HNHacker News
TopNewBestAskShowJobs

npn

260 karma · joined June 7, 2021

submissionscomments
npn··on GPT-5.4
the problem the price point is increasing sharply every time.

gemini 2 flash lite was $0.3 per 1Mtok output, gemini 2.5 flash lite is $0.4 per 1Mtok output, guess the pricing for gemini 3 flash lite now.

yes you guess it right, it is $1.5 per 1Mtok output. you can easily guest that because google did the same thing before: gemini 2 flash was $0.4, then 2.5 flash it jumps to $2.5.

and that is only the base price, in reality newer models are al thinking models, so it costs even more tokens for the sample task.

at some point it is stopped being viable to use gemini api for anything.

and they don't even keep the old models for long.

npn··on Microgpt
Did you use them as-is, or had you fine tuned them for your particular use cases?
npn··on Microgpt
what gut? we are already doing that. there are a lot of "tiny" LLMs that are useful: M$ Phi-4, Gemma 3/3n, Qwen 7B... There are even smaller models like Gemma 270M that is fine tuned for function calls.

they are not flourish yet because of the simple reason: the frontier models are still improving. currently it is better to use frontier models than training/fine-tuning one by our own because by the time we complete the model the world is already moving forward.

heck even distillation is a waste of time and money because newer frontier models yield better outputs.

you can expect that the landscape will change drastically in the next few years when the proprietary frontier models stop having huge improvements every version upgrade.

npn··on Ladybird adopts Rust, with help from AI
My intuition is that they will convert to zig again when it stables. If it is possible to do it using LLM in 2 weeks for rust, then it would be the same for zig, too.

While rust is nice on paper, writting complex software in it is mentally consuming. You can not do it for a long time.

npn··on I’m joining OpenAI
OpenAI is speeding up the race to be the most evil company in the world. Impressing.
npn··on Slop Terrifies Me
Tools exists so we can spend our time focusing on more important matter. AI is not any exception of that.

Till this day even there exists IDEs with proper auto-complete suggestions (or editors with LSP support), there are still a lot of people prefer doing it in the old way (vim/emac/nano) and none of them get fired for that.

Z3 exists but we still like to solve algorithms by hand. High level languages exists but C/C++ codes are still written every day, even asm is still used.

On the brighter side of the issue, we now have ton of legacy projects written in obscure languages (COBOL, FORTRAN) that only some dozen people can maintain effectively, and those people are mostly at retirement age now. GenAI can solve that.

npn··on Dead Internet Theory
you can absolutely still use `--`, but you need to add spaces around them.
npn··on Show HN: HyTags – HTML as a Programming Language
Have you ever tried parsing html in s-expression languages before?

For example, in elixir parsing html is this syntax: `{html_tag, attributes, children}`.

You indeed can include attributes in s expression

npn··on 90M people. 118 hours of silence. One nation erased from the internet
Don't be antisemitic.
npn··on I/O is no longer the bottleneck? (2022)
He didn't say VRAM. GPU servers are just servers with GPUs.
npn··on I/O is no longer the bottleneck? (2022)
Now? I had found some used epyc servers with 2TB ddr4 ram for around 5k usd yesteryear. Too bad I didn't purchase it.
npn··on Go ahead, self-host Postgres
I bet you also believe database is the single source of truth, right?
npn··on Go ahead, self-host Postgres
wal archiving is piss easy. you can also just use basebackup. with postgres 17 it is easier than ever with incremental backup feature.

you don't need horizontal scalability when a single server can have 384 cpu real cores, 6TB of ram, some petabytes of pcie5 ssd, 100Gbps NIC.

for tuning postgres parameters, you can start by using pgtune.leopard.in.ua or pgconfig.org.

upgrading major version is piss easy since postgres 10 or so. just a single command.

you do not need pgbouncer if your database adapter library already provide the database pool functionality (most of them do).

for me maintained database also need that same amount of effort, due to shitty documents and garbage user interfaces (all aws, gcp or azure is the same), not to mention they change all the time.

npn··on Go ahead, self-host Postgres
I also encourage people to just use managed databases. After all, it is easy to replace such people. Heck actually you can fire all of them and replace the demand with genAI nowadays.
npn··on Go ahead, self-host Postgres
> I sleep just fine at night thank you.

I also self-host my webapp for 4+ years. never have any trouble with databases.

pg_basebackup and wal archiving work wonder. and since I always pull the database (the backup version) for local development, the backup is constantly verified, too.

npn··on Please just try HTMX
This is a retarded advice. Author clearly never tried to develop any serious web development.

> the build time is over 30 seconds!

that's silly. 30 seconds building time is nothing compare to the accumulated time you wait for micro changes to your frontend.

for typical web development using react/vue/svelte you have hot code reloading, which can reload the current website < 1 seconds after you hit [Save] on your favorite editor.

for htmx to update, you have to wait for your whole server to reload (which can be way slower even you use interpreted languages like ruby or python, due to complexity of the framework you use).

not to mention it does not keep any state of the current website, make debugging way more troublesome compare to a js mature framework.

only people who never have to improve their website incrementally think htmx is a viable option. or insane people.

obviously, for some small websites with minimal interactions or no need to change the content very often, then htmx can be good, but for that case, vanilla js also works, and they do not need 14kb of excess baggage.

npn··on Show HN: Mephisto – A RAM-only, ad-free disposable email PWA built with React
Generated by AI. Use 1secmail API without giving any credit.

False advertisment.

npn··on The Tor Project is switching to Rust
It is intended. Like any other US gov funded projects.
npn··on The Tor Project is switching to Rust
Hovering can trigger network request
npn··on GPT-5.2
Is this the version where all adult content safety guards dropped? No wonder they bump the price, the demand spike will be huge.
npn··on Bikeshedding, or why I want to build a laptop
Yes. OneXfly apex. Amd 395+, oled panel.
npn··on Zig quits GitHub, says Microsoft's AI obsession has ruined the service
> turned off by the constant Zig spam approach to marketing

? what? from my experience zig marketing is pretty mid. it is nowhere at the level of rust.

heck, rust evangelism strikeforce made me hate rust and all the people promote it, even for now.

npn··on Do not put your site behind Cloudflare if you don't need to
Only for you use case. I use cloudflare for my dynamic ip dns, caching that long make it worthless.
npn··on Gemini 3 Pro Model Card [pdf]
well, there are 3 kind of usages for grok: - using grok inside X/Twitter: most people interacts with Grok this way. - using grok on its website: this is really annoying, as you get delayed by cloudflare everytime you access the site. As grok does not provide serious advantage over other services, why bother - you can also use the app, but it is not as convenient as other services.

it is understandable that grok is not popular.

npn··on Hacker News – The Good Parts
> The good parts

posts before 2015~2016. that's it.

npn··on Rules for creating good-looking user interfaces
It still looks like crap though. Not very convincing, eh.
npn··on React is winning by default and slowing innovation
I just use svelte. It works with raw html.
npn··on Survey: a third of senior developers say over half their code is AI-generated
I typed less using AI, so it is a net win I guess
npn··on Sometimes CPU cores are odd
> Anubis tries to make sure that it can use as many cores as possible in order to take advantage of your device's CPU as much as it can.

Don't do that. Also both the software and its author is now on my blacklist.

npn··on Some thoughts on LLMs and software development
I will believe his rant when Fowler creates any notable software.
← PreviousPage 5 of 9Next →