iiinova — where your digital self keeps evolving

Bram Zijlstra @bramzijlstra.com · 18/05/2025

I barely post anymore here. The engagement is so low for the non-biggest channels, barely existing.

040

Bram Zijlstra @bramzijlstra.com · 03/04/2025

Alignment is a two way street

020

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

Every now and then I read posts about 90% of coding will be done by AI this year and I can't lie it makes me a bit nervous. Then a study likes this drops and I can relax a bit

140

Reposted by Bram Zijlstra

Sung Kim @sungkim.bsky.social · 31/03/2025

You’ve probably heard about how AI/LLMs can solve Math Olympiad problems ( deepmind.google/discover/blo... ). So naturally, some people put it to the test — hours after the 2025 US Math Olympiad problems were released. The result: They all sucked!

917350

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

I think this a big part of it. Developers talk about their experience with GenAI coding, but we come from way different places. I do a lot of solo prototyping for work, a lot of my code is a conversation starter. In this stage llms make you. 2-3 times faster, but in later stages it's maybe 20-30%

010

Bram Zijlstra @bramzijlstra.com · 29/03/2025

There are plenty of cases where survivorship bias doesn’t apply. We just don’t remember them.

120

Bram Zijlstra @bramzijlstra.com · 28/03/2025

Housekeeper: "I no longer think people should clean their own apartments" Uber Driver: "I no longer think people should own their own car"

040

Bram Zijlstra @bramzijlstra.com · 26/03/2025

I'm 179.5 cm, the metric equivalent. It's the most trustworthy length, no one would lie about it.

010

Bram Zijlstra @bramzijlstra.com · 26/03/2025

Errors with a 200 status code is like gift wrapping a turd

000

Bram Zijlstra @bramzijlstra.com · 26/03/2025

Have mixed feelings as well. One the one hand debugging while vibe coding is super frustrating. On the other hand, it has been amazing for throwaway demo's and prototypes. I can aim for things I wouldn't have the time for normally.

010

Bram Zijlstra @bramzijlstra.com · 25/03/2025

Excel is one of the safest messaging technology out there. No message leaves your device and no risk of the other person leaking information.

000

Bram Zijlstra @bramzijlstra.com · 25/03/2025

An infinite amount of monkeys with an LLM can write the entire the complete works of Shakespeare

110

Bram Zijlstra @bramzijlstra.com · 20/03/2025

My doctor told me he's into vibe surgeries lately

060

Bram Zijlstra @bramzijlstra.com · 18/03/2025

I like them, it's a good heuristic to not read any text around it

020

Bram Zijlstra @bramzijlstra.com · 18/03/2025

I think the reason why LLMs are overconfident is because we keep saying telling "You are an expert in" literally anything

130

Bram Zijlstra @bramzijlstra.com · 10/03/2025

I wrote my bachelor thesis on paradoxes in self-reference and I dit that. Felt obligated to that.

080

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

Electron apps are the Monkey's Paw to someone wishing cheaper RAM

000

Bram Zijlstra @bramzijlstra.com · 28/02/2025

Anyone know best practices / tips for improving LLM quality on classifying long texts? For shorter input few-shot learning and finding the best examples work well, but this not very practical with longer texts. #databs

020

Bram Zijlstra @bramzijlstra.com · 24/02/2025

We'll have AI alignment before MS Word alignment

040

Bram Zijlstra @bramzijlstra.com · 18/02/2025

An agent is an LLM that uses tools. A tool is someone who keeps saying '2025 is the year of the agents'

010

Bram Zijlstra @bramzijlstra.com · 14/02/2025

You said 'use a mixture of experts' so I invited every stakeholder in the company to the meeting

130

Bram Zijlstra @bramzijlstra.com · 08/02/2025

If I ask chatgpt for some advanced text processing tasks it will recommend contrived string manipulation instead of calling an llm. If I then ask it to use gpt it will use the openai library of an earlier major release

020

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

Today was a day where all emails found me well

020

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 · 07/02/2025

That's fine, just run it several times and in parallel. No downsides

010

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

Really want to like cursor but I am completely Pycharm brained it seems. Anyone else have the same ? Which copilot did you go for?

010

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

If you would make a coffee table book of these I would buy it

000

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

Haha so LLM labs tell us AGI is around the corner but we have to tell the LLM that they exist?

100

Bram Zijlstra @bramzijlstra.com · 06/02/2025

Humanity's Actual Last Exam

040

Bram Zijlstra @bramzijlstra.com · 06/02/2025

It's like going to a realtor and they suggest to live under a bridge

100

Bram Zijlstra @bramzijlstra.com · 06/02/2025

I'm always surprised that ChatGPT suggests old school NLP / string manipulation tricks instead of suggesting to call OpenAI. Seems like upselling 101 to me.

110

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

“Our company has moat” The moat:

020

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

I like my language models like my drinks, distilled to the point where it produces gibberish

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 · 30/01/2025

It’s like being a hoarder but also your diplomas, tax documents and passport are somewhere in the filth

110

Bram Zijlstra @bramzijlstra.com · 30/01/2025

This sounds like a wholesome version of trainspotting

010

Reposted by Bram Zijlstra

Mike Masnick @masnick.com · 29/01/2025

So, look. I'm sure I'm in the minority here on Bluesky in believing that training AI systems isn't copyright infringement. But, also. Dude. There's no way OpenAI can make this argument without looking very, very silly.

3847447

Bram Zijlstra @bramzijlstra.com · 28/01/2025

People have such binary brains and find it difficult to fathom that LLM's can be both super useful and unnecessary or useless. It's like saying "electricity is useless, I don't need it for cutting vegetables"

010

Bram Zijlstra @bramzijlstra.com · 28/01/2025

I’m fine with following, US news is world news. My biggest pet peeve is when Americans can only talk about other countries’ politics by relating it to the US. Don’t think I’ve seen a single mention of the South Korean martial law without immediately turning it into US politics

020

Bram Zijlstra @bramzijlstra.com · 28/01/2025

I doubt that a lot of people remember dogfooding with chatbots. I love llms but I haven’t encountered any that I like to use, and I don’t think that the product leaders use them themselves

030

Bram Zijlstra @bramzijlstra.com · 27/01/2025

Will copilots (eventually) negatively impact open source library development? Major updates will break any copilot, so it disincentives doing so.

220

Bram Zijlstra @bramzijlstra.com · 27/01/2025

Ethan, you talk a lot about interesting use cases of text based llms, what are in your opinion the value of image generators? Besides the wow effect I don’t really found a lot of use for them

010

Bram Zijlstra @bramzijlstra.com · 27/01/2025

No thanks, the bridge is open source and thus free. I've been driving over the bridge for a few days now. Curious to hear how my local weights act in the interest of the CCP

130

Reposted by Bram Zijlstra

Gergely Orosz @gergely.pragmaticengineer.com · 27/01/2025

Baffling to see the disclosure around DeepSeek Last I checked the tech industry, we celebrate small teams pushing the industry forward, coming up w novel ways to build software more efficiently. And sharing it. Yet now there's a class of folks who think this is some bad thing?

4361957