Sorry for the tangent, but as a chart nerd it is so interesting to me the way that LLMs produce charts for articles like this - it's like they have no concept of a 3rd party (the reader). So the writing and the vis gets overloaded with the particular context of the conversation, even if it would be irrelevant to a reader.
Like the x-axis label of the first plot mentions that it shows gridlines every 8 ticks - I don't think that's a choice I've ever seen a person make. The next plot does the same thing, but in the title: "(dotted lines: evenly spaced reference)". Again, would anyone making a chart ever include such a specific detail in the title? Probably not beyond undergrad level.
I encounter a similar thing when trying to work with LLMs to write articles - they cannot help including information or detail from the conversation, rather than empathising with the reader and filtering out what they might already know or not care about. They may be capable of solving decades-old maths problems, but on this particular axis they seem to be floundering at the level of a fresh graduate that's desperate to talk about what they've done rather than what their audience needs to know.
This happens with everything LLMs write - code comments, readmes, user facing UI text, etc. It just can't help but include every irrelevant detail from the context.
It's definitely gotten worse over time as well. Looking at older README files I generated using Opus/Sonnet 4.5, its actually quite shocking how clear and concise they are. Whereas eg, Opus 5 will write paragraphs and paragraphs about random things I prompted, or a partial solution that was then scrapped for something else.
My theory is that the increasing strength in performing long-context tasks has led to this particular quirk.
> Like the x-axis label of the first plot mentions that it shows gridlines every 8 ticks - I don't think that's a choice I've ever seen a person make
The spacing being 8 is important and seems not so bad to call this out in an extra place.
Imagine LLMs didn’t exist for a moment. Is this the part of the article you’d find most intriguing? That the grid lines being an unusual distance apart is written both on the axis and in the description? Would it have seemed inhuman on a blog post about debugging a weird display issue on an e-reader?
> They may be capable of solving decades-old maths problems, but on this particular axis they seem to be floundering at the level of a fresh graduate that's desperate to talk about what they've done rather than what their audience needs to know
Have you ever worked with people beyond graduate level? Ever seen scientific posters at a conference?
I got an X3 a couple of weeks ago, they are dirt cheap and the form factor is amazing. It's so easy to leave it in the pocket and just pull it out when you have an idle moment somewhere out in public. I wasn't certain how comfortable reading would be on such a small screen but it works well. With crosspoint, you can sync your page position with koreader on a larger form factor device.
those LookUp Tables are the most difficult thing to get from a display manufacturer; never thought to let an AI tune one itself using image feedback, incredible!
Yeah...its taken a TON of hand tuning (with AI of course) to test all the LUTs on 15 different devices across b&w, grays, anti-aliasing grays, etc. Huge pain. Finally in a good spot after this recent set of PRs got merged in. I had some collapsed grays that I never noticed before that OP here helped me identify were missing.
I assume these fixes did not go yet in the latest release (1.6.0) and are going to be part of a later release? Looking forward to it, specially since anti-aliasing improvements were mentioned.
Been really enjoying the X3 lately. This is a great piece and ideally how a good blog post written about an experience with AI should go – authentically written, and not AI generated. I had fun reading it.
I’m interested in your debugging of this device. I’m not interested in what your llm did in the process. Just mention at some point you used one; "Claude led me on this dead end" is not interesting content.
Okay ill give you points because I never heard of Xteink X3or the open-source CrossPoint. I have a Kobo reader/pen note taker and a Kobo ereader (somewhere) one is too large and one is just meh. Im going to try one of these but it would have been easier to pick something if you actually said HOW DOES IT WORK OUT OF THE BOX AS AN ACTUAL EREADER!
The reader functionality was not the point of TFA. If you want a product review, there are plenty, including the prior HN discussion linked in this thread.
That said, I read enough in the HN thread to get the sense of not bothering with “out of the box” and just flash Crosspoint Reader on it. I played with stock for a few minutes, it was functional, but not notable. Crosspoint took ten minutes to flash, and is head-and-shoulders above the stock FW. Better reader functionality, better font rendering.
I recently purchased an Xteink 4 Pro. My understanding is the built-in software is somewhat ok. You can find reviews on YouTube if you want to know more.
For installing CrossPoint, that is very easy. The CrossPoint website walks you through the installation, which only takes 5-10 minutes. After that, you can put content on the device using the build-in SD card, via Wifi, and a few other methods.
reviews I've seen have been very positive, with the caveat that "out of the box" includes reimaging it with open source firmware. make sure you get a device that can be reimaged this way.
Sorry for the tangent, but as a chart nerd it is so interesting to me the way that LLMs produce charts for articles like this - it's like they have no concept of a 3rd party (the reader). So the writing and the vis gets overloaded with the particular context of the conversation, even if it would be irrelevant to a reader.
Like the x-axis label of the first plot mentions that it shows gridlines every 8 ticks - I don't think that's a choice I've ever seen a person make. The next plot does the same thing, but in the title: "(dotted lines: evenly spaced reference)". Again, would anyone making a chart ever include such a specific detail in the title? Probably not beyond undergrad level.
I encounter a similar thing when trying to work with LLMs to write articles - they cannot help including information or detail from the conversation, rather than empathising with the reader and filtering out what they might already know or not care about. They may be capable of solving decades-old maths problems, but on this particular axis they seem to be floundering at the level of a fresh graduate that's desperate to talk about what they've done rather than what their audience needs to know.
This happens with everything LLMs write - code comments, readmes, user facing UI text, etc. It just can't help but include every irrelevant detail from the context.
It's definitely gotten worse over time as well. Looking at older README files I generated using Opus/Sonnet 4.5, its actually quite shocking how clear and concise they are. Whereas eg, Opus 5 will write paragraphs and paragraphs about random things I prompted, or a partial solution that was then scrapped for something else.
My theory is that the increasing strength in performing long-context tasks has led to this particular quirk.
Perfectly stated and also explains why it slips past the author (they just nod and let it slip pass).
This is so tiring.
> Like the x-axis label of the first plot mentions that it shows gridlines every 8 ticks - I don't think that's a choice I've ever seen a person make
The spacing being 8 is important and seems not so bad to call this out in an extra place.
Imagine LLMs didn’t exist for a moment. Is this the part of the article you’d find most intriguing? That the grid lines being an unusual distance apart is written both on the axis and in the description? Would it have seemed inhuman on a blog post about debugging a weird display issue on an e-reader?
> They may be capable of solving decades-old maths problems, but on this particular axis they seem to be floundering at the level of a fresh graduate that's desperate to talk about what they've done rather than what their audience needs to know
Have you ever worked with people beyond graduate level? Ever seen scientific posters at a conference?
I got an X3 a couple of weeks ago, they are dirt cheap and the form factor is amazing. It's so easy to leave it in the pocket and just pull it out when you have an idle moment somewhere out in public. I wasn't certain how comfortable reading would be on such a small screen but it works well. With crosspoint, you can sync your page position with koreader on a larger form factor device.
I'm not convinced on the price. It's about 75 euro delivered for me.
It's about half the price of a kobo Clara bw, which has much better specs, higher ppi, waterproof etc.
I am tempted though.
Related to the Xteink X3 reader mentioned in TFA:
https://news.ycombinator.com/item?id=49596629
those LookUp Tables are the most difficult thing to get from a display manufacturer; never thought to let an AI tune one itself using image feedback, incredible!
Yeah...its taken a TON of hand tuning (with AI of course) to test all the LUTs on 15 different devices across b&w, grays, anti-aliasing grays, etc. Huge pain. Finally in a good spot after this recent set of PRs got merged in. I had some collapsed grays that I never noticed before that OP here helped me identify were missing.
I believe someone built a Linux print driver that let you print straight to these and it would upload the file.
This one
https://news.ycombinator.com/item?id=49617255
I assume these fixes did not go yet in the latest release (1.6.0) and are going to be part of a later release? Looking forward to it, specially since anti-aliasing improvements were mentioned.
Will be in the upcoming 1.6.5 which will probably release this week. RC is out now
Correct.
Been really enjoying the X3 lately. This is a great piece and ideally how a good blog post written about an experience with AI should go – authentically written, and not AI generated. I had fun reading it.
wow, an e-reader with opensource firmware? been living under my kindle rock for a while, this sounds exciting.
Join us!
If you’re interested in this, you should check out the Modos project.
I’m interested in your debugging of this device. I’m not interested in what your llm did in the process. Just mention at some point you used one; "Claude led me on this dead end" is not interesting content.
Okay ill give you points because I never heard of Xteink X3or the open-source CrossPoint. I have a Kobo reader/pen note taker and a Kobo ereader (somewhere) one is too large and one is just meh. Im going to try one of these but it would have been easier to pick something if you actually said HOW DOES IT WORK OUT OF THE BOX AS AN ACTUAL EREADER!
The reader functionality was not the point of TFA. If you want a product review, there are plenty, including the prior HN discussion linked in this thread.
That said, I read enough in the HN thread to get the sense of not bothering with “out of the box” and just flash Crosspoint Reader on it. I played with stock for a few minutes, it was functional, but not notable. Crosspoint took ten minutes to flash, and is head-and-shoulders above the stock FW. Better reader functionality, better font rendering.
I recently purchased an Xteink 4 Pro. My understanding is the built-in software is somewhat ok. You can find reviews on YouTube if you want to know more.
For installing CrossPoint, that is very easy. The CrossPoint website walks you through the installation, which only takes 5-10 minutes. After that, you can put content on the device using the build-in SD card, via Wifi, and a few other methods.
reviews I've seen have been very positive, with the caveat that "out of the box" includes reimaging it with open source firmware. make sure you get a device that can be reimaged this way.
The author never intended their article as a review. So why not look up "xteink x3 review"?
You're on HN ffs...