Sabitlenmiş Tweet
Fernando Irarrázaval🔸
13.2K posts
Fernando Irarrázaval🔸
@Cucho
https://t.co/kf71l4jKWI Head of Data at Enveritas (YC S18). Prev @JPAL & @WorldBank. Master in Economics @Kennedy_School. 🔸 10% Pledge
Santiago, Chile Katılım Nisan 2008
1.6K Takip Edilen2.1K Takipçiler
@xxxxxxx03474898 @awaissbluth d) haciendo crecer el gasto más lento que otras administraciones.
(la deuda neta se ve en el gráfico y también parece crecer más lento, no?)
Español
¿como freno el aumento de la deuda?
(a) aumentando ingresos
(b) disminuyendo gastos
(c) vendiendo patrimonio, disminuyendo caja y chuteando pagos a proveedores y pagos.
Alternativa (c)
Es decir, es sólo para la foto, no es una frenada real.
Nota: todo quien queira informarse de economía real, no debe mirar una sóla variable (deuda bruta), debe mirar todos los factores: deuda neta, pagos comprometidos (doblemente grave cuando no quedaron provisionados, como paso), etc
Español
El mismo CFA que citan muestra que gobierno de Boric es el que logra frenar aumento de la deuda. Titulares de los diarios lo ponen como si fuera el culpable de la deuda (que además está dentro de rangos adecuados mundiales) pero gráfico muestra el mejor manejo de los últimos años
Español
Fernando Irarrázaval🔸 retweetledi
- Drafted a blog post
- Used an LLM to meticulously improve the argument over 4 hours.
- Wow, feeling great, it’s so convincing!
- Fun idea let’s ask it to argue the opposite.
- LLM demolishes the entire argument and convinces me that the opposite is in fact true.
- lol
The LLMs may elicit an opinion when asked but are extremely competent in arguing almost any direction. This is actually super useful as a tool for forming your own opinions, just make sure to ask different directions and be careful with the sycophancy.
English
@Jasey0890 @scottjwillis @OlisaOsega 100%. Basically:
- he has a low number of shots compared to other arsenal strikers
- has an average shot -> shot on target conversion
- has a very high SoT -> goal conversion
And people are using that last stat to say nobody passes him the ball.
English
English
Gyökeres has scored 16 goals for Arsenal this season from 26 shots. Do you know how bad attacking play has been this season that our striker has had only 26 shots all season and we are in March. 16 goals in 26 shots also shows that Gyökeres is very clinical. Give him the ball ffs
English
@OlisaOsega He’s taken 70 shots. I have no idea where you would get 26 from. He has 43 in the Premier league, 22 in the champions league, 4 in the league cup and 1 in the FA cup.
English
@ben_j_todd People are way too harsh on pie charts. For this, it is a great fit.
English
It took over 40 versions and far too many hours to design this book cover.
Everyone sees the pie chart…but they miss the other reason we went for it.
There are haters, but I'm pretty happy with it in the end.
English
Fernando Irarrázaval🔸 retweetledi
VIDEO | “Este no es un país que se cae a pedazos”: Ignacio Briones analiza el escenario fiscal y los desafíos estructurales de Chile adnradio.cl/2026/03/25/est…
Español
How tf do you have +100 commits per day? Serious question.
My agents don't commit unless I ask them to. Even so, I max out at 20-30 meaningful commits a day. Am I missing something?
English
Fernando Irarrázaval🔸 retweetledi
FT Exclusive: Traders made bets worth half a billion dollars in the oil market about 15 minutes before Donald Trump’s post touting 'productive' talks with Iran sent the price of crude tumbling ft.trib.al/5jnFcCt
English
El Temucuicui de Kast
Cadem_cl@Cadem_cl
👎#Cadem 11 días alcanzó a sostener el presidente Kast la alta aprobación con la que comenzó su gobierno (57%). La crisis por el alza en el precio de los combustibles impactó de manera significativa su relación con la opción pública y la desaprobación es por primera vez más alta
Català
@jthenriquezc Pero China, Rusia, Francia y Reino Unido también pueden bloquear. Eso cambia la ecuación.
Español
Voy a repetirlo hasta que me canse. EEUU no le va dar apoyo en el consejo de seguridad. Luego, es inviable. Lo que diga o no Ignacio walker es i-rre-le-van-te.
Catalina Fernández Carter@catafernandezc
Ex Ministros/as y ex Subsecretarios/as de Relaciones Exteriores lamentan retiro de apoyo a candidatura de Pdta Bachelet, indicando que "tiene condiciones que hacen que su candidatura sea considerada una de las que tierne mayores posibilidades de éxito" y que esto es un "bochorno"
Español
@steipete what's in your prompt? this output looks like you told it to be very efficient with words
English
Pretty much every PR I review:
0) review [codex does it's thing and finds issues]
1) is the issue clear? [if not, trash PR]
2) is this the best possible fix? [95% of the time no]
3) continue discussion, consider tradeoffs, usually rewrite PR
Most folks send too localized, small fixes that would end up making the project unmaintainable.
English
Fernando Irarrázaval🔸 retweetledi
Whether the next president is a Republican or a Democrat, can we have someone with at least a little class next time? Because this is embarrassing.
English
@anmolmoses @steipete @openclaw agreed. looking more closely, gpt-5.3-codex was evaluated on different tasks from the rest. pinchbench.com/?view=graphs&g…
English
Interesting benchmark on which model is best for @openclaw pinchbench.com
English
@steipete @openclaw Weird. gpt-5.3-codex was evaluated on different tasks from the rest.
pinchbench.com/?view=graphs&g…
English
If I'm understanding this correctly, they weren’t evaluated on the same tasks?
pinchbench.com/?view=graphs&g…
Moritz Kremb@moritzkremb
There's finally a proper benchmark for @openclaw model performance. I just found that @kilocode built an open source benchmark that tests models across 23 real world openclaw tasks like scheduling meetings, writing code, triaging email etc gpt-5.3-codex is sitting at number one. tbh that matches my experience. gemini 3 flash in second place. didn't expect that. curious to see where gpt-5.4 will land on this.
English
@moritzkremb @openclaw @kilocode Very surprised. I have had better results with opus 4.5 and 4.6. Will test again gpt-5.3-codex.
English
There's finally a proper benchmark for @openclaw model performance.
I just found that @kilocode built an open source benchmark that tests models across 23 real world openclaw tasks like scheduling meetings, writing code, triaging email etc
gpt-5.3-codex is sitting at number one. tbh that matches my experience.
gemini 3 flash in second place. didn't expect that.
curious to see where gpt-5.4 will land on this.
English
Had a retro with my team just now. The whole hour was about AI.
We're figuring out how to review code we didn't fully write. How to stay owners of systems we didn't fully design. How to be productive without losing the thread of what we're building.
No clean answers.
English
It wasn't always the same at one point, in fact some of us were called crazy for a while until Claude eventually acknowledged a degradation in model quality
I however don't think Claude is purposefully nerfing models upon release. Could be a symptom of infra issues, exacerbated with the most recent outages likely due to an influx of new users
English
We need some whistleblowers to confirm our gut feeling they do this, it's getting annoying
If no whistleblowers show up we can all sleep in peace again
Dean Fiacco@DeanFiacco
@levelsio They nerf these models after release. It’s insane
English