ImaCake

Its a bit like how we have gps connected maps in our pockets. Anyone born after 2000 has only a loose concept of being lost (within range of a *G tower and >1% battery) and treats paper maps as a hobby. Our children will grow up assuming their phone can do maths and programming better than human ever could.

>In five years, we'll all have phones that connect physical locations again, instead of (or as well as) people. And we'll be raising a generation of kids who don't know what it is to be lost, to not know where you are and how to get to some desired destination from wherever that is. >Think about that. "Being lost" has been part of the human experience ever since our hominid ancestors were knuckle-walking around the plains of Africa. And we're going to lose it — at least, we're going to make it as unusual an experience as finding yourself out in public without your underpants.

https://www.antipope.org/charlie/blog-static/2007/05/shaping...

Is this sarcastic or authentic cope? I don't understand how people can still be in denial about LLMs this late in the game.

>It just takes too much discipline

While I think your argument has merit, I would be careful to discount any human skill based on people being too lazy to do it.

>If you have a specific question, reading LLM text can be more interesting than anything in the world.

I've noticed this as well. When I get a bot to generate something I am happy to read it (although I often skim read it). But god forbid I have to read someone's LLM generated report! Come to think of it, if someone marks the text as LLM generated, especially after they have plonked a hand typed summary or sentence at the top, I am quite happy to read it. Maybe my beef is being expected to read LLM text the same way I would read high effort human prose? I certainly read LLM text differently to human text - you know the first sentence is going to glaze you and the second paragraph will just repeat the first paragraph so you can skim those annoyances.

For search I am perfectly happy with 3.6 Flash, its fast and servicable, and maybe being too smart is almost a problem for search and research. Mostly the edge google has here is that websites blocked for claude or chatGPT (if it respects the block) are unblocked for gemini.

This is true for literally anything you read though? It wouldn't be worth reading anyone's blogs or books if they all had the same experience. Maybe the guy who helped invent Electron and Claude Code has something interesting to say.

Yes, which is why I am happy to state it having not read that deeply into the evidence either way. I would argue that stating "Inequity between male and female athletes is a result not of inherent biological differences" is a very strong statement that the scientific consensus has many counter-factuals available to disprove.

His website is generally fantastic: https://www.gregegan.net/

My personal favourite on there is Oceanic. I really should get around to reading some of his books, but they are pretty intense bedtime reading!

I went and read it and it doesn't seem to matter to me that it's about ultra-endurance events? Of course differences in stride length and endurance physiology will matter for run performance.

The full paragraph:

>Inequity between male and female athletes is a result not of inherent biological differences between the sexes but of biases in how they are treated in sports. As an example, some endurance-running events allow the use of professional runners called pacesetters to help competitors perform their best. Men are not permitted to act as pacesetters in many women's events because of the belief that they will make the women "artificially faster," as though women were not actually doing the running themselves.

To me both cultural and physiological differences are likely in play. I am not even sure how you could effectively seperate culture from physiology here (hard to do a randomised controlled trial on stride length - I guess Mendelian Randomisation regression might work?). Anyone confidently asserting only one or neither of these is at play is probably not stating that from a position of logic.

I think we underestimate how much skill we put into our prompts to build these personal apps or pretty much any of these few-shot vibe coded projects. Not that a complete noob couldn't do this, but they would get there much much faster if they had heard of Kotlin and downloaded an .apk from github. This isn't meant to dispute the Eunice claim or even the magic of a bot writing better programs with little more than an idea and an hour. Prompting lowers the barrier to entry, but those of us with experience are still having an easier time of it than the beginners.

>which is good ideas

... and politics! To me LLMs are just another piece of evidence that it was never the lack of resources that made Microsoft produce bad user experiences.

Does work pay for your subscription? I think many workplaces would worry about personal accounts for work projects.

EDIT: but thanks for a useful answer to OP's question. As a programming dabbler I also find it hard to imagine how to push much past the 5hr limit on $20USD subscription.

Maybe that's true, I don't know. I make no claims of an unbiased sample. Merely a counterfactual to the parent comment's claim that they aren't useful.

I know a few very serious and well compensated programmers with decades of experience. These people work on prominent open source projects, climate modelling, linear programming, large org enginneering. Every single one of them uses LLMs heavily and swears by the increased productivity of using them.

Yeah its quite the contrast to the recent QGIS core "no AI" policy that very pointedly had no emotive language at all.

They REALLY don't like AI, to the point of "wish I didn't need to rely on GCC and GRUB because they have AI in them" levels of hate AI. See this further down the linked page:

>The fork in question, of U-Boot, is heavily AI vibe-coded and thus useless to Libreboot. This is not mere conjecture; in said aforementioned U-Boot mailing list thread, the author stated thus. We will not, in any way, encourage or facilitate use of AI. It has been made clear that several projects Libreboot uses also have some AI-assisted commits; GCC, GRUB, etc. This is not our fault, but it is my intention that I and others will audit these projects in the future, de-slopping them and possibly even forking them entirely for that very reason.

We have a small child so eating at the table is almost a default for some meals. But we went from basically never eating together at home to deliberately sitting and talking while eating for breakfast and dinner. There's no rules around not having devices available and we don't force the conversation. I think being relaxed about it has helped make the practice stick. Mostly its just semi-guaranteed time to talk and do something together every day.

> Before we come to a final conclusion on this

Science can be patient but those of us who eat should act now. Something is making us fat, and it sure looks like its the UPFs.

> is that UPFs aren’t inherently bad

My reading of Ultraprocessed People was more that the author was very careful to only pin on UPF's what could be proved to be true. The between the lines read certainly felt like "you should probably not eat UPFs".

I would humbly add that any of Michael Pollan's books, especially "In Defense of Food" are a good answer to how we should eat if we can't eat McDonald's. I have certainly found restarting the dining table tradition to have incalculable benefits for my family.

Yes, but that could double the cost by default! Which is probably why they don't. On reflection, coding bots do implement something like this with the verification tic they seem to have.

Thanks for this wonderful reply. I don't have anything to add but did want to show I appreciate it!

This is why I started a diary. It's hard to understand who you are, what your interests are, what you were thinking about in the moment and how that changes over time. You have to some corpus of work that you can look back on to see that.

A lot of us aren't actually using enough tokens to reach the value add of a $20USD/month subscription versus just paying for the API. Without even mentioning the chinese competitors, you can get a lot of tokens for $10 of ChatGPT Luna, which is more than capable of solving the kinds of problems I have at home (where I use hermes harness).

Plan mode does have different permission controls though? Its difficult for me to parse the docs on this but it does seem to imply more than just a prompt although less than a full sandbox. https://code.claude.com/docs/en/permission-modes#analyze-bef...

Depends on the harness too. The MS copilot 365 chat interface happens to parse PDF and docx files before handing them to the model. Absolutely fantastic if your PDF is 1300 pages of concatenated reports and you want to find a single detail in it that you can't easily ctrl+f for.

Plenty of code is not publically available, and yet that has not stopped them from RLHF'ing good coding bots. You can surely see the parallels to any other field including finance.

My vague understanding of the context window limitation is that it is largely a constraint of the model architecture. So maybe they have special extra long ctx, but it might just be a hard limit of the model itself.

Yes. If they could prove where the de-identified data came from then it wouldn't be de-identified. There's a whole field of statistics dedicated to this problem and often applied to things like national census data.

I am absolutely struggling to sell my workplace (which is entirely knowledge work) on the usefulness of LLMs for proofreading let alone on automation of hairy parts of our workflows. So yeah, even people who should be able to see what is coming are not looking.

Maybe they didn't ask or the anthropic people said no? Plenty of other explanations here.