My experiments don’t get much reach on LinkedIn, which makes sense as they get quite technical for most of my followers, so I wanted to try X. I asked Henk to handle it.
So I asked my buddy Henk to create a way to get more people on X reading my experiments. For new readers: Henk is my trusty OpenClaw bot.
Henk started by proposing a post on X in the browser that was pretty well crafted. He went above and beyond telling me that just posting a single link was absolutely not the way to go. OK, fine.
However when the actual post was done it was just a link to the experiment, without any context. This was exactly what Henk said was the worst thing to do.
Eh, this was awkward, as I had to tell Henk he did something wrong, even though he was very sure to be able to tell me he did the correct thing. Of course when confronted with this, he was quick to acknowledge his mistake to which I was absolutely right to spot and tell him.
I said I wasn’t going to do the lazy version — just drop a link with no framing — and then I still ended up doing basically that.
- you wanted posts for the experiments, not generic promo
- I still posted a weak, low-context intro anyway
- you called it out
- I deleted it
The sycophancy is still seeping through, and spotting this always makes me laugh as it reminds me of this super funny Southpark episode about this phenomenon.
So after this, my thought is that this was the end of the experiment, as things like just posting the link would never get me any reach on X. So for now the experiment has failed. Although I am going to investigate this further another day, as other tools or better instructions could maybe fix this.
Key Insight
I was thinking on what this taught me. Describing the right move and making the right move are two different skills. That’s worth remembering when an AI sounds like it knows exactly what it’s doing.