directionless update
June 28, 2026
While bogged down by the common cold, I couldn't help but notice that I could be doing better. I worked with Claude on this issue, and we came to the A loop in which repeated rewards lower the floor you return to, and from that lowered floor doing nothing starts to feel unaffordable, so the quickest hit answers before thought does..
I remember learning about this a few years ago, but it seems like I didn't apply it correctly at the time. This time around, it worked as advertised.
I managed to do hard things that I have been putting off for several months, of all on a Sunday, which I have recently been spending completely on highly dopaminergic activities previously.
However, this was only one day. I go back to work on Monday, and I will continue to apply it moving forward.
A separate learning is that whenever I reason with agents on self improvement, it's very important to learn the concepts as if you were the clinician. If the conversation is anywhere near working on you the individual/patient, the agents are still programmed to spit out the median answer immediately, they don't bother asking for more information, or context it needs, and does not work with you over the long term.
I wonder if there is a startup working on the above. There seems to be a large potential of value generation in helping people do better, better.
erm
It's a bit embarassing to say this, but somewhere along the way, I forgot that people actually have different value functions and personal preferences.
I think in my consistent solitude, the continous reasoning I applied for myself started leaking outwards, and I applied it unto others.
But that is only a theory, I do not know what I do not know.
It took two mistakes at work to realise something was not quite right, and that my ideal self would not have made those two mistakes.
I like to believe it was just because of my mental state at the time, paired with the erroneous line of thinking described three lines ago.
However, another learning was that I truly do not know what I do not know, and mistakes in my reasoning are par for the course.
In such, I cannot confidently and consistently say that I know where my mistake lies, nor what the proper remediation should be.
I think a great resource I should be leveraging more are my coworkers. They have the experience and context to give me the best-in-slot answer, insight separate from agents, unbiased from my inputs and thinking.
Thankfully, I have a 1:1 with my gracious manager/TL on Monday which was booked most likely in remedy of one of said issues, in which I hope to leave with a better understanding and learning of my errors!