Hacker News

Article URL: https://github.com/Joaoha/Bonsai

Comments URL: https://news.ycombinator.com/item?id=49043023

Points: 2

# Comments: 0

So i am a newly graduated person and in need of some cash so id be delighted if yall gave me a commision for something realistic, id be highly glad for anything

Comments URL: https://news.ycombinator.com/item?id=49043020

Points: 1

# Comments: 0

How do I approach making this legible? Is there a “magical AI” solution these days? I know the basics of image processing and filtering but I can barely read my own handwriting so I suspect I’d need help from AI or someone even if I managed to isolate the signal.

P.S. I hope I’m not doxxing my great great grandfather or anything.

https://ibb.co/4gC2X1fX

https://ibb.co/7J1jf4dV

Comments URL: https://news.ycombinator.com/item?id=49043001

Points: 2

# Comments: 1

Article URL: https://www.damninteresting.com/project-babylon-gerald-bulls-downfall/

Comments URL: https://news.ycombinator.com/item?id=49042999

Points: 2

# Comments: 0

I run a UK ltd, ecommerce, personalised products. Started in March, scaled to ~£150K/m in 2 months.

But then... "After conducting a routine review of your Stripe account... we've found that it presents a high level of risk for disputes." Payments switched off immediately. Payouts held for 120 days, balance released in November.

My dispute rate is 0.18% across about 3,900 payments. Their own threshold is 0.75%. We're on 4.7 stars on Trustpilot from 130+ reviews.

After a few hours before the shutdown email, a Stripe salesperson emailed me a repricing form because our volume was growing and they wanted to discuss custom rates.

I appealed tonight and sent fulfilment records matching every order to a tracked delivery, plus bank statements. The dashboard said the review would take until 27 July. It got denied about 90 minutes later. Account is now closed permanently.

Has anyone dealt with this?

Comments URL: https://news.ycombinator.com/item?id=49042989

Points: 1

# Comments: 0

Article URL: https://www.youtube.com/watch?v=cMnjphv3DQo

Comments URL: https://news.ycombinator.com/item?id=49042941

Points: 1

# Comments: 0

Article URL: https://shownotesgenerator.com/

Comments URL: https://news.ycombinator.com/item?id=49042934

Points: 1

# Comments: 0

Article URL: https://onewayinterview.com/

Comments URL: https://news.ycombinator.com/item?id=49042922

Points: 1

# Comments: 0

Article URL: https://terradev.cloud/tessera

Comments URL: https://news.ycombinator.com/item?id=49042917

Points: 1

# Comments: 0

Article URL: https://www.microsoft.com/en-us/corporate-responsibility/topics/open-weight/

Comments URL: https://news.ycombinator.com/item?id=49042884

Points: 3

# Comments: 0

Article URL: https://www.gr.inc/releases/introducing-backsearch

Comments URL: https://news.ycombinator.com/item?id=49042879

Points: 1

# Comments: 0

Article URL: https://www.tandfonline.com/doi/full/10.1080/15298868.2025.2593298#abstract

Comments URL: https://news.ycombinator.com/item?id=49042875

Points: 2

# Comments: 0

I used Comic Chat in 1996 and never got over it. So when Microsoft put the source up last week, I wanted it back the way I remembered it: same engine, browser tab, no install, works on a phone.

I do not write or read code. I'm the proverbial "product manager." So, I have to thank both GPT 5.6 and Fable 5 / Opus 4.8 (just before they released Opus 5 today). I aimed to stay faithful to the original as much as I could and extend it intelligibly.

It's not IRC, but you can connect with up to 10 people in a secure (E2E encrypted), ephemeral group chat. Sorry, but I'm not sure who uses IRC anymore. I don't. ;)

One link, up to ten friends, everyone typing into the same strip live. Names, lines and uploaded backdrops are encrypted in your browser with AES-GCM through WebCrypto before any of it leaves. The relay is a Cloudflare Durable Object that orders events and passes them along, and it cannot read a line of your comic. A room carries separate keys for the script and for uploaded images, plus a third derived with HKDF that proves you belong without the real key going on the wire. Keys are non-extractable and ride in the URL fragment, which a browser never sends to a server. Security was not an afterthought with this.

It is one HTML file. Nothing is emulated, there is no WASM, and the constants are reproduced from the source by hand rather than compiled. There is a "For nerds" tab in the app with the routines, the geometry and the crypto behind the shared rooms, etc., so I will let the page answer that rather than paste it here.

Comic Sans is kinda optional. Barely.

I'm down for constructive feedback and viable suggestions / bug reports.

Enjoy!

Comments URL: https://news.ycombinator.com/item?id=49042872

Points: 1

# Comments: 0

Article URL: https://www.cnbc.com/2026/07/23/amazon-makes-sellers-label-ai-generated-people-in-images-after-ny-law.html

Comments URL: https://news.ycombinator.com/item?id=49042870

Points: 4

# Comments: 0

My point is about structuring and indexing an arbitrary dataset, for preprocessing before eg segmentation/clustering, neighbor querys...

In 1D, sorting is trivial because scalar values have a total, unambiguous order. In 2D and higher dimensions, there is no single way to order points, and points can't all be compared >=/<= in a consistant way. Space-filling curves like Hilbert or Z-order curves still can order Nd points by mapping them to a 1D index, but flattening the space sacrifices much of the original geometric structure.

Recently, I’ve been thinking about partial orders for d-dimensional vectors. For vectors u and v, let denote u >=_k v if u_k >= v_k (comparing along the k-th coordinate axis only).

In 1D, a sorted array satisfies x[i+1] >= x[i]. Extending this naturally to a multi-indexed d-dimensional array x[i, j, k, ...], we can define a dataset as "cartesian sorted" if:

- x[i+1, j, k, ...] >=_1 x[i, j, k, ...] - x[i, j+1, k, ...] >=_2 x[i, j, k, ...] - ...and so on.

In other words, the array of multi indexes is sorted along every row, column, and slice under its corresponding coordinate projection.

What makes this definition interesting is that their exist a very simple algorithm to perform Nd sort in the "cartesian" sense. Just iterate: 1. Perform 1D sorts along every row (apply the permutation i' = sigma_jk(i) on the multi indexes that make all the rows sorted in the >=^1 sense. 2. Perform 1D sorts along every column. 3. Repeat across all dimensions until convergence.

While sorting along one axis can disrupt the order along another, the process is guaranteed to converge fast to a fully "cartesian sorted" state (I have a short proof if anyone is curious).

---

Questions for the HN Community:

1. Is there existing literature or a standard name for this approach? Searching for multi-dimensional sorting mostly yields spatial indexing, radix, or binning approaches, but maybe I missed something obvious ?

2. What are the fundamental limitations that make this approach less common in practice than space-filling curves or k-d trees, if they are some in your opinion ? I thought it would be commonly used already, but it doesn't seems to be a standard tool.

3. Has anyone experimented with similar (or maybe better) multi-index coordinate sorting techniques? I’d love to hear your thoughts or critiques on the concept.

Comments URL: https://news.ycombinator.com/item?id=49042861

Points: 1

# Comments: 0

Article URL: https://www.loweringthebar.net/2026/07/florida-bans-oddly-posed-alligators.html

Comments URL: https://news.ycombinator.com/item?id=49042860

Points: 1

# Comments: 0

Article URL: https://www.dimmd.app/

Comments URL: https://news.ycombinator.com/item?id=49042824

Points: 1

# Comments: 0

Article URL: https://www.outsideonline.com/culture/active-families/tell-your-kids-truth/

Comments URL: https://news.ycombinator.com/item?id=49042811

Points: 1

# Comments: 0

You write a standard solution, just like on LeetCode, and run it through the CLI. It identifies the problem by ID or title, executes your code against local test cases, and shows the result.

It currently supports around 1000 problems and multiple languages, including Python, C++, Rust, Java, Go, TypeScript, Swift, and others.

The project is still an MVP. System design, SQL, and concurrency problems are not supported yet, but support for more problem types is planned.

Made in Haskell!

Comments URL: https://news.ycombinator.com/item?id=49042788

Points: 1

# Comments: 0

Article URL: https://signfile.io/

Comments URL: https://news.ycombinator.com/item?id=49042782

Points: 2

# Comments: 0