It triggered for me on a completely pedestrian game design prompt a couple of days ago. I’ve sent feedback and continued with Opus, but that was really unexpected
This supports my feelings of what works best when working with the models. Especially together with the directional bias in recall that’s mentioned here in the comments.
First, the model attention is actually limited, so less rules is usually better, but that’s common knowledge already. Or maybe it’s as common as common sense, and a lot of people still employ lots of rules and try to cram everything in one step.
Second, it’s often quite sufficient to just namedrop a technique and LLM will work differently. For example, when debugging, LLMs tend to try to brute force the problem and often end up in the weeds. Just add “use scientific method for debugging and keep journal file” is usually sufficient to improve their skill here.
Another example is refactoring. Just add “use Mikado method”, and it’s sufficient to wholly change the approach and produce much better results.
Yeah, probably different halves of the book. I've read it a few years ago, and it looks like I've just ignored/forgot all that "bending reality" stuff. It's not my cup of tea, so I remembered only the good parts of it
I find all current LLMs to have pretty poor spatial awareness. It is becoming better, but still very poor. How are you dealing with that? Got any special tricks, any advice?
Touching shoulders and shaking hands palm-down is “persuasion” bullshit.
Can’t say anything about other books, but Never Split the Difference is about hearing and understanding other peoples’s wants and opinions. Not this pickup-artist-like bullshit.
I’ve genuinely read only Never Split the Difference from your list, and it’s kind of the opposite from manipulation.
The book teaches how to actually hear people even in the very emotionally charged situations, how to properly ask them questions to understand their point of view and their needs.
If I understand my son’s needs and can give him what he wants in exchange of him giving me what I want, how is that a manipulation? I can yell at him, impose sanctions (eg no minecraft for two days) and we both will be greatly dissatisfied. Or we can both get what we want, which is a win-win.
The post matches my experience as well, I am asking a question like “does A work like this and that”, and Claude responds with “you’re conflating A and B! Only A does this and that, and B does that other thing!”
Well, I am perfectly aware of B and that other thing and did not conflate them at all. I also achieved enlightment, so I don’t argue with Claude here, just ignore the obnoxiousness and move on.
Ah, okay, my bad. Got too focused on the name. Googled the dates, Satya became CEO in 2014 and Sundar became CEO in 2015, so it’s actually not that different, especially when we look at the events more than a decade later.
Sundar Pichai does not work in Microsoft, though. A bit weird to anchor the MS timeline on his position. When he became the CEO, actually? I don’t remember the year even approximately
Can this be a Siri-like core? Set me a timer, tell me what’s the weather, etc. Here is transcribed text and available list of tools for the model to call, and voice the output.
The last one is about involving less people. You don't have to read it as "shut up and keep your thoughts for yourself". I read it more like "Do we really need to have six people working on this feature/present in this call?"
What I meant is that only sometimes I am faster than Claude with debugging. When it's a standalone problem, a report in Sentry, and I just know immediately where I need to go to fix it. Then it's faster to do myself, than telling Claude what's the problem and where to look and wait.
Bugs happen during feature development, as you say, but then Claude is in the context, and I don't need to tell it where to go, it sees the bug with failing tests, or smth similar.
BTW. One thing that helps my Claude with debugging harder problems is that I tell it to apply scientific method to debugging. Generate hypotheses, gather pros/cons evidence, write to a journal file debug-<problem>.md, design minimal experiments to debunk hypotheses.
You can add that as a skill, and sometimes it will pick it up automatically, but it works wonders just as a single sentence in the input.