Fix JuanPabloF -- 🌈🇺🇦

Tweets about functions and types. Likes programming languages and abstract algebra. Likes aren’t endorsements. He/him

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

When you NEED to abstract over passing a parameter.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@mweinbach @edzitron Unless they want to stop training models entirely, then they will need to amortize capex into API pricing.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@Miki__Yoshi @ZacksJerryRig As temperature increases, radiation becomes more efficient at dissipating heat.

Note how the Y-axis is logarithmic. GPUs must be kept at under 100C, so heat emission will be very inefficient compared to an orbital nuclear reactor.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@CreatedByJannn Amateur mistake.

You never serve user content from your app domain.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@brandonjcarl @zrithvi @deepseek_ai @MiniMax_AI Regarding (3): are you talking about layer offloading with sparse activation or something else?

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@Samhanknr @danluu Unless you have infinite VC money, you'll need to eventually amortize training costs into inference costs. The economics change substantially in that case.

Not doing so means frontier labs stop training new models and their moat is gone.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@danluu Inference is profitable, yes. However, frontier labs are caught in a treadmill where they need to constantly train models to stay ahead of open weights models.

Inference costs necessarily need to amortize model training. Otherwise open weight models catch up and eat their lunch.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@enajas 1. Limitando la expansión geográfica de la ciudad. Incluso el centro norte está lleno de predios abandonados y terrenos baldíos y sub-utilizados

2. Otorgando autonomía recaudatoria y de ejecución de obra pública a los barrios de Quito. Adicional al Municipio del DMQ.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@enajas Sería increíble reducir el número de parqueaderos en la Shyris y añadir cientos de arupos.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@CryptoCyberia @abra55513382 Basically says LLMs can be profitable if they stop training better models.

Interviewer then asks about open source models catching up and eating his lunch in that case. Dario doesn’t have a good answer and rambles until stopped by the interviewer.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

How is "Various units" a reasonable axis label?! I expect nothing from CloudWatch, yet it manages to disappoint me.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@specialCaseDev Especially when they happen to rhyme 😍

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@Feru Que no falte el tigrillo!

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@HKGuwu @int_index CI is supposed to tell you if your code is broken in production. If you introduce behavior deviations only for CI, your software has the potential to only work in CI.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

> linear
> early returns

Pick one

[htmx.org / CEO of #Schizophreniamaxxing (same) @htmx_org]

as CEO of IF statements allow me to address this slander: the problem here is the deep nesting

the fix is to invert the if conditions & use early returns to linearize the whole thing

also extract local vars for the test conditions to make debugging easier

plodding, linear code

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@ShriramKMurthi Euronet targets touristy areas. You often find them in the most random places. It’s one of the warning I give people going to Europe

Ignoring your rejection is par for the course for them.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@tollyferro @NiceDay42 @JLarky This is mostly because `return` is a mistake. Its goto-like semantics are the gross part.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@DebugSteven Personally, I did think about this. It’s rough socially when growing it out, since it WILL look bad and people won’t like it at first. It’s an experience all bearded people remember.

Many cis men can only grow it well into adulthood, so it changes your sense of self somewhat!

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

@DebugSteven It looked awful for a few weeks and I was tempted to shave it every morning. It got worse with time and everyone was telling me to shave.

I persisted, went to a good barber shop and now I love it! Definitely more work than clean shaved and it can easily look bad if unkempt.

[Fix JuanPabloF -- 🌈🇺🇦 @lambdase]

The Scala urge to add an ugly PartiallyApplied class for partial type application for a function with a single call site.