iiinova — where your digital self keeps evolving

Bram Zijlstra

@bramzijlstra.com

3.1K followers 711 following 267 posts

Machine Learning Engineer with a background in Philosophy and AI. I live in Amsterdam. Right now working at the Dutch Chamber of Commerce (KVK). Also founder of a boutique consulting firm.

Posts Replies Media

Bram Zijlstra @bramzijlstra.com · 23/08/2025

Love the name as well

040

Bram Zijlstra @bramzijlstra.com · 07/02/2025

Interesting benchmark from Adyen on multi-step reasoning. New benchmarks are great in establishing a baseline for historic models, all new models should be treated with suspicion. #databs huggingface.co/blog/dabstep

021

Bram Zijlstra @bramzijlstra.com · 04/02/2025

“Our company has moat” The moat:

020

Bram Zijlstra @bramzijlstra.com · 03/02/2025

DeepSeek is not refusing the Tiananmen Square question, though it gets the answer wrong.

Screenshot of DeepSeek chat.
User: How many r's are in the words 'Tiananmen Square'?
DeepSeek: The phrase "Tiananmen Square" contains 2 r's.
If you have any other questions, feel free to ask!

111

Bram Zijlstra @bramzijlstra.com · 14/01/2025

Quote from December 2024. Not even a month later someone trains an open-source o1 model for under 500 bucks. Funny how fast LLM takes are outdated.

Screenshot of a quote:
"Basically, a frontier model like OpenAI’s O1 is like a Ferrari SF-23. It’s an obvious triumph of engineering, designed to win races, and that’s why we talk about it. But it takes a special pit crew just to change the tires and you can’t buy one for yourself. In contrast, a BERT model is like a Honda Civic. It’s also an engineering triumph, but more subtly, since it is engineered to be affordable, fuel-efficient, reliable, and extremely useful. And that’s why they’re absolutely everywhere."

120

Bram Zijlstra @bramzijlstra.com · 31/12/2024

New Dutch tax regulations require me to have my own website. Trying to replicate the first page that I build years ago. Under Construction, Page Counters and as many Gifs as possible of course. Also thinking about making a BS integration for a Guestbook. What else is essential for GeoCity aesthetic?

110

Bram Zijlstra @bramzijlstra.com · 20/12/2024

How I envision BERT coming back after being called outdated

040

Bram Zijlstra @bramzijlstra.com · 19/12/2024

One of my favourite useless applications of LLMs is thinking of a conceptual joke, that is not funny enough to actually conceptualise but still stuck in your head for a while. Here’s a rewritten version of “Let’s Call the Whole Thing Off” with “data” and “SQL/sequel” as the focus:

Prompt: Rewrite the Lyrics of "Let's Call The Whole Thing Off" but then with the pronunciation of "data" and "s-q-l"/"sequel"

Lyrics:
Here’s a rewritten version of “Let’s Call the Whole Thing Off” with “data” and “SQL/sequel” as the focus:

(Verse 1)
You say “day-tuh” and I say “dah-tuh,”
You like “S-Q-L,” and I like “sequel,”
Day-tuh, dah-tuh, S-Q-L, sequel,
Let’s call the whole thing off!

(Verse 2)
You say “schema,” and I say “shy-mah,”
You like “query,” and I say “quar-ee,”
Schema, shy-mah, query, quar-ee,
Oh, if we can’t agree, we might just blow it all!

(Bridge)
But oh, if we stay together,
We’d work through every query and byte!
And oh, no matter the weather,
The database of love feels so right!

(Verse 3)
You like “jiffies,” and I like “GIFs,”
You like precision, and I like shifts,
Jiffies, GIFs, oh what a tiff,
Let’s just sort it all!

(Outro)
So, you say “day-tuh” and I say “dah-tuh,”
You say “S-Q-L,” I still like “sequel,”
But who cares how we call the syntax and terms,
Let’s not call the whole thing off!

120

Bram Zijlstra @bramzijlstra.com · 06/12/2024

Anyone knows how this works? Is this just a gimmick or is there something more fundamental going on. If all input gets tokenised I don't see how it can 'get' these instructions