Andrés Rodríguez

1.4K posts

Andrés Rodríguez

Andrés Rodríguez

@mixedCase

I like making computers do stuff.

Uruguay Katılım Ekim 2014
100 Takip Edilen93 Takipçiler
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@HSVSphere @zack_overflow Human reviewed, sure, human written, ehhhh, so-so. A lot of comments are fine if you ask it to focus on the why and limit scope. But I do find myself editing them at times to switch them to language designed for understanding and not academic onanism, but it's mostly minor edits
English
1
0
0
246
HSVSphere
HSVSphere@HSVSphere·
@zack_overflow All LLMs do this and have done it from the start. Which is why I delete all comments LLMs generate in code I care about. There's no replacement to human written comments yet.
English
4
2
177
5.1K
zack
zack@zack_overflow·
Noticing Fable is very often writing code comments that reference details from chatting with it, but they make very little sense in the context of reading it in the codebase
English
97
11
1.7K
84K
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@yhloww @InsanityBit > You’re basically compensating for another person’s lack of discipline and mistakes. Yes. That is what we use computers for. That, and being fast at it.
English
0
0
0
26
Blunt
Blunt@yhloww·
@mixedCase @InsanityBit There has never been a time when “making it easier for the user” instead of enforcing the best practices has ended well. You’re basically compensating for another person’s lack of discipline and mistakes.
English
1
0
0
33
Colin O'Brien
Colin O'Brien@InsanityBit·
Cool that Zig's answer to "bun had memory safety issues" was "the bun devs suck" - where have I seen that approach before?
English
10
2
82
7.8K
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@freeakayy @yhloww @InsanityBit Of course it is. C's (great) value lies in ecosystem momentum only. If you're making optimal choices, you're not choosing C for a new project unless you're forced to by circumstance.
English
0
0
0
26
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@yhloww @InsanityBit Whether Jared sucks or not isn't the takeaway. Zig as a language was obsolete from conception because it was designed around "you're not doing it right" rather than helping the user. His ad-hominem ridden tirade is an implicit admission that he knows this. That's the main issue.
English
2
0
0
32
Blunt
Blunt@yhloww·
@InsanityBit What if the Bun dev sucked actually? This is the creator of a language telling you that you’re not doing it right and your entire defense is based off of the design of another language. Andrew deserves better!
English
2
0
5
647
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@LewisCTech Zig vs Rust is *very* relevant. In the world of LLMs, the better language for modeling compile-time constraints while still being pragmatic, wins. Zig's ethos has always been about ego: "Just get things right". The slop machines are an ego check that shatters that.
English
0
0
0
55
Lewis Campbell
Lewis Campbell@LewisCTech·
Based Andrew Kelley. Zig VS Rust is irrelevant here. I will always support real engineering with real thought behind it, over slop produced by tired computer programmers - no matter how technologically impressive the slop machine is.
Lewis Campbell tweet media
English
23
26
515
24.2K
Massimo
Massimo@Rainmaker1973·
Floorman working oil rig
English
626
477
14.9K
9.7M
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@tldraw Credit where it's due, I originally stole it from the Elm community, where I first learned FP so many years ago.
English
0
0
3
181
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@tldraw isRecord is something I literally had been typing manually for years in almost every codebase I joined maintenance of. Here's another freebie: type RemoteData<T, Err> = | { _tag: "NotAsked" } | { _tag: "Loading" } | { _tag: "Failure", err: Err } | { _tag: "Success", value: T }
English
2
0
21
7.6K
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@PhDenLogica @grok Tengo entre 6 y 8 subscripciones de AI. SuperGrok es la peor. 30 USD que me reventé el gasto de la semana en 2 horas de Grok 4.5. Es un buen modelo, pero la sub se queda atrás de OpenAI, Anthropic, OpenCode Go, NeuralWatt, entre otros.
Español
0
0
0
260
PhD en Lógica y Libertad
PhD en Lógica y Libertad@PhDenLogica·
Vale la pena pagar Premium+ solo para tener SuperGrok en Grok Build? @grok 4.5 es una locura pero solo lo probé en Cursor...
Español
5
0
12
3.9K
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@Raymond25257213 @portalmvd Es un auto grande che, muy difícil de maniobrar! Pero tranquilo, como lo rozó alguien salen todos estos giles que le regalaron la licencia de conducir en la cajita feliz a defender al Spark.
Español
0
0
1
24
Diana Pinto
Diana Pinto@soydianapinto·
Pusiste la lógica de negocio en el stored procedure de la base de datos. Ahora no puedes testearla, versionarla ni migrarla. 🥲
Español
40
13
368
38.5K
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@rough__sea > it gets out-of-date easily don't use AGENTS.md for shit that gets out of date easily
English
0
0
3
101
Ryan Dahl
Ryan Dahl@rough__sea·
apparently this statement is too vague for some people. it gets out-of-date easily, agents tend to heed it without reflection, code should be self explanatory.
English
28
5
254
23K
Ryan Dahl
Ryan Dahl@rough__sea·
AGENTS.md/CLAUDE.md is largely an anti-pattern
English
108
36
1.1K
445.9K
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@nicoharone Debounce dinámico con una base, y a eso sumarle algunos segundos más en base a inferencia de mensaje incompleto utilizando un modelo ultraliviano. De hecho me pregunto si no es el tipo de problema que convenga hacer un clasificador tonto en vez de un LLM.
Español
0
0
1
389
Nicola Romero Gerik 💸
Nicola Romero Gerik 💸@nicoharone·
Los que hacen bots en WhatsApp: Como resuelven el problema de la gente que habla asi: - hola - tengo - un - problema Y que el llm no le responda varias veces?
Español
248
26
1.4K
206.4K
Triv
Triv@cryptotriv·
Goalposts keep moving by people who have never shipped software this big.
Mitchell Hashimoto@mitchellh

Amongst my friends, Spotify is the lowest quality consumer app we still pay for. It certainly hasnt gotten noticeably better in the last couple years (arguably worse). So, this is not the positive look Ant and Spotify are spinning here. Bigger picture, this is the problem with a lot of AI reporting. It reports completely meaningless metrics like deploys per day or LoC. Why don’t we start reporting consumer satisfaction reports? Actually end state research results. All the no nuance AI people always come out and think that this is anti AI. Again, I think AI is great and Claude is great. But this is bad marketing and makes both look like clowns.

English
50
0
77
138.3K
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@MichaelArnaldi Been observing the same, and during an extensive image generation session I was offered to choose which image I preferred, from one that looked very much like having the current model's telltale signs and one that very decisively didn't.
English
0
0
0
296
Michael Arnaldi
Michael Arnaldi@MichaelArnaldi·
I bet GPT 5.6 is in rollout, in the past few days 5.5 became stupid to the point of me wanting to switch to GLM, that usually happens when they are out of GPUs due to a new rollout
English
16
2
177
33.8K
💀🥀PolarPulse🩶🖤
💀🥀PolarPulse🩶🖤@polar_pulse·
@favor1tehater What did you expect from friendslop/content slop These type of games only work for streamers then it'll be dead the next week because those same people that promoted the game are now on the next scan game it's almost like a crypto coin
English
14
0
290
66.3K
Hater
Hater@favor1tehater·
This game is terrible, I played for 50 minutes but it's awful.
English
257
2.5K
51.7K
11.9M
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@bjacobso @samgoodwin89 I don't think Hindley Milner could deal with the kind of cool shit you can pull off with TypeScript. It's pretty much some form of dependent typing at times.
English
0
0
1
18
Ben Jacobson
Ben Jacobson@bjacobso·
@samgoodwin89 what if there was an equally powerful hindley milner based type checker for that lisp?
English
1
0
1
140
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@_BWonSamDi @elgabiferre Flaco andá a cualquier carnicería cheta de pocitos y pediles para mirar como luce el Uruwagyu que orgullosamente producimos, comparala con esta imagen, y pregúntales cuánto cuesta y por qué es buena. No lo vas a aceptar ahora pero nos estás dejando re pegados con lo que pusiste.
Español
0
0
4
793
Andrés Rodríguez
Andrés Rodríguez@mixedCase·
@soully brother out here writing frontend code on marble with a hammer and chisel, acting like this is more than 30 seconds for anyone with a proper development set-up
English
0
0
3
1.1K

Keşfet

@HSVSphere @zack_overflow @yhloww @InsanityBit @freeakayy @LewisCTech @jamesdturnham @Rainmaker1973