a prompt built under another name falls to TSUS607.
  • The permission rule in v2 is a type, so it holds wherever the data-access function is the only way to the rows. Raw D1 calls in application code are still possible; TSUS601 catches their string-built queries, not their missing checks. The audit's permission analysis is a name-based heuristic with call-site ordering, and it cannot follow calls through dispatch tables.
  • match() costs closures; use a switch on hot paths.

v1: what changes

  1. Removes null and undefined. In their place, Option<T> and Result<T, E> become first-class types. This is the centerpiece: the removals below only work because these replacements exist.
  2. Removes classes. Data is plain objects; behavior lives in functions.
  3. Removes var.
  4. Introduces an SDK for macros, used for pattern matching and variable destructuring over Option, Result, and tagged unions. (TC39 already has a pattern-matching proposal in the pipeline; this accelerates where the language is heading.)

The interop boundary

Every existing DOM and npm API returns undefined, throws, and hands you classes. Ultra-Strict code never sees those directly. At the boundary, a checked bindings layer converts T | undefined and T | null into Option<T>, and throwing calls into Result<T, E>. Inside the boundary, the strict rules hold everywhere. In source, the boundary is the @tsus/runtime/boundary helpers (fromNullable, wrapSync, wrapAsync) plus any file named *.boundary.tsus, the only place casts and default exports are allowed.

Why not just an eslint preset?

Banning classes, nulls, and var can be approximated with tsconfig and lint rules today. Two things need a real compiler:

  1. The macro SDK. This is the honest cost of the proposal: TypeScript's types are fully erasable, which is why esbuild and swc can strip them in a single fast pass. Macros break that property. The bet is that with AI writing the code, we should spend build-time budget on expressiveness and checkability rather than on keeping type-stripping trivial.
  2. The boundary layer, which needs type information to generate the Option/Result wrappers.

Why TypeScript?

Back in 2008, Google shipped Chrome and V8, and the industry built its edge compute on top of that runtime. Cloudflare Workers runs V8 isolates directly. Node, Deno, and Bun's ecosystem all trace back to it. JavaScript and ESM are the substrate that's already deployed everywhere.

TypeScript is the typed layer the ecosystem already adopted. Rather than asking everyone to move to a new language, Ultra-Strict is a subset-plus-macros of a language AI models already know deeply, targeting a runtime that's already everywhere.

Ideal UX

Any file with a .tsus extension (TypeScript Ultra Strict) is processed under the strict rules and compiled to plain JavaScript. .ts and .tsus files coexist in one project, so adoption is per-file: an AI agent can write new modules in .tsus while the legacy .ts code keeps working untouched. Importing a .tsus module from .ts just works; importing .ts (or any npm package) into .tsus goes through the boundary layer, which rewrites the types so nullable returns arrive as Option<T> and throwing functions arrive as Result<T, E>.

What a .tsus file looks like:

TypeScript
import { readFile } from "node:fs/promises"; // boundary-wrapped: returns Result

type User = { name: string; email: Option<string> };

const loadUser = async (path: string): Promise<Result<User, IoError>> => {
  const raw = await readFile(path, "utf8")?; // ? returns the Err early
  return parseUser(raw);
};

// macro from the SDK; leaving out an arm is a compile error
const greeting = match (user.email) {
  Some(email) => `Reach me at ${email}`,
  None => "No email on file",
};

How the macros compile

match lowers to a call whose arms are an object keyed by variant tag:

TypeScript
match (opt) { Some(x) => a, None => b }   // -> __tsus_match(opt)({ Some: (x) => (a), None: () => (b) })
match (opt) { Some(x) => a, _ => b }      // -> __tsus_match_or(opt)({ Some: (x) => (a) }, () => (b))
match (path) { "/" => a, _ => b }         // -> __tsus_match_lit(path)([["/", () => (a)]], () => (b))

@tsus/runtime types that object with a mapped type over the subject union's tag field, so TypeScript itself requires every variant (unless there is a _ arm), rejects arms the union cannot hold, and narrows each arm's binding to its variant's payload. This works for Option, Result, and any user union with a string tag. Literal patterns lower to a list of [literal, handler] pairs; each literal must be a value of the subject's type, and _ is required unless the literals cover a finite union like "GET" | "POST". When the match has a contextual type (a declared return type, an annotated const, a typed parameter), every arm is checked against it, so Err({ kind: "missing" }) infers the declared error union and a wrong arm is reported on that arm; with no context, the match's type is the union of its arms. If any arm awaits, every arm becomes async and the match is a Promise; await inside a string or comment does not count.

expr? works like Rust's: on an Err or None it returns that value from the enclosing function, otherwise it unwraps.

TypeScript
const port = parsePort(raw)?;
// -> const __tsus_t0 = __tsus_try(parsePort(raw)); if (__tsus_isBail(__tsus_t0)) return __tsus_bail(__tsus_t0); const port = __tsus_t0.value;

The enclosing function must declare a Result, Option, or Promise of one as its return type, and the propagated error must fit it. In async code ? applies to the awaited value (await load()?). ? binds like the conditional operator, so it covers everything to its left: write (a()?) + (b()?), not a()? + b()?. Optional chaining (?.), ??, ternaries, and optional properties are untouched.

Diagnostics

Code Meaning
TSUS000 Syntax error after macro expansion
TSUS001-TSUS005 Banned class, null, undefined, var, throw
TSUS006 Malformed match (bad pattern, mixed variant and literal patterns, duplicate arm)
TSUS100 Runtime only: a match with no arm for its value (only reachable if the checker was bypassed)
TSUS200 Any TypeScript error, mapped to the .tsus position; tsCode holds the TS code
TSUS201 Non-exhaustive match: missing variants, or literal arms with no _
TSUS202 A match arm for a variant or literal the subject cannot hold
TSUS203 A match subject that is not a tagged union or a literal type
TSUS300 ? outside a function declaring a Result/Option return type (or at top level, or in a match arm)
TSUS301 ? applied to something that is not a Result or Option
TSUS302 ? propagating an Err/None the function's return type cannot hold
TSUS303 ? in a position that may not run (?: branch, right of &&/||/??, loop condition)

Every TypeScript-derived code (TSUS2xx, TSUS301, TSUS302) carries the original TS code as tsCode in tsus check --json.

Framework integrations

packages/plugin (@tsus/plugin) is one transform core with a thin adapter per host, all sharing the CLI's preprocessor and checker:

  • @tsus/plugin/webpack: a loader. In Next.js, one rule in next.config.mjs makes .tsus importable from any page or component (examples/27-nextjs, verified by next build emitting the rendered string into static HTML).
  • @tsus/plugin/vite: a Vite plugin, which also covers Astro, SvelteKit, and Nuxt through their vite.plugins config (examples/28-astro).
  • @tsus/plugin/esbuild: an esbuild plugin. Bun's plugin API is esbuild-shaped, so the same object registers with Bun.plugin; with a one-line bunfig.toml preload, bun run executes .tsus imports natively with no build step (examples/29-bun).

The adapters deliberately leave import specifiers untouched: the host bundler resolves .tsus to .tsus chains through its own pipeline and each hop hits the transform again. The tsus CLI stays the standalone path (and the CI path, via tsus check --json).

Tooling:

  • tsus build compiles to JS with source maps, tsus check type-checks without emitting. Both run as an esbuild/swc plugin so existing bundler setups pick up .tsus files without config changes. (Today the plugins run the syntactic rules only; the semantic type-check lives in tsus check.)
  • Editor support ships as a TypeScript language service plugin, so VS Code and every TS-aware editor get diagnostics, go-to-definition, and macro expansion previews for free.
  • Diagnostics are machine-first: every error has a stable code, a JSON output mode, and a suggested fix, so an agent can consume tsus check --json in a loop without parsing prose. Human-readable output is a rendering of the same data.
  • No new package manager and no new registry. npm packages work through the boundary layer, and published .tsus packages ship compiled JS plus .d.ts files, so consumers don't need to know the source language existed.

Evidence

evals/ holds a harness that tests the thesis directly: a model writes the same 30 small programs (config parsing, an LRU cache, retries with typed errors, a state machine, date ranges, an inventory ledger, and so on) in .ts and in .tsus, the compiler's errors go back to it for up to three repair rounds, and hidden tests the model never sees decide which version is correct. The tasks were chosen because null, undefined, exceptions and unhandled cases cause real bugs in them. Every task ships a reference solution in both languages and a plausible buggy solution that the hidden tests must catch (node evals/validate.mjs).

Results are in evals/RESULTS.md. So far that file holds only the harness self-test, which runs the reference solutions through the pipeline and says nothing about the thesis. No model run has been recorded yet; node evals/run.mjs --pilot runs the 10-task pilot once the Claude Code CLI is logged in or ANTHROPIC_API_KEY is set. Even then, one model, one sample per task and 10 to 30 tasks is a pilot, not proof. evals/README.md lists the caveats, starting with this one: results recorded before the compiler foundation landed came from a tsus checker that did not type-check, so .ts got more compile-time help than .tsus; tsus check now runs a full TypeScript type-check too.

Known limitations

The four limitations the MVP shipped with are closed: tsus check type-checks the whole project with ts.createProgram, a non-exhaustive match is a compile-time TSUS201, await in arms is detected on the parsed AST, and the tokenizer is TypeScript's own scanner with regex and template rescanning. What is still rough, roughly in order of how likely it is to bite:

  1. The bundler plugins (@tsus/plugin/*) transform one file at a time and only run the syntactic rules. Type errors surface in tsus check, not in next build or vite build.
  2. tsus check uses fixed compiler options (strict, ES2022, bundler resolution, DOM and Node types) and ignores any tsconfig.json.
  3. ? binds like the conditional operator, so a()? + 1 is a syntax error; write (a()?) + 1. It cannot appear in a match arm (arms are closures), a ?: branch, the right side of &&/||/??, or a loop condition. Its return-type check is syntactic: the declared type must be spelled Result<...>, Option<...>, or Promise<...> of one, not an alias.
  4. Regex detection is a one-token lookbehind: after ) or } a / is read as division, so if (x) /re/.test(s) misparses. Commas in an arm body's type arguments are handled by a heuristic, and an arm body cannot be a statement block.

For agents

The language is designed to be written by models, so the toolchain treats an agent as its primary user.

Drop-in primer. AGENTS.md is a one-page description of the whole language (types, match, ?, the boundary, every diagnostic code) meant to be copied into any project that uses tsus. tsus init does that for you.

Commands.

Command What it does
tsus check [dir] --json Every diagnostic has a stable code, start/end offsets and line/col ranges in the original .tsus file, a hint, an explain pointer, and a fix (or null)
tsus fix [dir] Applies every safe fix, re-checking between passes. --unsafe adds fixes that can change behavior, --code TSUS405 limits it to one rule, --dry-run reports without writing
tsus explain <CODE> The rule, why it exists, a rejected and an accepted example, for every code: the strictness rules, the macro and type-check codes (TSUS000, TSUS006, TSUS200 to TSUS203) and the ? codes (TSUS300 to TSUS303). --list prints every code, --json the raw entry
tsus init [dir] Scaffolds AGENTS.md (plus a CLAUDE.md that imports it), a strict tsconfig.json, package.json and a starter src/main.tsus. Never overwrites; --force refreshes only AGENTS.md
tsus migrate <file.ts> Codemod: T | null and T | undefined returns become Option<T>, throwing functions return Result<T, E> with a literal or tagged E, simple classes become a data type plus functions, var becomes const or let. Writes <file>.tsus next to the input and reports, with line numbers, everything it did not convert (including whatever the checker still rejects in the output)

A fix is { description, safe, edits: [{ start, end, line, col, endLine, endCol, newText }] }, where offsets index the original file as UTF-16 code units. "Safe" means runtime behavior is unchanged: == to ===, let to const, var to const/let when nothing depends on hoisting, any to unknown, deleting a suppression comment. Unsafe fixes ("safe": false) are offered where a mechanical rewrite exists but intent matters: an enum becomes a union of literals with its members inlined, a default export becomes a named one, a floating promise gets await, param.sort() becomes param.toSorted().

Rules beyond the original five. Each has a stable code, a rejection example, and tsus explain text, all generated from one registry (packages/cli/src/rules/registry.js). npm run docs regenerates docs/rules.md and the code table in AGENTS.md; npm run docs:check (part of npm test) fails if they drift or if a rule has no example.

Code Rule Fix
TSUS401 no any safe: unknown
TSUS402 no as casts, except as const and in boundary files none
TSUS403 no non-null ! none
TSUS404 no enum unsafe: union of literals
TSUS405 no == / != safe: === / !==
TSUS406 no default exports, except in boundary files unsafe: named export
TSUS407 no reassigning parameters none
TSUS408 let that is never reassigned safe: const
TSUS409 no mutating parameters (property writes, array mutators, Map/Set mutators, Object.assign) unsafe for sort/reverse/splice
TSUS410 Result error types must be string literal unions or tagged unions none
TSUS411 no floating promises (including void p and forEach(async ...)) unsafe: await
TSUS412 no unused Result none
TSUS413 no @ts-ignore, @ts-expect-error, @ts-nocheck safe: delete

The rules run on the macro-expanded source with the TypeScript parser and a per-file binder, so names resolve through real scopes (a shadowed parameter is not a parameter). TSUS410 to TSUS412 follow relative .tsus and .ts imports to see what an imported function returns or what an imported error type is; when a full ts.Program is handed to the checker, they use its types instead.

Boundary files. A file named *.boundary.tsus is where the outside world is let in: as casts (after a runtime check) and default exports are allowed there, and nothing else is relaxed. Cloudflare Workers' export default { fetch } lives in one (examples/23-cloudflare-worker), as does a checked JSON config loader (examples/53-boundary-file).

The examples

examples/ is the test suite: 53 small projects, each with an example.json declaring what must happen. npm run examples builds and executes all of them and fails on any drift (npm test also checks that generated docs are current).

Core language and runtime

Example Shows
01-hello Smallest possible .tsus file compiling to plain ESM
02-option-basics Option<T> with map/andThen/unwrapOr
03-result-basics Result<T, E> with map/mapErr
04-match-option, 05-match-result The match macro over Some/None and Ok/Err
06-tagged-union User-defined tagged unions with a _ wildcard arm
32-regex-literals Regex literals containing braces, quotes, a backtick, and a fake match block
33-await-in-string await inside a string in an arm stays synchronous; a real await makes the match async
34-try-operator ? on Result and Option, several per expression, in async functions, next to ?., ??, and ternaries
36-literal-patterns Literal arms over string, number, and boolean subjects; finite unions need no _; arms typed by the declared return type

Rejected on purpose (these pass by failing the checker with the right code)

Example Code
07-banned-null TSUS002
08-banned-class TSUS001
09-banned-var TSUS004
10-banned-undefined TSUS003
11-banned-throw TSUS005
30-non-exhaustive-match TSUS201, for a missing tagged-union variant, a missing Err, and literal arms that miss a union member
31-wrong-arm-type TSUS200 (an arm of the wrong type, a payload used as the wrong type) and TSUS202 (a misspelled variant)
35-try-misuse TSUS300, TSUS301, TSUS302, TSUS303
40-banned-any TSUS401
41-banned-as-cast TSUS402
42-banned-non-null TSUS403
43-banned-enum TSUS404
44-banned-loose-equality TSUS405
45-banned-default-export TSUS406
46-banned-param-reassign TSUS407
47-banned-let-never-reassigned TSUS408
48-banned-param-mutation TSUS409
49-banned-untagged-error TSUS410
50-banned-floating-promise TSUS411
51-banned-unused-result TSUS412
52-banned-ts-suppression TSUS413

Rejection examples can also pin exact positions ("diagnostics": [{ "code", "file", "line", "col", "tsCode" }] in example.json), which tests the source map back to the .tsus file.

Examples that list fixable codes in example.json also assert that those diagnostics carry a machine-applicable fix.

Interop and stress

Example Shows
12-tsus-imports-tsus, 20-deep-chain .tsus import chains, four levels deep
13-tsus-imports-ts, 14-ts-imports-tsus Mixed projects in both directions; .ts keeps its classes and nulls
15-npm-dependency zod feeding Result
16-boundary-fs, 26-json-boundary wrapAsync/wrapSync turning throws into Result
17-nested-match, 18-generics, 19-async-result Nested matches, generic signatures, await inside match arms
21-circular-imports Function-level .tsus cycles (fine, because ESM)
22-mixed-project ts + tsus + zod + boundary in one project
24-hostile-syntax Decoy match blocks in strings, comments, and template literals
25-shadowed-match A user function named match staying a plain call

Runtimes and frameworks

Example Shows
23-cloudflare-worker Cloudflare { fetch } module shape, tested against real Request/Response
27-nextjs Next.js 14 via the webpack loader; next build renders .tsus output into static HTML
28-astro Astro 4 via the Vite plugin, verified in built HTML
29-bun Bun runs .tsus imports natively via a bunfig.toml preload, no build step

v2 (plain .ts; see v2)

Example Shows
60-v2-security-demo The same Workers module in plain strict TS and in v2 with six planted bugs: plain passes tsc, v2 flags all six (and nothing else), the fixed version bundles with esbuild, and a harness runs the attacks against both
61-v2-basics match, matchLiteral, matchResult, Option as T | undefined, sql, html, validate; checked, bundled with esbuild and run with node's type stripping, same output
62-v2-rejected Every v2 code from TSUS500 to TSUS607, each pinned to the line it must land on, with nothing reported anywhere else
63-v2-audit tsus audit on a legacy file: the findings pinned to their lines, and the file unchanged

Agent tooling

Example Shows
53-boundary-file A *.boundary.tsus file using a checked cast; the rest of the program sees only a tagged Result
54-fix-roundtrip tsus fix on a copy (including edits inside match arms), compared byte for byte with expected/, then built and run
55-migrate tsus migrate on a legacy .ts module (class, var, nullable returns, throws) against golden output, then run; a second file shows the not-converted report
56-init tsus init into an empty directory, then the starter project builds and runs