Hey there, TypeScript team member here. The tsrs project has been extremely interesting to see come together. We've been working with the maintainers to better understand how tsrs is fast and what sorts of optimizations it uses are safe. We even did our own Rust vibe port and what we currently suspect is that a lot of the speed does not just come down to language choice, and some of what is going to make TS 7.1 faster and leaner (see the charts in the blog post) actually comes from optimizations discovered in tsrs.
We think the wasm and oxlint integration makes a lot of sense for Cloudflare's needs, so this investment makes sense. At the same time, we'll be working hard and collaborating to make the official TypeScript toolset faster and even more efficient.
Congrats to the tsrs team on making this official. We're excited and we look forward to continuing on collaboration!
No matter how many optimizations can be ported from tsrs to ts-go, Rust will be more efficient and enable the WASM use case.
The premise of porting the compiler from TS to GO was ease of migration due to Go's garbage collector. That doesn't hold anymore? AI doesn't care how hard or easy a migration is if a test suite exists.
> That doesn't hold anymore? AI doesn't care how hard or easy a migration is if a test suite exists.
Sure, but that only matters if migration is going to be the last thing anyone does on a project. Unless you think typescript 7.1 ought to be enough for anybody, it also matters what the typescript team wants to write/read/reason in.
I just vibecofed tsasm. tsrs is obsolete now and anyone who still uses it is hopelessly behind the times. The tsrs team should delete the repository immediately and join my project.
Too late I just shipped tsvhdl so you have a hardware implementation of the actual TS compiler. Why steal valuable CPU cycles from agent workflows, off load all transpilation and type checking to a dedicated TSC chip.
I am okay if agents can write code that is very safe in ASM. I am sure code written in Rust are naturally safe compared to asm or c. Waiting for tsasm to be significantly faster and safer than tsrs.
Did you? GitHub link? because that would be awesome! which asm target did you pick? I looked at using LLVMs' LVM because I didn't want to have to choose ARM or x86? What are some of the trade-offs you made along the way? Wait, are you making a joke and think you're funny? (Hate to be the bearer of bad news but you're not.)
> Why maintain an official compiler that's not even half as fast as the vibe-coded one?
Because at the end of the day, someone has to actually care and put in effort, even if the thing is machine generated, because it can go off the rails.
Because its impossible to stay motivated working on non trivial project when AI is running the show. Its the last thing Id want to do, its soul sucking and the code is terrible. Remember these guys have poured years of actual work, not slop into tsgo. You sloppers wouldnt understand that.
I never understand why people say things like this. It is not uncommon for languages to have multiple implementations, everything must not be abandoned for Rust man.
I assume that most of the speedup in ts-rs does not simply stem from using a language without garbage collection, but actually from doing smart optimizations that can be done in any langauge.
Similar to how uv in the Python world explained that their implementation being in Rust has very little to do with the most impactful speedups they did.
Another advantage of AI, though, is that it becomes easier to maintain more than one TypeScript compiler. Perhaps there's some benefits to having some diversity?
I walk to the supermarket. I'm pretty sure an F1 car would be noticeably slower.
On foot I cross the little residential street where I live, walk to the corner, cross the busier road using a Zebra that was installed by thoughtful "pedestrian first" local government, then walk to the next corner where the main entrance to the supermarket is.
In the F1 car lets assume somebody already pre-warmed it and so it's ready to go (F1 cars do not "just work" like your family car because that's not required in the F1 rules) I walk to the car park and get in, I need to accelerate quite hard to get the slippery F1 car up the steep drive, but then brake hard to stop and have somebody else check both ways since an F1 car isn't equipped for this in a side street. Now I pull out, takes no time to reach that street where pedestrian me crossed but I can't use that route in a car, so, go through another residential area, turn onto an old main road, turn again, to a roundabout, go into the underground carpark of the supermarket. Once I've parked I'm about as close to the front door as if I were walking from the Zebra. Huh.
My car doesn't require a dedicated team of engineers to start and has a trunk to hold groceries :P
The question isn't really if the speed is bad, it's what tradeoffs you need to make to realize it. The F1 car has a very narrow operating condition and so can make tradeoffs that are completely unreasonable for vehicles that need to do other things.
From what cloudflare is claiming, Tsrs gives me more speed without tradeoffs I care about, so why not have the free speed?
the cost between an F1 car and a "regular" car is high, both financially and functionally.
but tsrs and tsgo are not a similar comparison. they cost the same (you download a binary and run it) and are functionally identical. in that case why would you not use the faster & better one?
the question should be, do you always want your groceries 50% faster ?
Bad comparrison. The less time my computer spends doing things the less battery I use.
An F1 car goes faster but also uses more fuel to do so.
Any one tool being a bit faster doesn't make my batter last noticeably longer, but defaulting to using the efficient tooling gets me results I can see.
Now we're getting somewhere. This is a great example because obviously speed depends on the context.
Obviously a shovel is quite an improvement for home gardeners but big constructions need a completely different tool.
For +99% of TS projects going from eg 10ms to 1ms is not going to make any difference even though it's a 10x improvement. In part because most monitors refresh at 60hz (16ms per frame) but even with +120hz monitors this is not going to have an impact on any workflow. Even for CI network latency is going to be a much bigger factor than improving TS compilation beyond the performance that TSGO already provides.
I think it might have more meaningful impact on high volume use cases, i.e. where cumulative run time is still significant.
That’s already meaningful for at least a couple of companies I’ve seen investing in these ports. It might be niche for individual dev workflows now, but as those evolve it could become less niche (albeit probably still less uniform).
I mean... If I had an excavator it beats using a shovel? This is such a silly "gotcha" analogy because the only reason I wouldn't is cost. And cost is not a reason for tsgo to be slower?
"Would you rather have worse software for literally no benefit?"
We should aim to make software as performant, bug free, feature-complete, and maintainable as is cost effective. It's asinine to argue that software shouldn't be fast, not because of some trade off, but just because someone doesn't need it to be fast for small tasks.
Which, mind you, isn't even the case here. Thousands of businesses use typescript for millions of lines of code. Performance gains can mean tens of thousands of dollars in compute or hardware or time; tons of carbon saved from being emitted.
Yep cost is the reason, allegedly the ts team found go a better language for a port and both using a less ergonomic language and switching target every six months have associated costs
Please you port it if its so easy, the tests are available on gh. I wont be using an llm ported ts compiler anyways, and neither would many others
Its a great way to divide and kill the popularity of your project. I made sure bun is no longer used in my devices or projects that depend on it for example, I'm not alone in this.
+1 from what I’ve seen, porting apples to apples to rust only impacts performance by a few % (sometimes in the wrong direction). The main speed up comes from algorithmic improvements that can also be applied to the original project.
Thanks for the response and it is encouraging to hear that there's more to this project than vibe coding! Sounds like some interesting things were found in the process, and I'm all for vibe coding as a way to learn new things.
Im not because nobody actually ever learns anything. For every one person that learns something the other 99% of the population gets lazier and dumber, ponder how that might play itself out.
This is probably the most harmful tech ever created simply because how many people trick themselves into thinking they're smarter while actually becoming much dumber. We're actually cooked.
If I recall correctly, a few of both - there were instances where some work could be done more lazily/minimally (https://github.com/microsoft/TypeScript/pull/64476), and there were instances where we changed our allocation patterns and played around with some struct layouts. We originally strived not to drastically reorganize our data structures to make things easier to port, and we have more flexibility now. We're continuing to experiment more here.
Ah I was gonna ask what happened with tsc-rs by Theo but looks like you guys are merging it in, at least it won't be a waste of his $500k in token spend, lol.
It appears they're using a unique memory model (https://github.com/oxc-project/tsrs/blob/main/docs/PORTING.m...) to overcome the challenge that prevented the TypeScript team from choosing Rust to begin with: Rust has no GC and porting from a GC language to it often requires restructure of data.
Essentially, they fake a GC via never-freed arenas for long-living objects. It's interesting that Rust can do that (hopefully soundly)!
>Create with P::new(value). p.get() returns &'static T. [...] unless an explicit recycling site proves them dead
It's unsound, the LLM/human is doing the proving. They've hidden unsafe in an abstraction without removing the unsafeness. Using the arena allocator's lifetime would have been safe and easy (bumpalo recommends this AFAIK).
I'm guessing a sycophantic LLM either found or praised this load-bearing seam.
Calling this a breakthrough in memory management is an understatement. Call up the Turing Award committee. We just discovered a solution to the halting problem! Turns out that for certain applications the answer is unequivocally "yes"
We've have found an engineering solution to some of the cases for the halting problem, yes. It's still not solved science problem though. Science isn't engineering though.
Compilers often just build up more and more data structures in memory until the end when they then release everything more or less at once. With an arena you can then just throw away the whole arena at the end, assuming there are no resource handles that need to be closed in drop().
A long time ago I read an article about someone speeding up a C++ compiler that is itself written in C++ (was it g++? maybe clang?). They simply patched out delete so it did nothing. The compiler used marginally more memory but ran much faster, since releasing all the small objects one by one is just useless work at the end of the compilation.
A GCed language has a similar behavior and could therefore be fast in that case. It just builds up all the objects and then when the program exists it doesn't have to GC them all, it just throws away the whole process memory. So I can imagine one could "optimize" a compiler written in a GCed language by deactivating the GC.
Ah took me a while to understand the reason behind this.
It wasn’t “let’s build a faster Rust based Typescript compiler” at all. That was a nice side-effect.
They needed a Typescript compiler that can run on their WASM only sandbox. So the only existing option was to export tsc Go to WASM, but that comes with the full Go runtime.
And to avoid that cost, they experimented with a Rust compiler and turns out it’s even faster than the reference tsc/Go implementation.
It's exciting to see the speed gains. I use TypeScript every day on a large codebase, and it is too slow to run locally.
I don't love this change though:
> However, frontier AI models can now do large-scale ports almost independently in a matter of days
> I wrote zero code. Not "barely any". Zero
So we are going to move from a carefully handwritten port to an entirely AI-generated port. I know this is the way the industry is headed, I use coding agents a ton. But I still review their output because they often don't make sound architectural decisions. Maybe this is a sign that my way of working is going away. I'm not sure how to justify that kind of time spent when someone can do the same thing in a matter of days with barely any effort.
I am working on my own language now, I've spent most of this year on it. The idea that someone could just replace it in a matter of days is wild. Wonder how the TS team at Microsoft feels about this.
> The idea that someone could just replace it in a matter of days is wild.
They didn’t design a language that millions of human engineers enjoy (?, lol) using. They took an existing codebase and ported it. Yes, the porting part is easier now but only because another project and spec exist.
While this is true - and I see the same sentiment as i am about to make ad nauseam - it is kind of amazing how far the goal posts have moved in such a short space of time.
If TypeScript has the integration testing framework in place to verify that the compiler actually works (and it should) - then, presumably, you should be able to swap in a Chinese Room black-box and - at least - know that it works.
You do have to be worried that it's possibly installing trojan horses on your computer while compiling code, or any other number of unwanted side effects.
But you can and should be able to have integration testing in place to know that a compiler - at a minimum - actually works.
Totally fair. The thing is, though, agents often write code in ways I don't expect, and I don't think any human would write, which makes me think that code may also break in a way that no human expects or would have written a test for.
Your way of working is definitely going away. And quickly. The idea of reviewing code today is already a questionable use of time. In a few more months to a year it will clearly be a waste of time.
What strange times. A few years ago seeing a headline like this, my jaw would have literally dropped. Instead I'm wondering what kind of software projects will have meaning in the future if this headline feels common.
Wasm comes with a garbage collector though, no? So a garbage collected language that offloads all memory management to the runtime would actually be better than one that ignored the platform and does its own stuff.
Ignored is a little unfair to the Wasm committee, the Wasm GC is constrained by what is possible to ship in existing Wasm runtimes and the stuff needed by many languages is way beyond what those existing runtimes have available to them.
I think it made sense for them to compromise, at least on the initial version of Wasm GC, to get it out the door. I say this as someone who wants the features that got left out, like interior references :)
The JVM alone has, what, 4 different modern GCs now? Garbage collectors aren't really universal, so they can pose different issues for different languages.
Yes, and when you ship a Go WASM package it is already included. But it has a certain impedence mismatch with WASM since it is rolling its own stuff that WASM can't see, I believe the performance is a bit less good than it could be, and it results in very large WASM packages for small amounts of code.
Since a typescript compiler isn't "a small amount of code", you'd have to see specific numbers on the final WASM assembly size to know how relevant that is. I'd expect well-written Rust code to beat well-written Go code on final assembly size by some amount, which may or may not be relevant to any particular use case (since in a lot of cases one "couple dozen megabytes" is much the same as another when it really gets down to it, it's not like Go is going to be emitting gigabytes to Rust's hundreds of kilobytes or something). However I can't judge what AI-written Rust code versus well-written Go code would come out as. AI is not terribly good at being concise (he said, while writing this post between prompts fighting with an AI to make a PR much more concise than it currently is).
You already can run the Go tsc in WASM using mature tools; but it further amplifies the memory and performance difference between the Rust version (which maps more cleanly onto WASM's design).
1. Porting typescript is made easier by the enormous test suite they can copy for verification.
2. They're doing this for workers, so with a systems-level language, there's a much higher potential for CPU and memory savings (an order of magnitude on this first pass, apparently). Driving their compute costs down is very revenue-efficient at this scale.
> Just a few weeks ago we announced supporting native Rust in Workers with the new Emscripten target for wasm-bindgen. When you run entire software build and deployment systems in tight loops on Workers, the benefits of a small memory footprint pay off quickly.
Or we can just drop TypeScript completely and type-check JavaScript directly.
With HM-style type inference, row-polymorphism, sum types, type classes for overloading + operations, recursive types for object methods, and type annotations for edge cases.
typescript is two things: a language, and a type system.
that same type system exists in the javascript language (jsdoc). doing those type checks is the 'slow' part that makes go / rust appealing. not just stripping types to emit js.
I assume you mean jsdoc + tsc in js type checking mode (tsc --allowJs --checkJs --strict)?
inty infers more - including generics (parametric polymorphism), while tsc is more of a simpler local inference.
inty is more strict - it accepts less programs than tsc would. This is mostly on purpose (to our benefit) as some of those programs have unsound code like array covariant (downcasting array members of the wrong type). tsc makes it easier to slip through these bugs.
And if tsc + jsdoc is good then the same question remains (do we really need the TypeScript language?)
youtuber vibe codes a transpiler says they're not going to bother maintaining it - but big company like cloudflare takes it seriously enough to absorbe into their own official services.
What is happening?
What is even software dev anymore?
What didnt the labs or cloudflare do this themselves, I mean they did, but then if it's all vibe coded, why do they need someone else's vibe coded stuff?
Shameless plug of my Typescript C# port. It's a case study with some interesting (perhaps) notes from me in the README about approach, performance, and cost(~$75).
The problem with Python is that the whole ecosystem is based on CPython’s implementation details, so every change inevitably breaks something and requires effort of package maintainers to fix compatibility. There is no test suite an agent can follow.
I guess it will be soon that typescript won’t matter anymore since agents will be writing code is Rust compiled to WASM or similar for everything. typescript added human readability or JavaScript, if no human is coding, I don’t see the point of wasting tokens on this.
Ehm no, added type safety, and dealing with a type checker it's exactly what agents need.
> if no human is coding
Nope outside the bubble there are many still hand coding(me too)
> I guess it will be soon that typescript won’t matter anymore since agents will be writing code is Rust
Oh hold on a second isn't Rust the same as Typescript "adding human readability"?! Why not trashing Rust and just have LLMs output WASM directly? XD Wasm which is basically useless outside the browser or some microvms on the edge, and in the browser you don't even have an interface for directly rendering anything and still have to deal with the dom
You either are highly non technical, very junior or I just don't know
I'm tired of all the new tools that comes out every day, but I would say, this is basically the Vite team and if they implement it with Vite+, then I'm all for it, less decisions about tooling to be made, one config file, etc.
All of this just for enabling the WASM use case? Why not keep supporting Deno then that is written in Rust? To me it just sounds like a technical ai journey and a blow to the go port
I benchmarked tsrs a few weeks ago. I was testing out tsgo, ts-rust, tsrs, and bun for typechecking on our monorepo. tsrs is definitely awesome, it is very memory efficient when bumping up the number of threads. Not as efficient as bun, but it has incremental checks, which bun does not have.
I use oxc tooling and it's great to see the rust rewrite singularity in action and get to ride the fastest horses, but I would like that they made a branch for 7.2 as experiment to see how it compares. Maybe this ends in a win for Microsoft, they could cut operating expenses by keeping just Anders Hejlsberg in a hammock reciting invariants to claudpilot.
Obligatory reminder that their needs from a TS compiler are extremely different from a regular web developers iteration cycle. They are trying to run millions of these while saving on costs.
You almost definitely don't need to immediately plan to switch.
> This enables new possibilities when combined with agents: You are now able to run a development toolchain end-to-end in a fully sandboxed environment on Cloudflare.
True. Looks like main use-case is AI Agent Sandboxing
this is insane! saw a quote tweet of this by an infra engineer at chat gpt who said “ I'm looking at migrating over for the ChatGPT codebase and saw ~70% improvements on both wall time perf + peak cpu/memory usage.
then i went and tried this on my own open source harness. type checking went from 10 seconds to under half a second and uses a quarter of the memory. madness
You should say you're staying fungal infection free as it's named after the Rust fungus not the chemical reaction even if that's what people think these days.
Rust is basically the perfect language for LLMs because it has a strong, comprehensive, and strict compiler. For coding you really need a tight loop for models so they can iterate. C doesn’t accomplish this because it’s far too “anything goes” so bugs slip through. Dynamically typed languages don’t work either because their source code doesn’t contain enough context.
Hey there, TypeScript team member here. The tsrs project has been extremely interesting to see come together. We've been working with the maintainers to better understand how tsrs is fast and what sorts of optimizations it uses are safe. We even did our own Rust vibe port and what we currently suspect is that a lot of the speed does not just come down to language choice, and some of what is going to make TS 7.1 faster and leaner (see the charts in the blog post) actually comes from optimizations discovered in tsrs.
We think the wasm and oxlint integration makes a lot of sense for Cloudflare's needs, so this investment makes sense. At the same time, we'll be working hard and collaborating to make the official TypeScript toolset faster and even more efficient.
Congrats to the tsrs team on making this official. We're excited and we look forward to continuing on collaboration!
At this point, why invest further in TSGO?
No matter how many optimizations can be ported from tsrs to ts-go, Rust will be more efficient and enable the WASM use case.
The premise of porting the compiler from TS to GO was ease of migration due to Go's garbage collector. That doesn't hold anymore? AI doesn't care how hard or easy a migration is if a test suite exists.
> That doesn't hold anymore? AI doesn't care how hard or easy a migration is if a test suite exists.
Sure, but that only matters if migration is going to be the last thing anyone does on a project. Unless you think typescript 7.1 ought to be enough for anybody, it also matters what the typescript team wants to write/read/reason in.
I would love this to be typescript 8 and 8.1 story (:
I just vibecofed tsasm. tsrs is obsolete now and anyone who still uses it is hopelessly behind the times. The tsrs team should delete the repository immediately and join my project.
Too late I just shipped tsvhdl so you have a hardware implementation of the actual TS compiler. Why steal valuable CPU cycles from agent workflows, off load all transpilation and type checking to a dedicated TSC chip.
I am okay if agents can write code that is very safe in ASM. I am sure code written in Rust are naturally safe compared to asm or c. Waiting for tsasm to be significantly faster and safer than tsrs.
People massively overestimate Rust's "safety" - it has a very nice approach to managing data races, and. that's it.
It still, like every other language/tool we have that shares state between threads, has race conditions, for example.
That is a confidently incorrect assertion.
Did you? GitHub link? because that would be awesome! which asm target did you pick? I looked at using LLVMs' LVM because I didn't want to have to choose ARM or x86? What are some of the trade-offs you made along the way? Wait, are you making a joke and think you're funny? (Hate to be the bearer of bad news but you're not.)
Your answer is within the comment you wrote. AI is maintaining tsrs whereas humans are maintaining tsgo.
Your question boils down to "why would humans maintain something?" which could have a variety of answers depending on context.
No, the question is now: "How much human mental momentum, ego, and investment is there in and behind any particular project or its forks?"
I for one would be demotivated. Why maintain an official compiler that's not even half as fast as the vibe-coded one?
"Just" publish the language specs and be done.
These ports are borrowing large parts of the architecture, algorithms, and data structures from the official implementation.
I don't think a spec would facilitate them at such a low effort level and cost.
> Why maintain an official compiler that's not even half as fast as the vibe-coded one?
Because at the end of the day, someone has to actually care and put in effort, even if the thing is machine generated, because it can go off the rails.
Because its impossible to stay motivated working on non trivial project when AI is running the show. Its the last thing Id want to do, its soul sucking and the code is terrible. Remember these guys have poured years of actual work, not slop into tsgo. You sloppers wouldnt understand that.
I never understand why people say things like this. It is not uncommon for languages to have multiple implementations, everything must not be abandoned for Rust man.
Because TSGO is stable and compatible?
I assume that most of the speedup in ts-rs does not simply stem from using a language without garbage collection, but actually from doing smart optimizations that can be done in any langauge.
Similar to how uv in the Python world explained that their implementation being in Rust has very little to do with the most impactful speedups they did.
If an AI agent could vibe-optimise tsgo, why port it all again over to rust?
because Rust will be faster and can compile to WASM which the TS ecosystem needs
Another advantage of AI, though, is that it becomes easier to maintain more than one TypeScript compiler. Perhaps there's some benefits to having some diversity?
TSGO is already plenty fast and overkill for most projects.
Sure the Rust version is faster but the difference is probably only noticeable in the top 1% largest projects.
Fast is one thing, TSGO is fast enough I think, but my god does it suck up memory
I'm always mind-blown when I read stuff like "speed is overkill". Speechless!
do you go to the supermarket with an F1 car?
I walk to the supermarket. I'm pretty sure an F1 car would be noticeably slower.
On foot I cross the little residential street where I live, walk to the corner, cross the busier road using a Zebra that was installed by thoughtful "pedestrian first" local government, then walk to the next corner where the main entrance to the supermarket is.
In the F1 car lets assume somebody already pre-warmed it and so it's ready to go (F1 cars do not "just work" like your family car because that's not required in the F1 rules) I walk to the car park and get in, I need to accelerate quite hard to get the slippery F1 car up the steep drive, but then brake hard to stop and have somebody else check both ways since an F1 car isn't equipped for this in a side street. Now I pull out, takes no time to reach that street where pedestrian me crossed but I can't use that route in a car, so, go through another residential area, turn onto an old main road, turn again, to a roundabout, go into the underground carpark of the supermarket. Once I've parked I'm about as close to the front door as if I were walking from the Zebra. Huh.
My car doesn't require a dedicated team of engineers to start and has a trunk to hold groceries :P
The question isn't really if the speed is bad, it's what tradeoffs you need to make to realize it. The F1 car has a very narrow operating condition and so can make tradeoffs that are completely unreasonable for vehicles that need to do other things.
From what cloudflare is claiming, Tsrs gives me more speed without tradeoffs I care about, so why not have the free speed?
the cost between an F1 car and a "regular" car is high, both financially and functionally.
but tsrs and tsgo are not a similar comparison. they cost the same (you download a binary and run it) and are functionally identical. in that case why would you not use the faster & better one?
the question should be, do you always want your groceries 50% faster ?
No, but I sure wouldn’t mind if I opened my fridge and the food was already there.
If I could afford one, I absolutely would! I'd add a tow hitch so it had some storage and get it all street legal and add blinkers.
Oh boy that’s about the most off the mark car analogy I’ve ever seen and I’ve seen some
If you are travelling a long distance, say over 1000 miles/km, do you always drive? Or do you take an airplane (or a train)?
Bad comparrison. The less time my computer spends doing things the less battery I use.
An F1 car goes faster but also uses more fuel to do so.
Any one tool being a bit faster doesn't make my batter last noticeably longer, but defaulting to using the efficient tooling gets me results I can see.
Do you usually dig up your garden by hand with a shovel, or do you use a rototiller?
Now we're getting somewhere. This is a great example because obviously speed depends on the context.
Obviously a shovel is quite an improvement for home gardeners but big constructions need a completely different tool.
For +99% of TS projects going from eg 10ms to 1ms is not going to make any difference even though it's a 10x improvement. In part because most monitors refresh at 60hz (16ms per frame) but even with +120hz monitors this is not going to have an impact on any workflow. Even for CI network latency is going to be a much bigger factor than improving TS compilation beyond the performance that TSGO already provides.
I think it might have more meaningful impact on high volume use cases, i.e. where cumulative run time is still significant.
That’s already meaningful for at least a couple of companies I’ve seen investing in these ports. It might be niche for individual dev workflows now, but as those evolve it could become less niche (albeit probably still less uniform).
Do you dig your garden with an excavator?
Why are bringing Java into this?
I mean... If I had an excavator it beats using a shovel? This is such a silly "gotcha" analogy because the only reason I wouldn't is cost. And cost is not a reason for tsgo to be slower?
"Would you rather have worse software for literally no benefit?"
We should aim to make software as performant, bug free, feature-complete, and maintainable as is cost effective. It's asinine to argue that software shouldn't be fast, not because of some trade off, but just because someone doesn't need it to be fast for small tasks.
Which, mind you, isn't even the case here. Thousands of businesses use typescript for millions of lines of code. Performance gains can mean tens of thousands of dollars in compute or hardware or time; tons of carbon saved from being emitted.
Yep cost is the reason, allegedly the ts team found go a better language for a port and both using a less ergonomic language and switching target every six months have associated costs
And that's a fine argument to make! But this stupid garden analogy is nonsense.
Please you port it if its so easy, the tests are available on gh. I wont be using an llm ported ts compiler anyways, and neither would many others
Its a great way to divide and kill the popularity of your project. I made sure bun is no longer used in my devices or projects that depend on it for example, I'm not alone in this.
+1 from what I’ve seen, porting apples to apples to rust only impacts performance by a few % (sometimes in the wrong direction). The main speed up comes from algorithmic improvements that can also be applied to the original project.
Thanks for the response and it is encouraging to hear that there's more to this project than vibe coding! Sounds like some interesting things were found in the process, and I'm all for vibe coding as a way to learn new things.
Im not because nobody actually ever learns anything. For every one person that learns something the other 99% of the population gets lazier and dumber, ponder how that might play itself out.
This is probably the most harmful tech ever created simply because how many people trick themselves into thinking they're smarter while actually becoming much dumber. We're actually cooked.
Been with TS for many years now and the work your team's done over the years is amazing. Thank you
Thanks for the input, glad to hear both parties benefit.
Could you elaborate on what type of optimizations were added to TS 7.1? As in, were they more algorithmic or related to memory management?
If I recall correctly, a few of both - there were instances where some work could be done more lazily/minimally (https://github.com/microsoft/TypeScript/pull/64476), and there were instances where we changed our allocation patterns and played around with some struct layouts. We originally strived not to drastically reorganize our data structures to make things easier to port, and we have more flexibility now. We're continuing to experiment more here.
Ah I was gonna ask what happened with tsc-rs by Theo but looks like you guys are merging it in, at least it won't be a waste of his $500k in token spend, lol.
It appears they're using a unique memory model (https://github.com/oxc-project/tsrs/blob/main/docs/PORTING.m...) to overcome the challenge that prevented the TypeScript team from choosing Rust to begin with: Rust has no GC and porting from a GC language to it often requires restructure of data.
Essentially, they fake a GC via never-freed arenas for long-living objects. It's interesting that Rust can do that (hopefully soundly)!
From the file:
>Create with P::new(value). p.get() returns &'static T. [...] unless an explicit recycling site proves them dead
It's unsound, the LLM/human is doing the proving. They've hidden unsafe in an abstraction without removing the unsafeness. Using the arena allocator's lifetime would have been safe and easy (bumpalo recommends this AFAIK).
I'm guessing a sycophantic LLM either found or praised this load-bearing seam.
Never deallocating would have been sound (missile memory management: https://devblogs.microsoft.com/oldnewthing/20180228-00/?p=98... ).
> Never deallocating would have been sound (missile memory management: https://devblogs.microsoft.com/oldnewthing/20180228-00/?p=98... ).
Calling this a breakthrough in memory management is an understatement. Call up the Turing Award committee. We just discovered a solution to the halting problem! Turns out that for certain applications the answer is unequivocally "yes"
We've have found an engineering solution to some of the cases for the halting problem, yes. It's still not solved science problem though. Science isn't engineering though.
With sanitizers (https://github.com/oxc-project/tsrs/blob/main/notes/mem-recy...). We're back to C I guess. I hate this project now.
> load-bearing seam
Zing!
Wouldn't that eat the whole ram?
Compilers often just build up more and more data structures in memory until the end when they then release everything more or less at once. With an arena you can then just throw away the whole arena at the end, assuming there are no resource handles that need to be closed in drop().
A long time ago I read an article about someone speeding up a C++ compiler that is itself written in C++ (was it g++? maybe clang?). They simply patched out delete so it did nothing. The compiler used marginally more memory but ran much faster, since releasing all the small objects one by one is just useless work at the end of the compilation.
A GCed language has a similar behavior and could therefore be fast in that case. It just builds up all the objects and then when the program exists it doesn't have to GC them all, it just throws away the whole process memory. So I can imagine one could "optimize" a compiler written in a GCed language by deactivating the GC.
Ah took me a while to understand the reason behind this.
It wasn’t “let’s build a faster Rust based Typescript compiler” at all. That was a nice side-effect.
They needed a Typescript compiler that can run on their WASM only sandbox. So the only existing option was to export tsc Go to WASM, but that comes with the full Go runtime.
And to avoid that cost, they experimented with a Rust compiler and turns out it’s even faster than the reference tsc/Go implementation.
Another reason: type aware listing in oxlint.
Oxlint already has type-aware linting, integrated with TS7/typescript-go.
Yeah but now it can just all be in rust, which is a lot simpler.
It's exciting to see the speed gains. I use TypeScript every day on a large codebase, and it is too slow to run locally. I don't love this change though:
> However, frontier AI models can now do large-scale ports almost independently in a matter of days
> I wrote zero code. Not "barely any". Zero
So we are going to move from a carefully handwritten port to an entirely AI-generated port. I know this is the way the industry is headed, I use coding agents a ton. But I still review their output because they often don't make sound architectural decisions. Maybe this is a sign that my way of working is going away. I'm not sure how to justify that kind of time spent when someone can do the same thing in a matter of days with barely any effort.
I am working on my own language now, I've spent most of this year on it. The idea that someone could just replace it in a matter of days is wild. Wonder how the TS team at Microsoft feels about this.
> The idea that someone could just replace it in a matter of days is wild.
They didn’t design a language that millions of human engineers enjoy (?, lol) using. They took an existing codebase and ported it. Yes, the porting part is easier now but only because another project and spec exist.
While this is true - and I see the same sentiment as i am about to make ad nauseam - it is kind of amazing how far the goal posts have moved in such a short space of time.
It’s not a goal post. It was literally a fucking benchmark of the very first LLMs.
sorry, can please explain this? which benchmark are you referring to?
That's fair, I was thinking the same – porting must be simpler, especially if there's already a substantial test suite
If TypeScript has the integration testing framework in place to verify that the compiler actually works (and it should) - then, presumably, you should be able to swap in a Chinese Room black-box and - at least - know that it works.
You do have to be worried that it's possibly installing trojan horses on your computer while compiling code, or any other number of unwanted side effects.
But you can and should be able to have integration testing in place to know that a compiler - at a minimum - actually works.
Totally fair. The thing is, though, agents often write code in ways I don't expect, and I don't think any human would write, which makes me think that code may also break in a way that no human expects or would have written a test for.
Your way of working is definitely going away. And quickly. The idea of reviewing code today is already a questionable use of time. In a few more months to a year it will clearly be a waste of time.
What strange times. A few years ago seeing a headline like this, my jaw would have literally dropped. Instead I'm wondering what kind of software projects will have meaning in the future if this headline feels common.
According to voidzero, this port is up to 8x faster than TS 7.0 & ~70% less memory.
Microsoft should seriously consider porting typescript to rust at this point
I'm pretty sure they already considered it before settling on Go. Allegedly, they chose Go because it was easier to translate to.
That was pre-AI. Ironically if they waited just another year they could've done this port directly to Rust themselves.
Not really because this port is based on the Go version. I doubt it would be as easy for an LLM to port the TS 6 codebase.
I wonder how more frustrating is it to work on the compiler in rust vs go.
IIRC at the time of the port to go they wanted fast test and compile time, of course it's a different era now.
That's interesting. A go version to iterate on the code and a rust version to run the code might be an optimal split here.
Idk. Apart from the fact that benchmarks are generally of questionable value, what's the point? Is the TS compiler unbearably slow?
They say the point in the article - WASM
(Not sure if this could be a genuine remark or is more of a joke, but)
Why not vibecode a Go -> WASM cross-compiler, then?
Go needs a garbage collector while rust doesn’t.
Wasm comes with a garbage collector though, no? So a garbage collected language that offloads all memory management to the runtime would actually be better than one that ignored the platform and does its own stuff.
Apparently there were some design issues (raised and ignored) that made it unusable for many languages.
Ignored is a little unfair to the Wasm committee, the Wasm GC is constrained by what is possible to ship in existing Wasm runtimes and the stuff needed by many languages is way beyond what those existing runtimes have available to them.
I think it made sense for them to compromise, at least on the initial version of Wasm GC, to get it out the door. I say this as someone who wants the features that got left out, like interior references :)
The JVM alone has, what, 4 different modern GCs now? Garbage collectors aren't really universal, so they can pose different issues for different languages.
Yes, of course, but couldn’t you have/make one in WASM too? (I have no clue how memory is allocated there)
Yes, and when you ship a Go WASM package it is already included. But it has a certain impedence mismatch with WASM since it is rolling its own stuff that WASM can't see, I believe the performance is a bit less good than it could be, and it results in very large WASM packages for small amounts of code.
Since a typescript compiler isn't "a small amount of code", you'd have to see specific numbers on the final WASM assembly size to know how relevant that is. I'd expect well-written Rust code to beat well-written Go code on final assembly size by some amount, which may or may not be relevant to any particular use case (since in a lot of cases one "couple dozen megabytes" is much the same as another when it really gets down to it, it's not like Go is going to be emitting gigabytes to Rust's hundreds of kilobytes or something). However I can't judge what AI-written Rust code versus well-written Go code would come out as. AI is not terribly good at being concise (he said, while writing this post between prompts fighting with an AI to make a PR much more concise than it currently is).
You already can run the Go tsc in WASM using mature tools; but it further amplifies the memory and performance difference between the Rust version (which maps more cleanly onto WASM's design).
1. Porting typescript is made easier by the enormous test suite they can copy for verification.
2. They're doing this for workers, so with a systems-level language, there's a much higher potential for CPU and memory savings (an order of magnitude on this first pass, apparently). Driving their compute costs down is very revenue-efficient at this scale.
For workers, you only need to strip the types. There's no need to do the expensive checking.
Did you read TFA?
> Just a few weeks ago we announced supporting native Rust in Workers with the new Emscripten target for wasm-bindgen. When you run entire software build and deployment systems in tight loops on Workers, the benefits of a small memory footprint pay off quickly.
The success of "modern" CLI tools written in Rust (uv, oxlint, turbopack) show there's a clear demand for faster tools in general.
Also, faster feedback loop is crucial for AIs, so I guess this makes sense.
What they show is that JavaScript should never have left the browser into the backend.
Don't rewrite into Rust, Go, Java, .NET, Zig, Odin, whatever.
Start with them directly.
> Is the TS compiler unbearably slow?
It's not. Even the JS compiler is still fast enough for many projects given the single core improvements in hardware (eg Apple Silicon).
Faster feedback loops almost always deliver improvements elsewhere.
Up to 8x faster seems kinda of a crazy claim. 8x faster at what?! TS 7.0 is reasonably well optimized and compiles VSCode in well under 10 seconds..
This can type check and compile VSCode in less than a second?
For VSCode: 4x speedup vs TS 7, 3x vs TS 7.1
Yes, it's in the article.
Well, no, not exactly. The article says vscode is compiled in 5.28s in typescript 7.1 (nightly), and in 1.73s in tsrs (dev), which is 3x faster.
3x isn't so wild but still very impressive.
Or we can just drop TypeScript completely and type-check JavaScript directly.
With HM-style type inference, row-polymorphism, sum types, type classes for overloading + operations, recursive types for object methods, and type annotations for edge cases.
I've been working on this since before TypeScript existed. Latest incarnation: https://github.com/sinelaw/inty
Demo: https://sinelaw.github.io/inty/ (you may need to reload once)
typescript is two things: a language, and a type system.
that same type system exists in the javascript language (jsdoc). doing those type checks is the 'slow' part that makes go / rust appealing. not just stripping types to emit js.
isn't it better to use JSDoc ?
I assume you mean jsdoc + tsc in js type checking mode (tsc --allowJs --checkJs --strict)?
inty infers more - including generics (parametric polymorphism), while tsc is more of a simpler local inference.
inty is more strict - it accepts less programs than tsc would. This is mostly on purpose (to our benefit) as some of those programs have unsound code like array covariant (downcasting array members of the wrong type). tsc makes it easier to slip through these bugs.
And if tsc + jsdoc is good then the same question remains (do we really need the TypeScript language?)
So that's three now. This, tsgo-rs, and `bun check`
They mention in the post they are working with Theo to absorb tsgo-rs into tsrs.
Why would they bother merging his vibe slop if theirs is better?
so this is software now?
youtuber vibe codes a transpiler says they're not going to bother maintaining it - but big company like cloudflare takes it seriously enough to absorbe into their own official services.
What is happening?
What is even software dev anymore?
What didnt the labs or cloudflare do this themselves, I mean they did, but then if it's all vibe coded, why do they need someone else's vibe coded stuff?
How is any of this serious?
Big tech looks out after their own. That's all there is to it.
Prob a PR thing with a quick merge to /dev/null
Nice. I missed that. This is good news.
Shameless plug of my Typescript C# port. It's a case study with some interesting (perhaps) notes from me in the README about approach, performance, and cost(~$75).
https://github.com/stepwisehq/typescript-dotnet
I wonder if something similar has been attempted with Python. CPython is a project where I don't know how they are taking advantage of AI.
The problem with Python is that the whole ecosystem is based on CPython’s implementation details, so every change inevitably breaks something and requires effort of package maintainers to fix compatibility. There is no test suite an agent can follow.
I guess it will be soon that typescript won’t matter anymore since agents will be writing code is Rust compiled to WASM or similar for everything. typescript added human readability or JavaScript, if no human is coding, I don’t see the point of wasting tokens on this.
> typescript added human readability
Ehm no, added type safety, and dealing with a type checker it's exactly what agents need.
> if no human is coding
Nope outside the bubble there are many still hand coding(me too)
> I guess it will be soon that typescript won’t matter anymore since agents will be writing code is Rust
Oh hold on a second isn't Rust the same as Typescript "adding human readability"?! Why not trashing Rust and just have LLMs output WASM directly? XD Wasm which is basically useless outside the browser or some microvms on the edge, and in the browser you don't even have an interface for directly rendering anything and still have to deal with the dom
You either are highly non technical, very junior or I just don't know
I'm tired of all the new tools that comes out every day, but I would say, this is basically the Vite team and if they implement it with Vite+, then I'm all for it, less decisions about tooling to be made, one config file, etc.
I ran out of patience before models got good enough to get this done.
https://tsz.dev
Congrats to the team on getting this out
All of this just for enabling the WASM use case? Why not keep supporting Deno then that is written in Rust? To me it just sounds like a technical ai journey and a blow to the go port
I think it is a response to tsc-ts and scriptc from Vercel. CloudFlare and vercel CEOs have a personal competition going on.
I benchmarked tsrs a few weeks ago. I was testing out tsgo, ts-rust, tsrs, and bun for typechecking on our monorepo. tsrs is definitely awesome, it is very memory efficient when bumping up the number of threads. Not as efficient as bun, but it has incremental checks, which bun does not have.
Apparently only supported under darwin and linux for arm64 and x64? At least via npm.
Y'all building the same thing.
I wonder how fast "Doom in TypeScript" runs in tsrs now. XD
I use oxc tooling and it's great to see the rust rewrite singularity in action and get to ride the fastest horses, but I would like that they made a branch for 7.2 as experiment to see how it compares. Maybe this ends in a win for Microsoft, they could cut operating expenses by keeping just Anders Hejlsberg in a hammock reciting invariants to claudpilot.
I think we'll see TypeScript ported to minecraft ported to rust ported to typescript before GTA6.
It's a nit, but do you guys need to embed the TS version in semver? https://www.npmjs.com/package/tsrs?activeTab=versions
Obligatory reminder that their needs from a TS compiler are extremely different from a regular web developers iteration cycle. They are trying to run millions of these while saving on costs.
You almost definitely don't need to immediately plan to switch.
I expect I'll switch once Vite starts shipping it by default.
> This enables new possibilities when combined with agents: You are now able to run a development toolchain end-to-end in a fully sandboxed environment on Cloudflare.
True. Looks like main use-case is AI Agent Sandboxing
this is insane! saw a quote tweet of this by an infra engineer at chat gpt who said “ I'm looking at migrating over for the ChatGPT codebase and saw ~70% improvements on both wall time perf + peak cpu/memory usage.
this is great for getting work done faster but also much better if you want to run concurrent agents/worktrees at scale”. https://x.com/feedthejim/status/2112584843368570921?s=46
then i went and tried this on my own open source harness. type checking went from 10 seconds to under half a second and uses a quarter of the memory. madness
I thought its TS port of theo
Theo's port is tsc-rs[0] as reference, and is significantly slower and uses more memory than this new version.
[0] https://github.com/pingdotgg/ts-rust
A new one but they're merging it with Theo's.
These days the trend seems to be rewriting everything in Rust.
I'm staying stainless
You should say you're staying fungal infection free as it's named after the Rust fungus not the chemical reaction even if that's what people think these days.
Did you know? Even stainless steel RUSTs in sea wind.
so rust is now go-to slop language? can't you guys use C or something. I want rust back.
Rust is basically the perfect language for LLMs because it has a strong, comprehensive, and strict compiler. For coding you really need a tight loop for models so they can iterate. C doesn’t accomplish this because it’s far too “anything goes” so bugs slip through. Dynamically typed languages don’t work either because their source code doesn’t contain enough context.
What? You realize how many people now are excited about Rust that were not before? I am, for one.
So excited that you avoid writing a single line of it at all costs.
Oh god, I wish this Rust fetishism would end soon.
please support TSRX (https://github.com/tsrx-org/tsrx)