uhh. turns out we weren’t typechecking like half our code?
I need to write a blog post about this
I need to write a blog post about this
Add moduleSuffixes to tsconfig by mozzius · Pull Request #11146 · bluesky-social/social-app
Resolve platform-specific modules native-first ([".ios", ".android",
".native", ""]) during typechecking, matching the react-native
customConditions already ...
github.com
July 12, 2026 at 6:12 PM
uhh. turns out we weren’t typechecking like half our code?
I need to write a blog post about this
I need to write a blog post about this
The Complete and Easy Bidirectional Typechecking algorithm is neither complete nor easy
May 1, 2025 at 2:56 PM
The Complete and Easy Bidirectional Typechecking algorithm is neither complete nor easy
 that I thought people interested in programming language theory might find interesting.
The context is that Grace was originally based on a bidirectional typechecking system (specifically: [Complete and Easy Bidirectional Typechecking for Higher-Rank Polymorphism](https://arxiv.org/abs/1306.6032)) and as Grace evolved I began to run into a few limitations of that approach. Some of these limitations I could workaround with various hacks but recently these workarounds piled up to generate a bizarre bug.
### The problem
To illustrate the bug, consider this Grace program:
```python
let authorities = [ { domain: "google.com" }
, { domain: "localhost", port: 8080 }
]
for { domain, port = 443 } of authorities
in "${domain}:${show port}"
```
You can read this program as describing a list comprehension:
> Iterate over each record in the list of `authorities` and for each one bind the `domain` and `port` fields to variables of the same name, defaulting `port` to `443` if absent. Return `"${domain}:${show port}"` for each authority record.
I would expect a program like this to return:
```python
[ "google.com:443", "localhost:8080" ]
```
… but the bugged version would return:
```python
[ "google.com:443", "localhost:443" ]
```
What gives? The `port: 8080` field in the second record is being completely ignored.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:6ouaxjxgcbwzw6porooiafil/bafkreibc6i2s3qpeh33cxpo2qpiq55q2xuxrp2tzwq5otd26fpxxru2oom)
I wrote a thing!
Bidirectional Typechecking is Bidirectional
https://cybercat.institute/2025/01/28/bidirectional-typechecking/
Bidirectional Typechecking is Bidirectional
https://cybercat.institute/2025/01/28/bidirectional-typechecking/
Bidirectional Typechecking is Bidirectional
For a while I've had an intuition that typechecking should be optical, with subterms going forwards and types (and type errors) for subterms going backwards. In this post I'll make sense of this idea, by implementing a bidirectional typechecker for simply typed lambda calculus as an optic, using Haskell's Control.Lens.
cybercat.institute
January 28, 2025 at 4:48 PM
I wrote a thing!
Bidirectional Typechecking is Bidirectional
https://cybercat.institute/2025/01/28/bidirectional-typechecking/
Bidirectional Typechecking is Bidirectional
https://cybercat.institute/2025/01/28/bidirectional-typechecking/
Deno 2.6 is here:
🛠️ `dx` is the new `npx`
⚡ faster typechecking with tsgo
🔒 improved security with `deno audit --socket`
🦺 safer deps with `deno approve-scripts`
🚘 source phase import support
and more!
deno.com/blog/v2.6
🛠️ `dx` is the new `npx`
⚡ faster typechecking with tsgo
🔒 improved security with `deno audit --socket`
🦺 safer deps with `deno approve-scripts`
🚘 source phase import support
and more!
deno.com/blog/v2.6
December 11, 2025 at 5:14 PM
Deno 2.6 is here:
🛠️ `dx` is the new `npx`
⚡ faster typechecking with tsgo
🔒 improved security with `deno audit --socket`
🦺 safer deps with `deno approve-scripts`
🚘 source phase import support
and more!
deno.com/blog/v2.6
🛠️ `dx` is the new `npx`
⚡ faster typechecking with tsgo
🔒 improved security with `deno audit --socket`
🦺 safer deps with `deno approve-scripts`
🚘 source phase import support
and more!
deno.com/blog/v2.6
reading the ghc typechecking code
December 9, 2024 at 5:47 PM
reading the ghc typechecking code
Something subtle from @vscode.dev's recent release notes: VS Code core, and all its built-in extensions, are now type-checked with TypeScript 7 native previews!
code.visualstudio.com/updates/v1_1...
code.visualstudio.com/updates/v1_1...
May 19, 2026 at 5:57 PM
Something subtle from @vscode.dev's recent release notes: VS Code core, and all its built-in extensions, are now type-checked with TypeScript 7 native previews!
code.visualstudio.com/updates/v1_1...
code.visualstudio.com/updates/v1_1...
Working in a programming language with immature tooling sure makes me grateful for all the nerds who've put time and energy into making sure I could have nice things like typechecking, error handling, and robust stack traces
November 6, 2025 at 1:49 AM
Working in a programming language with immature tooling sure makes me grateful for all the nerds who've put time and energy into making sure I could have nice things like typechecking, error handling, and robust stack traces
the typechecking keeping bad data out of bsky
July 11, 2023 at 8:29 PM
the typechecking keeping bad data out of bsky
you can now use .ts files in the playground:
svelte.dev/playground/7...
caveat: only supports erasable syntax (www.typescriptlang.org/docs/handboo...). under the hood this uses bloomberg.github.io/ts-blank-spa...
(note that typechecking/intellisense are TODO)
svelte.dev/playground/7...
caveat: only supports erasable syntax (www.typescriptlang.org/docs/handboo...). under the hood this uses bloomberg.github.io/ts-blank-spa...
(note that typechecking/intellisense are TODO)
TypeScript! • Playground • Svelte
Web development for the rest of us
svelte.dev
April 9, 2025 at 8:09 PM
you can now use .ts files in the playground:
svelte.dev/playground/7...
caveat: only supports erasable syntax (www.typescriptlang.org/docs/handboo...). under the hood this uses bloomberg.github.io/ts-blank-spa...
(note that typechecking/intellisense are TODO)
svelte.dev/playground/7...
caveat: only supports erasable syntax (www.typescriptlang.org/docs/handboo...). under the hood this uses bloomberg.github.io/ts-blank-spa...
(note that typechecking/intellisense are TODO)
aron is notable for actually running his lean code rather than just typechecking it
May 28, 2026 at 3:55 PM
aron is notable for actually running his lean code rather than just typechecking it
Went down a rabbithole around why we use the TYPE_CHECKING constant in Python.
vickiboykis.com/2023/12/11/w...
vickiboykis.com/2023/12/11/w...
Why if TYPE_CHECKING?
Typechecking is brittle yet important
vickiboykis.com
December 12, 2023 at 12:31 AM
Went down a rabbithole around why we use the TYPE_CHECKING constant in Python.
vickiboykis.com/2023/12/11/w...
vickiboykis.com/2023/12/11/w...
"Contributing upstream: Both projects are active and easy to contribute to. While working on the migration I hit a bug in Pyrefly, and I recently fixed it upstream. It was a nice feeling to give something back to the tool we now depend on." ❤️🔥🥲
napari.org/island-dispa...
napari.org/island-dispa...
From Any to Certainty: A Typechecking Journey — napari blog documentation
napari.org
September 24, 2026 at 1:38 AM
"Contributing upstream: Both projects are active and easy to contribute to. While working on the migration I hit a bug in Pyrefly, and I recently fixed it upstream. It was a nice feeling to give something back to the tool we now depend on." ❤️🔥🥲
napari.org/island-dispa...
napari.org/island-dispa...
MPS coderules seems wild jetbrains.github.io/mps-coderule...
Typechecking Extended Lambda Calculus | MPS Coderules
jetbrains.github.io
January 8, 2026 at 3:18 PM
MPS coderules seems wild jetbrains.github.io/mps-coderule...
I FINALLY UNDERSTOOD BIDIRECTIONAL TYPECHECKING AND DE BRUIJN INDICIES AND EXISTENTIAL TYPES AND ISORECURSIVE TYPES ALL WITHIN LIKE 3 DAYS, LIKE, AAAAA I ACTUALLY GET IT NOW THAT IVE PUT PEN TO PAPER AND DERIVED IT BY MYSELF, NOW IM GONNA IMPLEMENT IT IN KOTLIN MAYBE (maybe not, i hate programming)
March 15, 2026 at 9:23 PM
I FINALLY UNDERSTOOD BIDIRECTIONAL TYPECHECKING AND DE BRUIJN INDICIES AND EXISTENTIAL TYPES AND ISORECURSIVE TYPES ALL WITHIN LIKE 3 DAYS, LIKE, AAAAA I ACTUALLY GET IT NOW THAT IVE PUT PEN TO PAPER AND DERIVED IT BY MYSELF, NOW IM GONNA IMPLEMENT IT IN KOTLIN MAYBE (maybe not, i hate programming)
if i had endless time, sure, I'd write these hundred thousand lines of Lean4 proofs by hand. It'd be great fun, spiritually enlightening, etc etc
if i had endless time, heck, I'd perform the typechecking by hand too. Gotta spend eternity somehow
if i had endless time, heck, I'd perform the typechecking by hand too. Gotta spend eternity somehow
I got into coding at age 8. Coding by hand does bring me a certain measure of joy.
But there is so much to do, and there is little joy to be had in just writing the same boilerplate over and over again.
But there is so much to do, and there is little joy to be had in just writing the same boilerplate over and over again.
it feels so freaking good to code, dude. like human to IDE.
how do ai-pilled people not find joy in doing this by hand. i genuinely dont understand.
how do ai-pilled people not find joy in doing this by hand. i genuinely dont understand.
August 2, 2026 at 11:12 AM
if i had endless time, sure, I'd write these hundred thousand lines of Lean4 proofs by hand. It'd be great fun, spiritually enlightening, etc etc
if i had endless time, heck, I'd perform the typechecking by hand too. Gotta spend eternity somehow
if i had endless time, heck, I'd perform the typechecking by hand too. Gotta spend eternity somehow
I bet there are all kinds of interesting research problems around improving error messages for constraint-based typecheckers. It’s too bad that I am unqualified to pursue this work, for the simple reason that I have never actually read a typechecking error message in my life.
February 8, 2025 at 5:41 AM
I bet there are all kinds of interesting research problems around improving error messages for constraint-based typecheckers. It’s too bad that I am unqualified to pursue this work, for the simple reason that I have never actually read a typechecking error message in my life.
i think especially when you combine it with PL level typechecking (pipedream) it becomes something one can't live without
I know, I know: dealing with SQL is hard and ORMs seem like a great choice to avoid dealing with it. But sticking to "plain old SQL" can actually have a great developer experience! Here's why you might not need an ORM after all
www.youtube.com/watch?v=XEJx...
www.youtube.com/watch?v=XEJx...
You don't need an ORM - Giacomo Cavalieri | Lambda Days 2025
YouTube video by Code Sync
www.youtube.com
December 9, 2025 at 5:47 PM
i think especially when you combine it with PL level typechecking (pipedream) it becomes something one can't live without
Yesterday's MSP101 talk, by me.
Leftovers, Rightunders: Typechecking Thinnings Compositionally
youtu.be/BABpO4epxOo
Leftovers, Rightunders: Typechecking Thinnings Compositionally
youtu.be/BABpO4epxOo
MSP 101: Leftovers, Rightunders: Typechecking Thinnings Compositionally (Conor Mc Bride)
YouTube video by MSP — Strathclyde
youtu.be
September 23, 2025 at 1:15 PM
Yesterday's MSP101 talk, by me.
Leftovers, Rightunders: Typechecking Thinnings Compositionally
youtu.be/BABpO4epxOo
Leftovers, Rightunders: Typechecking Thinnings Compositionally
youtu.be/BABpO4epxOo
A few days ago I began to write up a Bluesky thread about some recent typechecker fixes I had worked on and the thread got long enough that I was, like "I should turn this into a blog post" so I did
haskellforall.com/2026/05/a-bi...
haskellforall.com/2026/05/a-bi...
A bidirectional typechecking puzzle
Type inference challenges for real-world JSON
haskellforall.com
May 5, 2026 at 1:18 PM
A few days ago I began to write up a Bluesky thread about some recent typechecker fixes I had worked on and the thread got long enough that I was, like "I should turn this into a blog post" so I did
haskellforall.com/2026/05/a-bi...
haskellforall.com/2026/05/a-bi...
i love type checking i love type checking i just spent like 10h nonstop working on typechecking code and i'm still not done aaaaa
May 4, 2026 at 7:46 PM
i love type checking i love type checking i just spent like 10h nonstop working on typechecking code and i'm still not done aaaaa
python's typechecking is NOT AT ALL better than Luau LMAOOOOOOO
January 29, 2025 at 10:22 AM
python's typechecking is NOT AT ALL better than Luau LMAOOOOOOO
Diving in to Python typechecking and stumbled on this @vickiboykis.com post
vickiboykis.com/2023/12/11/w...
The tldr: "The result of the deep dive is that I now no longer trust anything in the Python typing ecosystem."
I'm now confused as ever, but on a higher level and about more important things
vickiboykis.com/2023/12/11/w...
The tldr: "The result of the deep dive is that I now no longer trust anything in the Python typing ecosystem."
I'm now confused as ever, but on a higher level and about more important things
Why if TYPE_CHECKING?
Typechecking is brittle yet important
vickiboykis.com
August 31, 2026 at 12:53 PM
Diving in to Python typechecking and stumbled on this @vickiboykis.com post
vickiboykis.com/2023/12/11/w...
The tldr: "The result of the deep dive is that I now no longer trust anything in the Python typing ecosystem."
I'm now confused as ever, but on a higher level and about more important things
vickiboykis.com/2023/12/11/w...
The tldr: "The result of the deep dive is that I now no longer trust anything in the Python typing ecosystem."
I'm now confused as ever, but on a higher level and about more important things
I'm beginning think that for bidirectional typechecking I only want an `infer` and `check` judgment, but not a `subtype` judgment, even if that's tricky to actually implement
August 5, 2025 at 4:55 PM
I'm beginning think that for bidirectional typechecking I only want an `infer` and `check` judgment, but not a `subtype` judgment, even if that's tricky to actually implement
@napari.org reports its Pyrefly check at ~0.7s warm / 0.9s cold, versus ~15–25s cold in its previous mypy setup.
Fast checks change habits: run it locally, fix the thing, move on. Compute left for science.
Thanks for the great write-up on your experience!
napari.org/island-dispa...
Fast checks change habits: run it locally, fix the thing, move on. Compute left for science.
Thanks for the great write-up on your experience!
napari.org/island-dispa...
From Any to Certainty: A Typechecking Journey — napari blog documentation
napari.org
September 23, 2026 at 4:54 PM
@napari.org reports its Pyrefly check at ~0.7s warm / 0.9s cold, versus ~15–25s cold in its previous mypy setup.
Fast checks change habits: run it locally, fix the thing, move on. Compute left for science.
Thanks for the great write-up on your experience!
napari.org/island-dispa...
Fast checks change habits: run it locally, fix the thing, move on. Compute left for science.
Thanks for the great write-up on your experience!
napari.org/island-dispa...