David Sainez

@davidsainez.com

woke tech bro he/him

We have an updated link below and would greatly appreciate if you took a few minutes to tell your Congress members to demand Sama's freedom. She is a Palestinian American citizen and a student, who is being unjustly and illegally held in Israeli military prison.

IMEU Policy Project@imeupolicy.bsky.social · 2w ago

Palestinian-American student Sama Safi is now being wrongfully detained in Damon Israeli military prison and has been enduring horrific conditions and abuse after a temporary release for medical treatment.

American culture is so weird when it comes to death. In Mexican culture, death feels more present and insubstantial and I think that ultimately helps people be more clear-eyed in situations like this. Dying is not a thing that rewrites your life.

Elizabeth Warren@warren.senate.gov · 4w ago

I got to know Senator Lindsey Graham traveling with him abroad in Afghanistan and Ukraine. We had just met to renew the fight for our Big Tech bill. Even though we disagreed on much, he was always willing to negotiate, with humor and wit. My heart goes out to his loved ones.

I’m at my mom’s house, checking in on her. I took a walk and saw two trans inclusive Pride flags on her block here in Biddeford, ME. It’s hard to be far from friends tonight on Pride, but it helped to learn community is near—just one of the ways these flags work. Thanks to everyone who flies them.

“Dear migrants, before I say any other word to you, I want to bow before your dignity. “You are not numbers or case files. “You are people — with a family and a home left behind, with dreams that no one has the right to scorn.” — Pope Leo XIV

Liz Soto and Savanna Batten are 2 of those sentenced to 50 years in prison for Prairieland. I visited them in a north Texas jail—11 months into their incarceration over a July 4 noise demo—to meet the Trump admin’s “Antifa terrorists”. from behind glass, Soto said, “They didn’t like my book club.”

‘This is injustice’: how leftist zines were used to sentence anti-ICE protesters to decades in prison

Advocates sound alarm after zines were used as evidence to convict protesters of terrorism charges tied to 2025 protest at Texas ICE facility

theguardian.com

crazy that if you harbor self-hate you will think that everyone hates you, but there's only one man in the US who can visit new york city and get boo'd from every corner of madison square garden and general city

Didn't get the job I was deeply hoping for, unfortunately. Need a remote TypeScript senior/staff-level position as soon as possible.

One thing that does not get talked about often enough is intelligence/token. For example, Qwen3.5 9B has a much higher artificial analysis score than Granite 4.1 8B. But if you compare the output tokens required to run the eval, it tells a different story.

A scatter plot comparing AI models by their Artificial Analysis Intelligence Index score versus the number of output tokens used. The chart plots six models—Claude 4.5 Sonnet, DeepSeek V4 Pro, Qwen3.5 9B, Granite 4.1 30B, Granite 4.1 8B, and Nova Lite—with a highlighted green quadrant representing the most efficient high-performance range.
Estimated Values (Intelligence Index score / Output Tokens):
Claude 4.5 Sonnet: 37 / ~7.8M
DeepSeek V4 Pro: 39 / ~14M
Qwen3.5 9B: 27 / ~78M
Granite 4.1 30B: 15 / ~4.6M
Granite 4.1 8B: 12.5 / ~4.0M
Nova Lite: 12.5 / ~7.0M
David Sainez@davidsainez.com · 2mo ago

Ah, right. Reminds me of this writeup: www.levs.fyi/blog/2-years.... They found that a model with weaker general benchmarks (nova lite) preformed about the same as sonnet 4.5 on the classification task.

This attempted refutation is incorrect. 1. They also looked at couples rather than households and found it was still true 2. It is also true for market incomes by age 30 (the early 20’s lag is easily explained by more education time) 3. The square root adjustment is standard statistical practice.

Bild
Andrew Parker@andrewparker.bsky.social · 3mo ago

In the paper, the "household sharing unit" *includes income from parents*. It's not "how much did this person in Gen Z make" it's "how much was the total income of the household the Gen Z person lived in divided by the square root of the number of members." 4/