LLMs need to be understood via your System 1, not your System 2.
LLMs need to be understood via your System 1, not your System 2. LLMs are inherently complex and hard to reason about. The only way to be able to use them...
Topic: llms
1,598 mentions · 615 chunks · 117 episodes
169.8× distinctiveness vs baseline
?
How much more common this term is here than in ordinary English. Higher values mean the topic is more characteristic of this corpus.
Topic summary
?
A short read on the topic's time range, peak episode, and strongest associations. Use it as the quick orientation before drilling into examples.
- llms appears in 615 chunks across 117 episodes, from 2023-11-06 to 2026-04-20.
- Its densest episode is Bits and Bobs 2/2/26 (2026-02-02), with 15 observations on this topic.
- Semantically it travels with ChatGPT, Claude, and prompt injection attack, while by chunk count it sits between Claude; its yearly rank moved from #3 in 2023 to #1 in 2026.
Over time
?
Raw mentions over time. Use this to see absolute attention, not relative rank among all topics.
Observations
?
The primary evidence view for this topic. Sort it chronologically when you want concrete examples behind the larger pattern.
Someone with more knowhow in a context will be like a wizard.
...ho are able to work at higher layers of abstraction will be more effective with LLMs.
People who know how to use LLMs are wizards.
People who know how to use LLMs are wizards. They can use LLMs extremely effectively, to do things that look like magic. Using an LLM effectively is a form of knowhow. Knowhow doesn...
Guillermo Rauch points out that with LLMs sometimes it's faster to create software than to Google it.
Guillermo Rauch points out that with LLMs sometimes it's faster to create software than to Google it. What a time to be alive!
The easiest way to "finetune" a model is to curate the context.
The easiest way to "finetune" a model is to curate the context. LLMs with a well-curated context will be significantly better at producing outputs that are useful. The tools that create the most value with LLMs will ha...
LLMs are not the thing; they create the conditions for the thing.
LLMs are not the thing; they create the conditions for the thing. They disrupt the cost structures, enabling the creation of a new thing. Back before the ...
Apparently a famous musician had an interesting tactic for developing new ideas.
...nly possible because he was so rich. But now anyone savvy enough can do it with LLMs!
LLM output is kind of like the tyranny of marginal information.
...e in the entire corpus of humanity you get the middle of the distribution. Will LLMs help people find new ideas / vibes? Or do LLMs just drive to the average of everything? The LLM is un-opinionated and bland. But where you drive it i...
One of the best use cases for LLMs: they can compile English to code.
One of the best use cases for LLMs: they can compile English to code. The implications of this are significant. English is at a much higher pace layer than traditional code, which allo...
Normal LLM models are like the brain's System 1.
... problems that the right expert could give a gut answer on and be right. Normal LLMs are great at these. Some problems are inherently multi-ply problems. Even experts in the field would have to sit down to think it through This is whe...
o1 is not like other LLMs; it's a different type of thing.
o1 is not like other LLMs; it's a different type of thing. It's more akin to comparing whole systems/scaffolding (e.g. Github Copilot Workspaces) around a model than comparing...
Just because LLMs can talk doesn't mean your app should have them talk.
Just because LLMs can talk doesn't mean your app should have them talk. A lot of UX patterns today using LLMs imply working with a human. When there's an implied human...
The only way to learn how to use LLMs is to get hands-on experience with them.
The only way to learn how to use LLMs is to get hands-on experience with them. The open-endedness is intimidating! But the more you use LLMs, the more of a feel for them you get, the bett...
Most software for specific needs is very simple.
...lish to software in the small. Small niche software is very low complexity, and LLMs can do a very good job with it.
Spontaneous software is clearly a high potential use for LLMs.
Spontaneous software is clearly a high potential use for LLMs. There are lots of amazing systems that have sprouted up for making spontaneous software. But all of those existing systems so far are about creating...
LLMs help patchwork thinkers more than highly ordered thinkers.
LLMs help patchwork thinkers more than highly ordered thinkers. The former jump between things with connections, and LLMs help do that for quickly. The la...
A famous Murray Gell-Mann quote: "Imagine how hard physics would be if atoms could think."
...in other systems, less in control of the output, and look down on them. But now LLMs can "think". Suddenly, corralling them becomes harder than writing code!
It's amazing how much competition there is for LLMs.
It's amazing how much competition there is for LLMs. Everyone standardized on OpenAI's API, which allows fast switching between providers. Whichever knowledgebase users store their data in for the LLM ...
If you try to fit LLMs into your existing workflow, it will fail.
If you try to fit LLMs into your existing workflow, it will fail. But people who are willing to change their workflow can realize its power. Programmers can change their ow...
Without the internet it would be impossible to have LLMs.
Without the internet it would be impossible to have LLMs. The "assembly condition" of LLMs is language --> books -> internet -> LLMs. LLMs could not exist without any of those steps in that chain.