Arnoldo

32 | Ingeniero en Computacion | No karate, no life

[Arnoldo @Jascrer]

Thy Flesh Consumed - Hell Beneath - UV First levelwith I try in UV and it was quiteto a challenge #DOOM #NintendoSwitch

[Arnoldo @Jascrer]

I've found so many incorrect approaches to a problem that I can't ever not read the plans...

[Maximilian @maxedapps]

Even if you don't read the code, you should read the plans & reports your agent may give you.

For example, when running code reviews, you will ALWAYS get findings. Probably with some "must fix / critical" ones.

Read them. Besides false positives, there often will be issues that are not issues for what you're building.

In my experience, agents simply can't say "yeah, it's fine, ship it". They will ALWAYS find stuff.
Don't fall into the trap of iterating more and more when you should ship.

[Ally-Seraphine main @allywooww]

Tengen Toppa Gurren Lagann es de los mejores animes que existen, no pienso discutirlo con nadie

[Aram Hăvărneanu @aramh]

Don't fall for this snakeoil. Using LLMs (which are universal function approximators) to approximate GCC, while impressive in a perverse way, is not the same thing as what they want you to think.

Ok, LLMs can produce a very poor implementation (of zero commericial value) of a very well understood problem while burning through a lot of tokens. Why am I supposed to be impressed?

[Aram Hăvărneanu @aramh]

Non-issues that people have brought up in regards to Anthropic's C compiler:

1. It uses GNU as and GNU ld. Irrelevant, GCC also uses GNU as and GNU ld! And respectable software like CompCert also uses GNU as and GNU ld (even worse, it calls those through GCC!).

2. It can't compile 16-bit x86 code. It's not feature complete, that's fine.

3. Code generation quality is bad. Again, that's fine.

Actual issues that people should bring up in regards to Anthropic's C compiler:

1. It doesn't do type checking. TBH, this makes it borderline fraudulent to call it a "C compiler". Now, I would be willing to brush this under "incomplete" moniker if it weren't for the next point.

2. The code is huge, fragile and resists any further modification. The article explains this. The LLMs can't add new features to it. It's the embodiment of technical debt. What exactly are you supposed to do with this artefact? It's also a disproportionately huge amount of code compared to the functionality. Code is debt.

3. I see 7 people named in the Anthropic article (plus a vague reference to "many other people"), this is not the set and forget type of thing they try to imply. People have been working full time to try to get this to work even in its current (broken) state. And they describe a feedback loop put together with spit and duct tape. Good luck scaling and maintainig this!

4. Compiling Linux is impressive but it fails to compile "Hello, World!", c'mon man. This just shows that the type of errors you get from LLMs artefact are very different then the errors people make even in the presence of specs and exhaustive tests suites. This is a hidden cost that people have no idea about its impact.

5. It uses GCC as an oracle. This won't work for a new ISA (or for a new language), but let's ignore that. The reality is that it's perfectly fine to use another compiler as an oracle and you would do it yourself if you'd write a compiler by hand. However, it is misleading because in real life, for new engineering problems you simply do not have this luxury. Not only you don't have an oracle, you don't have certified tests. Usually you don't even have a spec!

[Aakash Gupta @aakashgupta]

Sounds incredible until you read the fine print. The compiler generates less efficient code than GCC with all optimizations disabled. It doesn’t have its own assembler or linker. It can’t produce a 16-bit x86 code generator. And Carlini himself says it has “nearly reached the limits of Opus’s abilities.” New features and bugfixes kept breaking existing functionality.

So what did $20,000 and two weeks actually buy? A compiler that passes 99% of GCC’s torture tests but can’t match the output quality of a tool that’s had 37 years of human engineering. That’s the constraint nobody’s pricing in.

The real story is in the cost curve, not the capability demo. $20,000 for 100,000 lines means $0.20 per line of generated code. A senior compiler engineer costs roughly $150/hour. At maybe 50 polished lines per hour for something this complex, that’s $3/line. AI just did it at 15x cheaper, and it will only get cheaper from here.

But the code isn’t equivalent. The AI version needs a human to finish the assembler, fix the linker, optimize the output, and prevent regressions. Those are the hardest 20% of the problem, and they represent 80% of the engineering value. Anthropic built the demo. Shipping the product still requires humans.

This tells you exactly where we are in the autonomous software timeline. AI can now produce impressive first drafts of complex systems at trivial cost. Turning those drafts into production software still requires the judgment that costs $300K+ per year in compiler engineer salary. The gap between “compiles the Linux kernel” and “replaces GCC” is measured in decades of accumulated engineering wisdom that no model has internalized yet.

The companies that understand this will use agent teams to generate the 80% and hire engineers to finish the 20%. The companies that don’t will ship $20,000 compilers that produce slower code than a free tool from 1987.

[✨Wild Nimue✨ @wildnimue]

Hostias, qué pereza dais, de verdad.

Es justo lo contrario, y Tolkien lo dejó clarísimo mil veces, PESADOS. Odiaba la alegoría política y no estaba escribiendo ningún panfleto sobre civilizaciones superiores defendiéndose del bárbaro exterior. La historia no va de hombres fuertes salvando Occidente, va de gente pequeña, sin poder, sin épica y con miedo, intentando que el mundo no se vaya a la mierda. El mensaje central es que el ansia de dominación siempre corrompe, incluso cuando viene disfrazada de buenas intenciones. El poder no es la solución, es el problema, y por eso el Anillo no se usa jamás, se destruye.

Los héroes no son conquistadores ni líderes, son jardineros, amigos leales y pringaos agotados que tiran para delante a base de cooperación, cuidado mutuo y pura cabezonería.

Y sí, Tolkien era cristiano, pero la obra no es una alegoría cristiana ni pretende adoctrinar a nadie. No hay sermones, no hay salvación por obediencia, no hay jerarquías morales dictadas por dios alguno.

Así que no sé, igual antes de ir de profunda y provocadora convendría aprender a leer un texto sin proyectar cuatro ideas mal masticadas, porque alardear de nula comprensión lectora con esa seguridad tampoco es tan guay como te crees.

[Arnoldo @Jascrer]

Las estadísticas aunque vengan de un medio confiable parece que siempre van a confundir y eso mismo pasa cada 4 años en este país... háganse un favor y lean "How to lie with statistics" para entender mejor y no dejarse guiar por el número mas grande y en negrita

[cristian cambronero @cambronero]

El CIEP tiene q hacer algo con la comunicación eficaz de sus encuestas. Es impresionante lo generalizada q está la confusión entre la encuesta aleatoria y el panel.

Es muchísima la gente q no distingue un instrumento del otro, y eso distorsiona su recepción y credibilidad.

[cristian cambronero @cambronero]

El CIEP tiene q hacer algo con la comunicación eficaz de sus encuestas. Es impresionante lo generalizada q está la confusión entre la encuesta aleatoria y el panel.

Es muchísima la gente q no distingue un instrumento del otro, y eso distorsiona su recepción y credibilidad.

[Arnoldo @Jascrer]

There is a very good documentar about this called Wall-E (2008)

[Sovey @SoveyX]

“Universal high income.”

I’m sorry, but that should absolutely terrify you.

It terrifies me.
I don’t sleep very well because I can’t turn my brain off.

Ideas like this are the reason I take trazodone… or I wouldn’t sleep.

Our tech leaders are saying the quiet part out loud now, out in public, like it’s a cute little vision board moment.

The future they’re building is one where machines do everything, humans get paid to exist, and we all quietly retire into “comfort” like it’s a retirement plan instead of a slow-motion exit ramp from meaning.

Humans need purpose.
Not vibes.
I don’t care how much abundance or money is involved.
I need to provide value.
Not “content consumption.” Not a permanent allowance.

This feels like slipping society into a warm bath… and handing us the pill bottle.

I don’t have the perfect solution. But I do know this, I really don’t love the direction this is heading.

This is not good for us.

[Arnoldo @Jascrer]

@serrrfirat @GenAI_is_real Up to what point is this sustainable? because unspoken rules are really hard to wrasp and even then to describe them is not an easy task.
Current llms behave good with clear rules but on fuzzy rules they can hallucinate and these unspoken rules are good candidates of the later.

[f4mi ‼️ @f4micom]

in a world where subscription models for stuff that used to be free or local are becoming the norm and we are being treated like we HAVE to pay rent to some fuckass company to run shit on “the cloud” instead of owning our computer hardware or media or whatever the fuck my resolution for 2026 is to just not do that as much as humanly possible

[Progress Bar 2026 @ProgressBar202_]

2025 is 100% complete! Thank you all so much for following, and I hope you have a nice 2026.

[Gravantus @Gravantus]

This is the LAST day you can retweet this.

[Data Simpson @Data_Simpsons]

¡Carambilla! ¡1º de Enero! ¡Más vale darle a esos impuestos!

[José Valim @josevalim]

The key takeaway this year: static typing doesn't prevent runtime exceptions. If you assume it does, you'll write systems that will fall apart when bugs or unexpected conditions appear.

Types make you reason about errors statically, but mistakes and exceptions will still happen.

[Antonio Sarosi @antoniosarosi]

Curent state of AI:

- Helps you navigate codebases faster (finds key files, methods, tests, etc)

- Helps you write boilerplate code or repetitive but not quite copy-paste code

- Helps you write scaffolding code

- Can help write normal code but you better review it if you don't want to kill yourself 3 weeks from now

- Can help catch flaws in code that linters don't (if prompted specifically to review)

- Helps with writing design docs

- Helps with understanding complex data structures and algorithms used in other codebases but you better read the source code to verify it's not hallucinating information

That's it, it doesn't kill any software engineering anywhere at all.

And it won't until I see a model that I can prompt to write SQLite from scratch or something and it works first try without me guiding the model.

[Arnoldo @Jascrer]

@allywooww Ellos están esperando a debatir a Laura, Laura no va a debatir xq sabe q la desarman, entonces busca "alianzas" y votos en la calle.... y siendo muy crudo los debates no se si ganan gente, andar en la calle y tirar publicidad puede tener mas impacto... +popularidad +votos

[Hank Yeomans @HankYeomans]

The real reason people are upset about what @karpathy is saying is because what he’s saying is that everyone has a ways to go, has to keep working at it and will have to evolve for years. He’s saying “sorry that easy button you thought you were gonna hit and get to claim you’saved humanity’ you don’t get to hit yet. You don’t get to bad prompt your way through it.” He’s saying “what a time to be alive, but we’ve got work to do and there’s way, way too much hype around. Being an optimist that’s also grounded in current reality should he cheered. If you can’t think critically you don’t belong leading anyone, anything, or saving humanity. I’m a huge optimist about AI, but let’s be real, some things people come out with are shit.

[Arnoldo @Jascrer]

How do I battle the urge to create a terminal app that uses curl and jq to avoid the bloat all these clients bring??? Testing APIs should be simple

[JNS @_devJNS]

what do you use for testing your APIs?

[Arnoldo @Jascrer]

@gabriel_af70n El que lo encontro: