written-beyond

My life changed after I got tested for vit D and started talking supplements. I was severely deficient. I am now sufficient and everything changed for me.

100/10 for mobile usability. Panning, Zooming, selecting and moving was so seamless I thought I was tripping out.

You can get really good results with Chinese models. You're putting Opus and GPT on too high of a pedestal.

Claude code is good, Opus as general model is a hot hallucinating piece of garbage. I asked it to review a single page 50 rows, 2 column excel file. It hallucinated almost everything in the file. It repeated that for the next two files I asked it to review, these were tiny files, barely 20 kb.

The funny part is Opus was the one which generated the files in the first place. This was Opus 4.7 High. So no thank you, Anthropic.

I thought these TPUs were primarily used for inference?

Just interesting to see how SO is evolving.

Do we really need to wonder? it's probably cloudflare.

I considered going to hetzner at one point but I read a lot of stories around hetzner that didn't inspire confidence. Primarily that they're not really that much cheaper than going to other companies offering something similar.

If some people can chime in with their positive experiences I might switch.

I shudder when I read DRM after building a embedded display raspberry pi build. There was a problem with Raspberry pi kernel incorrectly detected an HDMI display was connected even though I was trying to display to DSI. The only work around I had was to write a kernel module that would disable HDMI as a possible interface.

The base fee doesn't need to be monthly, they can take $50 dollars as a one time registration verification fee. That should be enough to cover their compute costs for the year, specially when you pool in that from multiple customers. Who is spamming with 100 monthly emails? How much compute do they need to verify you aren't spamming. They can bake all of that into the pay per use price, they choose not to and I'm glad cloudflare is offering this.

The problem is how bullshit transactional emails are when you're outside of AWS. If you're not expecting to use 10,000 emails a month but would like the option to go over the free tier without committing to 10,000 more. Just let me pay per use FFS.

I agree, software (software startups) has always been the golden child of investors because of how cheap it is compared to hardware or any other physical good.

Good software is expensive regardless of the involvement of LLMs because you need someone to take responsibility. Large companies will save a buck because there may be fewer people needed to take said responsibility, but it's probably a marginal saving compared to the overall scheme of things.

I literally said this exact thing verbatim with someone who was worshipping ai for code. This is an insane idea, really cool.

One bit of feedback, I'd love if this could be a little more aesthetic in the visualization. The method names are a little hard to read, white on light green.

Whenever I read John Deere my brain somehow adds Louis Rossman in there too.

The moon is tidally locked to earth, we only ever see one side.

Yes but it was capped by their restriction on sign ups/registrations. That could've easily been in the hundreds of millions if the app had public signups.

Idk if Instagram would exist if they were spending hundreds of millions a day.

My god it's the perfect app.

I've never said YES BUY immediately after reading a title before but this is it.

My phone frequently gets cluttered with tens of different note taking, calendar, Todo type apps because I am literally trying to solve this exact problem.

When I need to make a note I need something instant and now, but which lets me set an alarm/reminder with it too. I am a forgetful person, I find it extremely difficult to remember which notes app I am using because I usually drop them because they're never really solving my problem.

Thank you for this. I am going to install it and try it out over the week, if it's experience lives up to your description/screenshots then you have an immediate customer.

I mean a trivial solution to all of these work around a could have been each keystroke registers a single asterisk that goes away after a delay. You wouldn't reveal the length and you'd had a standard way of informing the user that their keystroke was registered.

I am aware of that but you forgot the other conditions. Keys sometimes don't register, I'm not sure why but I do experience missing keystrokes.

The passwords get updated irregularly with the org IAM so you aren't sure what the password even is. Pasting doesn't work reliably sometimes, if you're on windows you need to right click to paste in terminals, sometimes a shortcut works. Neither gives me any feedback as to what event was ever registered though.

Yeah but am I going to really open another ssh connection just to run an admin specific command. They also didn't provide an admin user, it setup with all of the extra security configurations. You couldn't even `su`

The number of times I've been stuck wondering if my keystrokes are registering properly for a sudo prompt over a high latency ssh connection.

These servers I had an account setup too were, from what I observed, partially linked with the authentication mechanism used by the VPN and IAM services. Like they'd have this mandatory password reset process and sometimes sudo was set to that new password, other times it was whatever was the old one. Couple that with the high latency connection and password authentication was horrible. You would never know if you mistyped something, or the password itself was incorrect or the password you pasted went through or got double pasted.

I think this is a great addition, but only if it leads to redhat adopting it which is what they were running on their VMs.

TBF my friends who were getting business degrees struggled with their 8 gigs pc. When they need to run something like SPSS next to a chrome instance, their ram got tight pretty fast.

You greatly underestimate the utility used, serviceable laptops have provided to broke students.

My first laptop was a decommissioned pos office dell ultrabook. By every metrics it would've been the worst option to choose, but since it had replaceable memory I was able to push it to 16 gigs and get through my computer science degree and many side projects. Computational speed was adequate for me, I ran Linux on it. It had an Intel U series 6th gen (12th gen was latest then) i5, an NVMe ssd and was always responsive.

If I were a student in this day, and all I could find were these laptops this is what I would think. 1 they're out of budget for most students in developing countries. 2 I will most likely out grow 8 GB ram faster than my laptops CPU performance. 3 I am limited to learning with what can run on apple silicon(most Linux distros excluding asahi). Finally I end up paying basically 50-60% of the cost of a decent machine and replaced it with a disposable one.

Maybe this machine is perfect for a specific set of users, students with higher income households or degrees which need better a better quality display.

I still advise every computing student I meet to get a under $200 old used laptop that has expandable memory and atleast an NVMe ssd. That way they can maximise their time learning and experimenting. Anything that needs more complex hardware can always be offloaded into your institutes machines. Once you're settled a bit and have a decent amount of cash to burn go ahead and buy whatever maxed out MacBook your heart desires.

But it really was a once in a lifetime experimental device. The aesthetics were a little whack, in an endearing sort of way.

Man I really loved MeeGo, I was in my early teens when I got the N9 specifically because it was such a beautiful experiment from Nokia. Amoled display, MeeGo os and the polycarbonate shell. The phone was hot garbage technically, would over heat had 0 apps and had idle battery times worse than a cup of tea left out on a cold day. However the phone was such a work of art I could accept just about any flaw.

Right after that I got a Blackberry Z10 and there's just something about the multitasking UI in both of these OSs' that just felt like it was the right way of doing it.

Blackberry OS 10 and MeeGo where so wonderful, I truly had a rich experience of mobile phone OSs' growing up.

I'm not sure about Jolla as much though. Like I enjoy having this additional option but I wished they digged deeper into features other than enhanced privacy. Not that I'm complaining, I enjoy having enhanced privacy but if they added more productivity features like the Blackberry Hub.

A developers job has always been reviewing and understanding code.

Code is literally always the last resort. Unless you're building solutions for other customer, most companies should attempt to minimise the amount of code they have. Because, and I repeat, it's a developers job to understand and review code. More code, more understanding needed, more reviews needed, more problems created.

I don't think we should actively stop people from discussing what's on their minds.

Every person has their own lived experiences, I think it should be common courtesy to at least give someone who puts in the effort into writing a, respectful non ai generated, comment a fair shot and being read.

It may be a coincidence but there has been an increase in thinly veiled humor in HN comments.

You've basically described Flutter and Jetpack compose(for desktop). The problem really does turn into effort to pay off, even if we stayed with JS and the rendering engine, figured out a way to compile JS into native code and completely stripping all of the unused functionality and doing the same thing with the rendering engine. All of that would need to be made, it's not like electron apps literally crash you machine. You have metrics, hundreds of millions of devices running electron apps at a daily basis. Unless you make your own company, I don't think anyone can convince their leadership to take such a decision.

Gemini 1.5 Announced the 1 million token context window in 2024. I admire this view of being forward looking towards new technologies, specially when we see the history of how bad people can be at predictions just by looking at history HN posts/comments.

If we look at back 2 years, companies weren't investing into training their LLMs so heavily on code. Any code they got their hands on was what was in the LLMs training corpus, it's well known that the most recent improvements in LLM productivity occurred after they spent millions on different labs to produce more coding datasets for them.

So while LLMs have gotten a lot better at not needing the entire codebase in context at once, because their weights are already so well tuned to development environments they can better infer and index things as needed. However, I fail to see how the context window limitation would no longer be an issue since it's a fundamental part of the real world. Would we get better and more efficient ways of splitting and indexing context windows? Surely. Will that reduce our fear of soiling our contexts with bad prompt response cycles? Probably not...