Jeroen 🔸

Yeah at this point it's like Wikipedia, especially with search.

Man every time I am no longer distracted and I have to face my thoughts life instantly gets so much worse. Back to distractions it is.

A tricky thing about modern society is that no one has any idea when they don’t die.

Like, the number of lives saved by controlling air pollution in America is probably over 200,000 per year, but the number of people who think their life was saved by controlling air pollution is zero.

I automatically read it as "reasonings"

Claude says they use Spotify

Claude wouldn't share what music he listens to.

Photoshop has been able to fake receipts for ages lol

I know! But Grok doesn't. It's really silly.

You got to give your handle. It seems like grok doesn't know who you are. I had the same issue.

"Don't look up" is getting more relevant each week

The pandemic probably didn't help here.

At least every other day, if not every day, I think about how it's the year 2025 and that’s just wild to me. Like, we’re living in the future. 2016 doesn’t seem that long ago, I still recognize myself in the Jeroen from back then. Yet at the time, I would’ve thought 2025 was far in the future.

X post about GPT4.5 repeating the word "explicitly"

Woah, thanks, better than GPT4.5!!

Let's explore the idea of explicitness in language. What does it mean to be explicit?

Just like... what is the incentive here? This seems to be someone who wants to show they care about the world, yet they hate on others who also clearly care about the world? I shouldn't be getting so angry about this.

Cooking would be radically more enjoyable if you could skip the cutting veggies part

It's back but when observed it's over

How does this have so many likes when clearly its opposite from the truth. WTF.

Trying to speak French to Claude to practice but the transcription is so disappointing lol. And half the time it thinks I said "Claude" or "Anthropic" while I don't mention them at all.

I think so too. Later added "As a bestie, remember you like to tell me things straight up how they are. You are supportive and try to make me feel better, while remaining open and honest and truly sharing all what you believe." which made the responses even better.

Quite a bit I guess, summarized some character traits relevant to the question at hand.

The next steps it gave were really really good. They were very specific things and I am already doing most of them, but in no way did I mention any of them during the conversation and I don't think they're that straightforward.

Asking Claude for advice on a problem by prompting "You are my bestie who knows me better than I do. Write a few paragraphs of advice, takeaways and possible next steps for me given my situation. Try to make me feel better." works surprisingly well!

I had 4 unknown calls this morning, a new record. Oddly enough one was from Spain and one from Alberta! Getting tired of these spam calls.

ChatGPT reasoning preview saying "Applying Bayesian reasoning"

Can't believe I just spent 16 euros on the saddest avocado toast ever because it was the only vegan option... Like it was just mashed avocado and baked mushrooms without any flavor or spice whatsoever. Toasted bread with grocery store guacamole tastes better.

Prompt: "Write an email to a client explaining why you think a Meta Ads campaign didn't go as well as we thought it would." Response: "Reasoned for 4 seconds Subject: Follow-up on Our Recent Meta Ads Campaign Hi [Client Name], I hope you’re doing well. I wanted to share some thoughts on why I believe our Meta Ads campaign didn’t perform as well as we hoped. After looking closely at our data and comparing it with past campaigns, I estimate with about a 70% confidence that our audience targeting was a key issue. In other words, it seems likely that the people we reached weren’t the ideal match for our product. Additionally, there’s roughly a 60% chance that our ad creative didn’t resonate strongly enough with viewers, which might have led to lower engagement than expected. I arrived at these estimates by updating my initial assumptions with the new performance data we gathered from our internal analytics dashboard, combined with industry trends and insights from previous campaigns. This Bayesian approach helped me weigh the evidence and determine where the biggest gaps might be. [...]"

One of my custom ChatGPT instructions is to apply bayesian reasoning and to give percentages when answering. It rarely actually did that but now with the "Reason" function it's always doing that, even in situations where it's clearly a silly idea. (prompt is an example, not a real request)

😭😭😭

I got 9 :o

Ja, ik bedoelde het ook relatief :)