Tired: I don’t read l the code at all anymore.
Wired: I only read the code. I don’t read clanker’s prose about it at all anymore.
1
6
183
It’s wild to me that this describes the exact approach I advocated as likely ideal for similar projects… like porting Bun to Rust.
Also somewhat wild that the nuance of AI’s role in such an approach survived whatever biases Grok might have on the topic.
Replying to @xah_lee @vittoriuz
Yes, it's a direct **port** (not rewrite) of the TS compiler to Go (typescript-go repo).
They built tools to autogenerate much of the Go code from the TS codebase, enabling near line-for-line translation in many places because Go's simple functions+data-structures style closely matches the original.
Hejlsberg said they tried direct AI translation but it risked subtle bugs/hallucinations on 500k+ LOC. Better: use AI/tools to generate deterministic helper programs for the port, then review for exact parity.
Hybrid: automation + careful engineering.
github.com/microsoft/typescr…
1
252
So stoked that shipping software is getting faster. I'm sure we were all getting impatient, waiting for... whatever the fuck is happening here.
1
1
142
This is, unironically, the best part of that song… which I otherwise can’t stand.
You’re unable to view this Post because this account owner limits who can view their Posts. Learn more
1
123
I don’t doubt this phenomenon exists, but I’d be cautious about ascribing intent or deliberate action to produce it.
Another very real possibility is that degrading quality over time is a natural consequence of the models’ capabilities, usage, and misplaced trust in both.
"Anthropic, OpenAl and Google release their new models with high quality from day one then slowly nerf them until the next model, so when the next model hits, its perceived as a bigger jump than it actually is"
sounds right what's happening
128
One of the most grating things I’ve noticed about AI/software engineering discourse: it’s drastically refocused discussion around how (mechanics of writing/shipping code), at the expense of what (products/projects) and why (problems those solve).
Sure glad we’re just shoving nondeterministic garbage into every crevice someone might try to find basic factual information. This is truly humanity’s greatest achievement.
67
Beardo Schmidt 🦗 retweeted
The autistic experience of being misperceived as complaining when you’re just nonchalantly describing something you’ve experienced.
Day 1 trying Effect: tried to use it the wrong way (knowingly disregarding limitations specified in docs); frustration!
Day 2: stepped back and used parts the right way, where I know they fit; joy!
Day 3: okay yeah I’m rewriting the stuff I have so far to use Effect
2
14
609
Ergonomic note: mapping between Effect (type) and Either/Option feels frictionful. I suspect this is because most of what I’m writing so far is composable lib stuff & I’m trying to avoid Effect.gen for it (bc I’m finding it leads to more predictable/composable APIs).
1
81
Beardo Schmidt 🦗 retweeted
I think I may be coming to the sad conclusion that autistic people are constantly misunderstood mainly because not enough people are interested in understanding us
Beardo Schmidt 🦗 retweeted
… how we are perceived, which can in turn impact our relationships and potentially other important aspects of our lives as well.









