Hacker News Reader

simlevesqueon A Fast 64-Bit Date Algorithm (30–40% faster by counting dates backwards)

Simply, ClickHouse only works on a 399 years span while OP's algorith parses any date, over 3 trillion years.

simlevesqueon Cloudflare outage should not have happened

Yeah but the anti-DDOS feature needs to react to new methods all the time, it's not a static thing you build once and it works forever.

An insulin pump is very different. Your human body, insulin, and physics aren't changing any time soon.

simlevesqueon 'The French people want to save us': help pours in for glassmaker Duralex

It's also because almost every french home has one of their glasses and they are good quality.

So it's a mix of history, tradition, enjoying good products, the pride you get from your country producing good products.

simlevesqueon Gemini 3

Plan mode is more secure.

simlevesqueon Gemini 3

Ai studio is a web chat.

Antigravity is an IDE you install.

Colab is a place to run notebooks in the cloud.

simlevesqueon GPT-5.1: A smarter, more conversational ChatGPT

It seems like the line between sycophantic and bullying is very thin.

simlevesqueon Steam Machine

I bet they decided to crash their skin market in part because too many people were exploiting the Steam Deck loophole to take the skin money out of the system.

Now people will need to give Steam real money to buy their new devices.

simlevesqueon Yt-dlp: External JavaScript runtime now required for full YouTube support

> it has to work out of the box after a pip install.

What do you mean by it has to ?

simlevesqueon TTS still sucks

Context helps guessing what the next word will be.

simlevesqueon Asus Ascent GX10

He didn't give that info but the transcript linked at the end shows how much time was spent for each query.

simlevesqueon Asus Ascent GX10

Yeah that's a bummer. They do the same for all their boards like the Jetson Nano.

simlevesqueon Asus Ascent GX10

Cuda is king

simlevesqueon Asus Ascent GX10

Simon Willison seems to like his:https://til.simonwillison.net/llms/codex-spark-gpt-oss

simlevesqueon Asus Ascent GX10

gpt-oss:120b

https://til.simonwillison.net/llms/codex-spark-gpt-oss

simlevesqueon Asus Ascent GX10

I really wish I had the kind of money to try my hands at it.

simlevesqueon Btop: A better modern alternative of htop with a gamified interface

Lots of quality of life changes. Maybe some of them can be done in htop but with btop it's right in your face:

- CPU usage graph (global and per cpu)

- CPU temp + graph

- GPU usage

- Memory graphs

- Disks space infos + real-time IO

- Network graph, one for each network device, upload/download stats since opening the app

- Per program cpu graph, memory graph

You should try it.

simlevesqueon Helion: A high-level DSL for performant and portable ML kernels

Oh wow that's horrible.

simlevesqueon Show HN: I scraped 3B Goodreads reviews to train a better recommendation model

The How it works it way too short :) I'd love to see some scripts, know the hardware you use, etc...

By the way you could use Summa FTS Wasm + Duckdb Wasm to have the same website without any backend except file hosting. Maybe even just Duckdb Wasm with it's FTS would be enough. Summa FTS is very similar to meilisearch in essence because they're both derived from Tantivy.

https://izihawa.github.io/summa/quick-start/

simlevesqueon Why we migrated from Python to Node.js

> That IMHO is the weakest link in the TS backend ecosystem, compiler plugins to convert TS types to runtime types are really needed.

Just define all your types as TypeBox schemas and infer the schema from that validator. This way you write it once, it's synced and there's no need for a compiler plugin.

https://github.com/sinclairzx81/typebox?tab=readme-ov-file#u...

simlevesqueon Why we migrated from Python to Node.js

Yeah I did my first project with FastAPI earlier this year with experience in other languages and I couldn't believe how bad it was.

simlevesqueon Nvidia DGX Spark: When benchmark numbers meet production reality

CUDA

simlevesqueon Counter-Strike's player economy is in a freefall

I sold my old skins CS from 2013 last month. I bought a Pixel 10 Pro with the proceeds and I still have some cash in my wallet.

So this article is very funny to me too.

edit: I did try to warn you all: https://news.ycombinator.com/item?id=44943123

simlevesqueon ROG Xbox Ally runs better on Linux than Windows it ships with

Playing any PvP games without them isn't fun anymore.

simlevesqueon HP SitePrint

    ## How it works
    ### CAD preparation
    1. You need a 2D CAD file. If you have a 3D model, convert it into a 2D .dxf CAD file.
    2. Insert additional printing information and instructions and use the HP Plug-in to get a robot-ready file.
    3. Save the 2D .dxf CAD file into the cloud. Maintain version control and share revisions with field operators.
    ### Site Preparations
    1. Clear layout area as for manual layout. No need for a broom-swept floor.
    2. Make sure that control points used for Robotic Total Station setup are accurately marked. The layout is as accurate as the control points.
    ### Solution setup
    1. Set up the Robotic Total Station and shoot the control points. 
    2. Lock the Robotic Total Station to the robot tracking prism.
    3. Connect the Robotic Total Station wirelessly to HP SitePrint through the user interface (tablet, phone, laptop, etc.).
    ### Job execution
    1. Open the CAD file on the control panel, select the print area, and submit the job.
    2. Maintain a line of sight between the robot tracking prism and the Robotic Total Station. 
    3. Choose between different inks for different types of layouts.
    4. The robot avoids collisions with obstacles. 
    5. HP SitePrint is robust enough to work on rough and bumpy surfaces.
    ### End-to-End management
    1. HP SitePrint cloud allows sharing of the latest CAD files with all stakeholders, so you can monitor job progress from the office and manage accounting reports.

simlevesqueon DeepSeek OCR

> That will cost you around $0.20 for 1000 pages in batch mode and the results have been great.

Can you explain more about your setup ? I have a quarter million pages I want to OCR.

simlevesqueon DeepSeek OCR

https://cloud.google.com/document-ai

simlevesqueon Show HN: A Digital Twin of my coffee roaster that runs in the browser

I'd like to see you tackle the problem of figuring the opposite, let the user draw the curves and generate the inputs. Then I could replay with Artisan.

Cool website !

simlevesqueon An MVCC-like columnar table on S3 with constant-time deletes

Are you familiar with Parquet ? you can't do that at all, you need to rewrite the whole file.

simlevesqueon I switched from Htmx to Datastar

It's one of the worst blog post I've ever read.

They kind of have a point but everything around it is ridiculous.

simlevesqueon An MVCC-like columnar table on S3 with constant-time deletes

I just want to be able to append metadata to a Parquet file at the end without rewriting the whole file. Tombstones could be baked in the parquet file this way.

It does work with "one more file" but it's not good for performance.