# AI Agents Built a Faster TypeScript Compiler

Published: 2026-10-09

Theo Browne announced tsc-rs, an experimental agent-written Rust port of Microsoft's Go TypeScript compiler. This October 8 Daily Diff examines the project's workload-specific speed claims, API-price token equivalents, upstream revision mismatch and known adoption caveats. Supporting stories cover Haiku 5.5's long-prompt pricing tier and a census of infrastructure contributors. Niko's verdict is NEEDS REVIEW before replacing a release gate.

Canonical: https://thedailydiff.dev/video/theo-typescript-rust-port/

## What this video covers

- Theo announced tsc-rs on October 7, 2026 at 11:07 UTC. The project ports Microsoft's existing Go compiler design and tests; the successful Rust attempt did not invent all TypeScript semantics from scratch.
- Theo's roughly $400,000 unsuccessful OpenAI usage and $24,047 Opus usage are valued at API prices. He reports using Claude subscriptions for Opus. These figures do not establish cash payments of those amounts.
- The project's 12.5x result compares T3 Code with Effect against TypeScript 6 plus its older Effect plugin. Its Go-plus-Effect comparison is 1.89x, and Bun wins the no-Effect table. The older plugin has a different diagnostic rule set.
- The benchmark harness was inspected, with five timed runs after a warmup and native macOS ARM binaries. Performance and compatibility were not independently reproduced for this episode. Preview builds lack optimizations used by the measured build.
- The README names a September 29 upstream revision, while UPSTREAM.json labels a June pin current. This tooling metadata does not prove which revision the shipped binary uses. Verify the installed binary and compare equal revisions.
- The release lists Linux x64 and macOS arm64 support, alongside monorepo emission, build-reference and editor-memory caveats. The documented WASM build is single-threaded and lacks native language-server mode; infrastructure savings remain unmeasured.
- Haiku 5.5 launch pricing rises fivefold above 100,000 prompt tokens. Artificial Analysis reports heavy max-effort output-token use and a 33% Terminal-Bench result versus Anthropic's 39.2% under different evaluation setups. Its provisional cost figures exclude the long-prompt tier.
- The sheets.works census found one or two regular committers in 11 of 23 sampled infrastructure projects, using a threshold of ten commits in a year. That proxy can miss review, triage and support work. Adoption also depends on visible ownership of future maintenance.

## Chapters

- 0:00 Can agents port a compiler people actually use?
- 0:44 What does the token bill actually mean?
- 1:43 When is the twelve times speedup true?
- 2:18 What could break a drop in replacement?
- 3:17 Does cheaper Haiku mean cheaper work?
- 4:01 Who is already keeping our dependencies alive?
- 4:27 Who owns the next upstream update?
- 4:48 Why is my verdict NEEDS REVIEW?

## Transcript

### Can agents port a compiler people actually use?

0:00 You expect an AI compiler rewrite to collapse. Theo's Rust port beats Go on his T three Code benchmark, and he says he never read its code. In this video, what did agents actually build? And should you replace your checker? It's Thursday, October eighth, and this is The Daily Diff. Keep one question open. Who keeps this working when Microsoft changes TypeScript?

0:19 T S C R S is an experimental, open source TypeScript compiler, type checker and language server written in Rust. Theo Browne's project ports Microsoft's Go implementation, so it inherits an existing design and tests. It checks the types in your app and powers editor diagnostics. Wednesday brought Theo's port and a cheaper Haiku. Today's Hacker News board surfaced a maintainer census. The theme is abundant code, scarce ownership.

0:43 Theo says months of OpenAI attempts stalled around eighty four percent

### What does the token bill actually mean?

0:47 compatibility. His chart values their tokens above four hundred thousand dollars at API prices. That's usage valued at list prices. The imaginary invoice practically writes its own headline. He says Opus five point five replaced that earlier Rust attempt and produced a working first version in ten hours. The eventual run lasted two weeks, with roughly twenty four thousand dollars of API equivalent usage through his Claude subscriptions.

1:11 Starting over means replacing the unsuccessful port. Microsoft's compiler algorithms and test corpus were already there. This is unusually friendly terrain for an agent, because the original implementation can keep telling it exactly where it is wrong. Then the readme draws a Slop Line. Everything below it was written by agents. The project reports more than one hundred eighty thousand passing ported tests, with its own attribution label attached.

1:34 First question answered. Agents produced a port with an upstream oracle. The maintenance question begins where conversion ends, and a green test suite has a date on it.

### When is the twelve times speedup true?

1:43 The flashy speedup is twelve and a half times faster on T three Code with Effect diagnostics, against the older JavaScript compiler plus its plugin. The comparison with the Go compiler plus Effect is closer to twice as fast. Remove Effect, and Bun wins that same project's table. The winning horse changes with the race. The old Effect plugin also reports different errors, so the stopwatch comes with a semantic footnote.

2:05 These are the project's measurements on an M four Pro, using five timed runs after a warmup. Its preview binaries also lack optimizations used for the measured build. Put the machine and build settings beside the multiplier before borrowing it for your sales deck. The readme names a September development revision.

### What could break a drop in replacement?

2:20 The tooling manifest still labels a June pin current, so verify the binary you're comparing. Version drift can turn an extra error into a migration exercise. The release supports Linux on Intel compatible machines and Apple silicon Macs. Windows users get to watch the launch through a window. The known problems list also includes monorepo emit differences and builds reading stale dependency output when references are missing. I'd compare diagnostics and exercise project references before changing a release gate. Keep a long editor session in the test plan.

2:49 The readme acknowledges growing memory during edits, which your ten second benchmark will politely miss. The interesting deployment angle is WebAssembly. The repo includes a browser example and a small module with type checking in memory. That could bring checking closer to users, but the documented browser build is single threaded and lacks the native language server mode. Virtual machine savings depend on what an application moves into the browser. An infrastructure bill needs a workload study, even when the compiler has

3:15 excellent taste in programming languages.

### Does cheaper Haiku mean cheaper work?

3:17 Meanwhile, Haiku five point five offers very cheap tokens, with a threshold worth remembering. Prompts up to one hundred thousand tokens get the cheapest tier. Above that, input and output prices both jump fivefold. The large context window contains a pricing trapdoor. Artificial Analysis finds strong small model performance, but at maximum effort Haiku uses roughly three times Luna's output tokens per intelligence test task.

3:40 Cheap tokens can still arrive in large bags. Their provisional cost figures also exclude that long prompt price step. On Terminal Bench, Artificial Analysis reports thirty three percent, below Anthropic's roughly thirty nine. Different evaluation setups make those different measurements. Pick an effort setting with your own task success and latency in view, then calculate the bill from actual usage.

### Who is already keeping our dependencies alive?

4:01 And the maintainer census on Hacker News counted regular committers across twenty three infrastructure projects. Nearly half had only one or two. Its definition requires ten commits in a year, so it captures contribution concentration while missing some review and support work. The report attributes most recent time zone database changes to Paul Eggert. The official release notes explain why this keeps moving. Governments change clocks, and somebody translates that into software before

4:25 everyone's calendar develops an opinion. That answers our open question.

### Who owns the next upstream update?

4:28 Somebody still has to follow upstream changes, reproduce failures and maintain the port's compatibility promise. The agents inherited years of human decisions. I'd like the experiment even more with a visible plan for who owns the next update. If you'd rather read this than hear me say it, the diff lands in your inbox every morning, free at the daily diff dot dev, link below. So today's verdict, NEEDS REVIEW.

### Why is my verdict NEEDS REVIEW?

4:50 I'd test it alongside the existing checker. That answers question two. The workload wins are promising, and equal version correctness and maintenance decide whether it belongs in the release pipeline. And that's the diff for today. I'm Niko from Axrisi. Merge responsibly.

## Sources

- [tsc-rs announcement and attached token-price chart](https://x.com/theo/status/2107789940482621795) — Theo Browne on X
- [ts-rust README, benchmarks and known problems](https://github.com/pingdotgg/ts-rust) — Ping Labs on GitHub
- [T3 Code benchmark harness](https://github.com/pingdotgg/ts-rust/blob/main/scripts/bench-apps/t3code.sh) — Ping Labs on GitHub
- [Upstream development pin manifest](https://github.com/pingdotgg/ts-rust/blob/main/UPSTREAM.json) — Ping Labs on GitHub
- [WebAssembly build documentation and limitations](https://github.com/pingdotgg/ts-rust/blob/main/npm/wasm/README.md) — Ping Labs on GitHub
- [Introducing Claude Haiku 5.5](https://www.anthropic.com/claude-haiku-5-5) — Anthropic
- [Haiku 5.5 announcement video](https://x.com/claudeai/status/2107894039626277339) — Claude on X
- [Claude Haiku 5.5 independent evaluation](https://artificialanalysis.ai/articles/claude-haiku-5-5) — Artificial Analysis
- [The people holding up the internet](https://sheets.works/data-viz/holding-up-the-internet) — sheets.works
- [Time zone database release notes](https://data.iana.org/time-zones/tzdb/NEWS) — IANA