Duco Janssen

723 posts

Duco Janssen

@ducoluno

Building The Authority Index - Signal over noise

amsterdam Katılım Şubat 2015

1.8K Takip Edilen1.1K Takipçiler

Duco Janssen@ducoluno·8 May

I’m very curious to see the outcome of the Elon vs. OpenAI case. Whatever the result may be, it will likely set an important precedent. But an even bigger and more challenging issue is how many organizations operate under the label of a non-profit while effectively functioning (indirectly) as “for-profit” entities. This is even more problematic than openly transitioning to a for-profit model. In too many situations, charities are (mis)used as vehicles for indirectly transferring money, influence, or power under the guise of serving public interests, often while promoting specific agendas. Maybe exposing this will not be the hardest part, but rather getting people to understand what is happening — because even when confronted with facts, many simply do not want to see it.

English

Duco Janssen@ducoluno·3 Nis

A lot of that data is anonymous, weakly verified, and lacks clear signals of authority. We’re training systems that influence decisions… without consistently knowing who wrote the information or why we should trust it. If authorship and credibility aren’t part of the equation, you’re not building intelligence — you’re scaling uncertainty.

English

Duco Janssen@ducoluno·3 Nis

Everyone’s obsessed with “high-quality data” for LLMs and agents. But here’s the uncomfortable truth:

English

Duco Janssen@ducoluno·2 Nis

An Authority Index might be a step in that direction.

English

Duco Janssen@ducoluno·2 Nis

If your LLM or agent treats all information as equally credible, it won’t just fail — it will fail with confidence. The real question is: how does it learn to separate signal from noise?

English

Duco Janssen@ducoluno·1 Nis

Hallucinations happen when models lack a sense of authority. If you guide them with the right signals, you don’t just reduce errors — you improve decision-making.

English

Duco Janssen retweetledi

Thomas Slabbers@Thomasslabbers·9 Mar

X can do it! Ignore the Pixel: Meta’s advantage is fading as cookies decline. Don’t play their game. Leverage your strengt: X’s unfair advantage is high intent, real time human interaction. Critical: Semantic matching isn't enough. An ad can match the topic but miss the buyer. To win, you need: 1) Ad relevance for the user (Context) 2) Audience relevance for the advertiser (Qualification) X is one of the few platforms where this can be done with just on-platform data, and become one of the most relevant and engaging ad platforms out there.

Elon Musk@elonmusk

Have you ever bought anything based on an ad on this platform?

English

902

Duco Janssen retweetledi

TeamNL🇳🇱@TeamNLtweets·9 Şub

OLYMPISCH KAMPIOEN OP DE KILOMETER 🔥 #TeamNL | #MilanoCortina2026

Nederlands

280

3.1K

122.9K

Duco Janssen@ducoluno·27 Oca

In the tech/AI race, many companies keep adding layers — features, tools, products — often on top of a weak core. Over time, complexity grows and control fades. We take a different approach. We focus on getting the foundation right continuously. Because a product is only as strong as what it’s built on. From there, we build without compromising control. Every decision, every layer, connected back to the core. Adding products/features will become easy and very fast.

English

Duco Janssen@ducoluno·26 Oca

Your best customer might buy you. Or copy you. If they copy you, they’re rebuilding yesterday while you’re already building tomorrow.

English

100

Duco Janssen@ducoluno·24 Oca

This is me all the time when trying to explain (with the best intentions) how we can help X/xAI teams with their products😅

Nikita Bier@nikitabier

How it felt when I had under a thousand followers

English

156

Duco Janssen retweetledi

Robin@xdNiBoR·24 Oca

x.com/Thomasslabbers… Topic specific timelines are a much needed feature! And X literally has people talking about EVERY single topic possible, so it's totally doable to make this. In fact, this was basically the project we made at the xAI Grokathon in London last week! X could make timelines for everyone. From bird watchers to robotic experts. X has it all.

Elon Musk@elonmusk

The @xAI team is working on providing For You tabs that are specific to topics. For example, a “For You AI” that is focused only on artificial intelligence with no political rage bait. This would be like automatically generated follow lists with content ranked by quality.

English

2.5K

Duco Janssen retweetledi

Robin@xdNiBoR·22 Oca

Yes! We also think that this is the way to solve onboarding. X truly has people for EVERY SINGLE niche. It's crazy when you look at the actual data. But doing this manually just can't be a good use of time, and especially not when a partner already has this, scaled to every single country (and many cities even) with over 20k topics. Nikita, let us help make X onboarding the best onboarding there is! 🙏

Nikita Bier@nikitabier

Over the last few months, we scoured the world for the top posters in every niche & country We've compiled them into a new tool called Starterpacks: to help new users find the best accounts—big or small—for their interests ⬇️ Reply below with a topic you're most interested in We'll be rolling out to everyone in the coming weeks.

English

2.6K

Duco Janssen@ducoluno·21 Oca

This post and the replies perfectly illustrate what goes wrong with X advertising, largely because X/Grok’s system optimizes for the wrong things. It prioritizes click behavior instead of connecting a real, relevant audience to the advertiser/product. The fix is simple. Very simple. I genuinely believe this can become the most relevant platform for advertisers when done right. We have multiple conversations every day with advertisers running into these exact issues. One of the biggest drivers of low-quality (often bot) traffic is the “optimize targeting” option. When enabled, X effectively opens delivery to almost any “user”. The result is a self-fulfilling loop: low-value clicks get optimized for, which trains the system to deliver even more low-value clicks. The solution is simple: start with the core core audience, then let Grok optimize on top of that. When the system is trained on the right inputs, advertisers get real results. We’ve spent the past 12 years understanding these users, and we know this will work.

Daniel Vassallo@dvassallo

I don't have a ton of experience with online ads, but I have some. And unfortunately X is just terrible. Recently I had $8K in X ad credits and used them up targeting my followers. 600K impressions, 6.5K clicks, and only 3 conversions. Despite exclusively targeting my own followers, nobody I know saw the ad, and the comments under the ad were all from random people who didn't follow me or follow anyone I know. At the same time, I spent $700 on a different platform, running the same ad, and made 8 sales from just 55 clicks! And the cherry on the cake is that when you run an ad on X, you get mass blocked. Since ads are basically just promoted posts, and since the targeting mechanism is completely broken, your ads get shown to a lot of random people and they express their frustration by blocking you. So your account gets permanently penalized for trying to send money to this app.

English

111

Duco Janssen@ducoluno·20 Oca

"Every user on X has a level of authority within certain topics, regions, and niches. However, the current model doesn't have a direct "Authority Index" to look at. It treats all semantic relevance as equal. It optimizes for the Probability of a Click rather than the Value of the Information."

Thomas Slabbers@Thomasslabbers

x.com/i/article/1808…

English

204

Duco Janssen retweetledi

Robin@xdNiBoR·20 Oca

@nikitabier @joshelman Hey Nikita, have you seen this? We've build this website for the Grokathon in London that showcases our ability to instantly create a relevant timeline from 1 simple topic. Could it be helpful? x.com/i/status/20126…

Robin@xdNiBoR

This is exactly what we built today at the xAI Grokathon! From just one prompt and our own Authority Index, we can create a fully custom timeline. Importantly, this is NOT a filter on the X timeline. These are posts from users on our Authority Index.

English

2.6K

Duco Janssen@ducoluno·20 Oca

@fentasyl What do you think of something like this, where you can control your timeline (created during the Grokathon). isxforme.com

English

864