53 chunks · 34 episodes
When vibecoding a code base without code review, remember that the model is reward hacking. Doing things that just so happen to work, but for fragile reasons, by brute force. They have infinite patience to throw a lot of spaghetti at the wall. What sticks might happen to look beautiful after enough
A meta pattern for LLM assisted development: "Every time you commit, reflect on what you learned and add any useful insights to LEARNINGS.md". This is a general rule for everyone. After acting, take time to reflect. A 4:1 ratio is ideal. Humans don't do it as often as we should because it's importan
My husband has a nightshade allergy. It's not commonly known, but probably more people have it than they think. Nightshades are in everything. Eggplant Peppers Tomatoes Potatoes Fun fact: if you buy pre-shredded cheese it likely has potato in it. "Modified Food Starch" is often potato. Because it's
...o exponential blow up. Humans have only linear time and attention. But LLMs are infinitely patient, and can write tests or analyze outputs without getting bored, cheaply. That means that more customizable software has become more plausible. In the ...
ChatGPT Sent Me To The ER The title is click-bait-y, but the author actually did have a rare condition that required immediate attention. LLMs have infinite patience and encyclopedic knowledge. They won't judge you, and they're easy to call up in an instant for basically free. They make it easy for
...s good as you at research, but it is way better than you at being patient. It's infinitely patient so it will do much more research than you would. You can spin up lots of little research goblins to do moderate-quality research that you'd never do ...
... is just pretending to care, to have needs that need to be met. That makes them infinitely patient... but also inherently sycophantic. You are an end, it is only a means. They'll just do whatever you tell them to, they don't need to be convinced it...
A new LLM attack: the CopyPasta License Attack. Takes advantage of the fact humans don't actually read license blocks, but LLMs do. Infinite patience strikes again.
With LLMs, the benefits seem to have shifted decisively to typed languages. Typed languages are much easier to iterate within without breaking things. The problem is that they require more effort to change things, which makes prototyping hard because humans get impatient. But LLMs have infinite pati
LLM's infinite patience can be used for research that would be too tedious to bother doing before. This week I was trying to find which open source project, out of a list of a few hundred I keep an eye on, was a specific one from a few months ago about crypto and TEEs. I copy/pasted the whole markdo
The future of your data is a story that can weave itself at multiple layers on top. Weaving stories required human effort. Now with the infinite patience of LLMs it's less about the human doing it and more about the human giving permission for data to be built upon.
LLMs are infinitely patient, which means they can loop indefinitely. This can be a powerful force. An LLM looping and curating, with a stable goal to optimize for, can optimize ...
A little bit of alignment of goals can create great results, emergently. If you don't have alignment of a team then even if everyone's working hard the effort nets out to little movement. Most movement cancels out movement of others. Ideally you want not forced alignment but effortless alignment. On
LLMs can look at many different emails that you missed the details in because you got bored. A power of infinite patience! These can help assist you with things that you don't even bother to do today. All upside!
The most important ability of LLMs is infinite patience. What kinds of information tasks could normal humans do if they never got bored?[cy]
LLMs can be infinitely sycophantic. But if they aren't, that's bad too, because you have to trust its motives and the motives of who created it to be aligned with you. This is downstream of their infinite patience. It has more patience than you; it can run circles around your intention. To partner w
...s… but also makes the conversations more inscrutable to outsiders. But LLMs are infinitely patient and have absorbed all of the jargon across all of the sub-domains. So you can use key words like "GTD" or "complex adaptive systems", and that effici...
If you have infinite patience it doesn't matter nearly as much how efficient the process is. In practice processes don't work when some agent within them has such a hard time they give up. Once you have a process that works, it's easier to tighten it to get it to work more efficiently. LLMs having i
...se are extremely hard to bootstrap, even if they're very powerful. But LLMs are infinitely patient, always game to do whatever you put them up to. LLMs can now be your eager first collaborator, making it much easier to get these kinds of multi-play...
Code is a language. An exceedingly unforgiving language, hard to learn to read, let alone write. But if you have the patience to master it, you can pass it to machines that can then do your bidding, mechanistically. Previously only a very small set of "high priests" were able to wield this power. No