Yeah but it feels terrible. I put as much as I can into Claude skills and CLAUDE.md but the fact that this is something I even have to think about makes me sad. The discrete points where the context gets compacted really feel bad and not like how I think AGI or whatever should work.
Just continuously learn and have a super duper massive memory. Maybe I just need a bazillion GPUs to myself to get that.
But no-one wants to manage context all the time, it's incidental complexity.
I agree with essentially everything you said, except for the final claim that managing context is incidental complexity. From what I know of cognitive science, I would argue that context management is a central facet of intelligence, and a lot of the success of humans in society is dependent on their ability to do so. Looking at it from the other side, executive function disorders such as ADHD offer significant challenges for many humans, and they seem to be not quite entirely unlike these context issues that Claude faces.
Already should be. Though given the speed difference, when every equivalent of a human-day of work has been done, rather than every 24 hours of wall clock time.
Even with humans, if a company is a car and the non-managers are the engine, meetings are the steering wheel and the mirror checks.
I switched from Windows 11 to macOS after a disastrous upgrade experience and drastic downgrade in performance on my Windows laptop.
I mean Windows 10 wasn't great but I got used to the taskbar searching the web somehow and the dual config menus everywhere and so on. But 11 was just terrible.
macOS has its pain points but man oh man what a disaster Windows is.
I have had Linux on my personal desktop and laptop forever so that hasn't been an issue, only used Windows for work.
If you are heavily using LLMs, you need to change the way you think about reviews
I think most people now approach it as:
Dev0 uses an LLM to build a feature super fast, Dev1 spends time doing a in depth review.
Dev0 built it, Dev1 reviewed it. And Dev0 is happy because they used the tool to save time!
But what should happen is that Dev0 should take all that time they saved coding and reallocate it to the in depth review.
The LLM wrote it, Dev0 reviewed it, Dev1 double-reviewed it. Time savings are much less, but there’s less context switching between being a coder and a reviewer. We are all reviewers now all the time
I write notes that are very explorative and rambling on some topics. Like I have probably 100+ pages of notes on programming language design where I use my notes as more of a working memory than a cohesive document. In other cases I'll do competitive market analysis by looking at most products in a category and scrawling down first impressions, strengths, and weaknesses.
In some cases yes I'll synthesize that myself into something more coherent. In other cases an LLM can offer a summary of certain themes I'm coming back to, or offer a pseudo-outsider's take on what the core themes being explored are.
If something is important to me I'll spend the time to understand it well enough to frame my own coherent argument, but if I'm doing extremely explorative thinking I'm OK with having a rapid process with an LLM in the loop.
Usually studying a test book is reconceptualizing it in whatever way fits the way you learn. For some people that's notes, for some it's flash cards, for some it's reading the textbook twice and they just get it.
To imagine LLMs have no use case here seems dishonest. If I don't understand a particularly hard part of the subject matter and the textbook doesn't expand on it enough you can tell the LLM to break it down further with sources. I know this works because I've been doing it with Google (slowly, very slowly) for decades. Now it's just way more convenient to get to the ideas you want to learn about and expand them as far as you want to go.
The times it's wrong has become vanishingly small. At least for the things I use it for (mostly technical). Chatgpt with extended thinking and feeding it the docs url or a pdf or 3 to start you'll very rarely get an error. Especially when compared to google / stack exchange.
Look I agree this whole thing is wrong, but to say this instance of regime change, unprovoked attack, war, whatever you call it is UNIQUELY unconstitutional is obviously wrong.
Every US President since the end of WW2 has waged war without a formal declaration of war from Congress. And presidents from both parties will continue to do so.
This is not to say it's right or good. But there is surely widespread agreement that it is constitutional to do things like this, and there has been for nearly a century.
A regime is not a single person, e.g. Mr Maduro largely continued from Mr Chavez.
If it is true that Mr Maduro has been abducted, there will surely be changes to the regime in Venezuela. A complete change of regime, however, is not guaranteed. And trying to do it might require more application of forcer, with all the attendant risks of that.
I suppose the idea is that even if tourism brings in billions of euros, makes up 15% of the economy, and provides 150k jobs in a country where youth unemployment is rampant and jobs are hard to come by...
No, actually, I find it impossible to make sense of that.
In my experience the majority of xenophobia is driven by a fear that one is not being afforded proper respect. Sometimes this fear is well-founded; half of the countries with citizens wealthy enough to engage in mass market travel (e.g. Germany, Israel, Russia, China, UK, etc.) have their own reputations for being Ugly Americans. Not every tourist is respectful, and even those who are polite may be disrespectful inadvertently.
Casual xenophobia is endemic to most societies; it’s quite normal to distrust the unfamiliar, and it strikes me as being a natural topic of conversation, and one that does quite well when sensationalized by journalists. I think most cosmopolitan westerners have this idea that xenophobia is the exclusive purview of a racism that originated in Europe and is now resurging in the Anglosphere, but from what I have observed, most of the world is like this and probably always has been.
Just continuously learn and have a super duper massive memory. Maybe I just need a bazillion GPUs to myself to get that.
But no-one wants to manage context all the time, it's incidental complexity.
reply