epihelix

This is hypothesised future deceleration, I'm guessing? Because we've seen the exact opposite of deceleration from closed models over the last seven months.

Not quite what you're suggesting, but at least Nextcloud (and other open self-hosting platform) servers are a thing.

All you need is for the app to provide the option of saving to local storage – your cloud server software provides a local storage endpoint and does the rest.

Right now, that site doesn't show this post, regardless of whether the filter is active or not ...

So, it's impossible to know whether your filter is working on this story yet, either.

What about me ...

While I can see that you feel very passionately about this, the reality is that it's the majority experience that will dictate adoption.

There may never be published research on productivity -- blinding in this instance is impossible, so I don't know how you'd ever do fully-controlled behavioural studies that carried any weight. It doesn't matter. If enough of us decide that LLMs are useful to us, then this form of coding will become the norm.

If that ends up causing more harm than good, then eventually there will be a course correction. But for now, for enough people that matter, LLMs are at least giving the perception of productivity increases. And our decisions and choices come down to the perception of reality, not reality itself (for better or worse).

So I think it's far more useful to take a pragmatic approach, as per TFA. Accept that LLMs have issues, but also bring advantages, and that LLM use in coding is here to stay. What we can do is remain aware of the bad, and make better use of the good.

As for you, personally ... if you mentally cannot deal with LLM output, then I think you have two choices. You can either learn to author system prompts, so that LLM output better fits your needs and no longer triggers you; or you can sit more and more on the outer, raging against the machine while the world changes around you.

Eventually, you'll be like a master craftsman in an era of mass-production. But that's potentially highly valuable in niche markets (consider a watchmaker working in Glashutte, for example), so you may yet win from this. Remember that every day, LLMs are making your own coding skills and knowledge more elite and therefore lucrative, sit back, and smile.

I briefly used Zip drives back in the day. They were amazing, when the only alternative portable media was a 3.5" floppy.

And then USB thumb drives came along, and made them obsolete pretty much overnight.

Gods, I can't believe his Nintendo married him. I mean, what a player!

I didn't know that people used to make buggy whips for free and give them away?

The thing about FOSS that everyone keeps forgetting is that it sits outside the market economy.

We write FOSS code because we love doing it, and love giving it away. If nobody else uses my code, I don't care - I'll still write it and still release it. That's the way it's always been, and LLMs will not change this (quite the contrary - they make it easier to realise and maintain FOSS projects).

that could be: give away solar panels and two way heat pumps to low-income families, tax the profits of oil and gas companies, enforce right to repair, remove tariffs on chinese car makers if they create jobs in europe, fund train networks to get tickets cheaper than flying, invest in renewable energy to make electricity cheap. theres a lot of options. all of them will lead to some inflation and public debt but there are easy ways to offload that to the rich.

The problem is that most of these (and potentially all of these) ideas sound good, but achieve very little. Some (e.g. right to repair, sadly) will achieve almost nothing at all. This would be a scatter-gun approach at best; at worst, it creates the impression of doing something while actually doing very little.

I don't know what the best solution is. Arguably a genuine carbon tax (one that taxes large producers for the previously-externalised costs of greenhouse gas production, while providing tax cuts to the general population) is the single most effect means of enacting real change. However, Australia demonstrated that even a scheme that results in real, substantial income tax cuts while generating measurable, real reductions in emissions is vulnerable to a scare campaign by a committed, ultra-right attack dog politician.

I strongly suspect that our only potential way out of this now is via geoengineering, and that's fraught with dangers. I am also worried that a lack of government interest in climate change will eventually induce a highly wealthy individual to undertake geoengineering privately, without oversight, scientific input or regulation. But I cannot see the world stopping this any other way, other than by reversing the effects of climate change via nuclear winter.

At least with Zotero Version 6, you could have just selected all your PDFs and drag-n-dropped them into the interface.

Like another poster here, trusting an LLM with my reference database -- the ultimate source of truth -- is not a step I'd be willing to take. All it takes is a single hallucinated reference, and your career would be forever tainted. It's not worth the risk.

Sadly, Zotero seems to have removed this killer import feature in later versions, which is the reason I keep using version 6. It feels like later versions have been a route to dumbing-down the interface, prioritising simplicity (and an ultra-low-contrast interface) at the expense of functionality. (If you can still drag-n-drop PDFs straight in with the new versions, someone please let me know?)

Unlike grepping your BibTeX file, Zotero provides a nice tabbed PDF viewer, useful automatic import options with file renaming, and saved dynamic searches. Those are the main reasons why I still use it -- it's a good tool to manage my BibTeX file via a nice GUI, and faster than manually grepping BibTeX from the CLI.

Also, as my unenlightened colleagues insist on using ridiculous tools like MS Word to author papers in, it's useful to have Zotero installed for the ability to link a BibTeX file to a WYSIWYG word processor when collaborating. Zotero helps you play well with others.

I don't think Zotero is perfect by any stretch, but I'm glad that at least it allowed academia to reject both Endnote and Mendeley. Anything that punches Elsevier in the face gets a tick in my book.

I suspect "AI 1844" would be more apt for this group.

There are a lot of HN users for whom this song (and its covers) was an integral part of our childhood. The Nikki French cover was one of the very first MP3s I ripped. The song is just part of growing up with the early personal computers and teaching ourselves how to code. Its kinda cannon.

If you're younger and you missed out on this, then I'm sorry. The older I get, the more I realise that the best of all the years really have gone by.

Use shorter prompts: In internal evaluations, replacing long, explicit system prompts with minimal prompts improved scores by roughly 10–15%, while reducing total tokens by 41–66% and cost by 33–67%.

When has this ever not been the case? I don't think this is a GPT 5.6 specialty!

Although it seems youd have to pay a lot to get an accurate one because I have a $250 mechanical Seiko and its time keeping is junk. It was mediocre when I got it and has gotten worse. It was $150 when I bought it so I suppose it would have been a good investment if it hadn't got beat up.

You know you need to service mechanical watches regularly, right?

A 7S26 movement (Seiko's mass-produced budget workhorse) isn't that accurate (I think -35 to +45s per day IIRC?). But if you paid $250 secondhand you most likely have a 6R15 or similar inside, which should keep between -15s to +25s per day at worst if regularly serviced. Often you can get much better performance from these movements than the specs imply.

But ... you need to service that poor thing. For a 6R15, every 5 years at minimum, but as an old watchmaker I knew used to say -- a watch will tell you if it needs servicing earlier. Sounds like yours has been trying to get your attention for some time :)

(Otherwise, it's like complaining that the Porsche you haven't taken to a mechanic in the last decade doesn't drive so well any more ...)

You will never get quartz accuracy from any mechanical watch, but that's hardly the point.

(The ETA 2824-2 movement in the page you linked to -- the movement that powers most mid-range mechanical watches -- is substantially more accurate than these lower-range Seiko movements, although it's more costly as well.)

I'm currently using >7 year old Android phone. The batteries on these things are child's play to replace (especially after you've done it the first time and removed the pointless adhesive in the battery compartment). I will consider upgrading once new devices return to feature parity with my current device (apparently never).

While I'm of course an edge case, the fact that Google, Apple and Samsung all provide >5 years OS support for devices now (and battery replacement services) suggests that many people hang onto their old phones for a long time.

That said, I'm not using 7 year old software on my phone. That would be insane. And my browser (fennec) was updated just a few days ago.

As a molecular biologist / genomics PI, these examples read fine to me. It's meant as an introduction, not a compendium of every possible nuance.

I had a software engineer do a bioinformatics PhD in my group several years ago, and I wish this had existed to show him. It would have saved all of us a lot of time, and him a lot of angst and confusion.

I used to think it would be the other way around, but having supervised grad students from both sides of the divide, I can say with some certainty that it's easier (in general) to teach molecular biology majors coding, than it is to teach software engineers cloning.

(There are exceptions of course, both ways.)

The reality is that to succeed in this field now, you need both sets of skills.

I'd be extremely surprised if either gender or the propensity to believe in religion are 100% nature or 100% nurture - anything related to the mind is a mix of both.

OsmAnd~ is great for hiking, when you need topographic contour info, 3D relief shading, etc.

CoMaps is a much simpler and faster interface, that I use for driving navigation. (I use a 7 year old S10e, though, so it's possible that on a modern phone you'd not notice the difference?)

YMMV, but they're both free, so it's easy to try and see for yourself :)

In research science, it's very normal to require zero aids at an interview, and has been for some time. No calculators, no laptops, no phones -- just you.

Yes, you'll also give a seminar with slides to present your prior work, but the whole point of the chalktalk is that it's you, and you alone, presenting your future plans. You're grilled by the faculty on your ideas, and you have to defend yourself without any props or crutches.

Try intimidating the LLM as it writes your code.

That Claude Code regex for curse words? Turns out Anthropic was just looking for the next coding Bobby Fisher ...

If poor Green Boots was staring up at the camera, empty eye sockets and all, I'd understand this. But there is nothing distressing in the image; the out-of-focus fuzziness makes the photo seem if anything substantially more macabre than it actually is.

Given Green Boots' fame, it was interesting to see what the actual scene climbers experience was. So I think the inclusion of the photo was justified, in these unusual circumstances.

In this modern world where highly effective birth control is cheap and straightforward, we really need to stop equating fertility rates with levels of sexual activity. You can have plenty of sex and not have a child; you can have very little sex and have a substantial number of children.

It's fascinating to me how personal choice never seems to enter into these discussions, even in relatively highly educated, first-world democracies. I actively chose to not have kids -- it was not an accidental by-product of iphones or any other proposed environmental factor.

This is not my field, but are we sure that reality is not quantised at some level?

Infinite is a very big claim.

How would you know? You have no external frame of reference; a virtual world of sensory signals would be identical from your perspective. (I agree that "reality" is the most parsimonious explanation by far, btw, but that's never been the point of the simulation thought experiment.)

I think the more interesting corollary of this article is that if we're living in a simulation, it's an impossibly, improbably detailed one. I really want some compute time on the HPC that's running it.

Thank you :) All wombats are in some trouble right now (even the bare-nosed or "common" wombat), but the Northern Hairy-nosed is right on the edge of extinction.

Wombats never get much attention, so it's awesome to see this article and the response it's got.

The thing that really scared me about the landing site for Claude Science was this promotional image of the software in action:

https://cdn.prod.website-files.com/6889473510b50328dbb70ae6/...

Very depressing. MDPI journals will be saturated with these slop papers (if they're not already). It shocks me that Anthropic thought that this was a good thing, and says a lot about their research integrity (or lack thereof).

Science has no idea yet how such disclosures should work yet.

Technically, most journals have a policy that LLM use should be acknowledged, but I agree we're still very much in the weeds about this right now. Much firmer guidelines should have been established years ago.

(I also have no issues with LLM usage in research either, btw -- I use LLMs to fact-check / proofread / discuss / sanity-check my conceptual work, to background myself in other research, and to refactor and assist with analytical coding. They can be a game-changer for medical research, when used rationally and sensibly.)

I don't know about you, but I think with words all the time when reasoning through complex ideas. Where do you think the phrase "thinking out loud" comes from?

Have you ever reasoned through a complex coding or maths problem? Tell me, how did you do that without words?

You (and TFA) are making a false dichotomy here. Yes, of course we think in images. But we also have an inner monologue that is critically important for much of our higher-level thought. How do you even write a HN comment without thinking through it in words?

Should we also have a referendum on reducing the tax rate to zero? It would also be an extremely important decision, directly affecting everybody's life. If asking people about anything, then isn't that what it should be?

Not all referenda that might win a "yes" vote are sensible to propose.

I sat one exam at uni (in person, invigilated) for History and Philosophy of Science, which had no time-limit. You could take as long as you needed, whether that was one hour, five hours or all day.

Pretty much everyone still took three hours. But you felt so much calmer, knowing you weren't racing against the clock.