bravura

Can you explain your use-case and why DuckDB is better?

Considering switching from pandas and want to understand what is my best bet. I am just processing feature vectors that are too large for memory, and need an initial simple JOIN to aggregate them.

The top line of the README says: "Embed an SQLite database in your PostgreSQL table. AKA multitenancy has been solved."

But I'm still having trouble trying to grok the intricacies of it. In a sense, I guess it has well isolated individual SQLite DBs and you'd have to go out of your way to join over them. With that said, does PostgreSQL manage and pool all the writes correctly? So you don't need to worry about SQLite concurrency issues?

I've never used esp32/arduino. How does it compare writing some low key Python on RPi and blitting an image to the display?

As someone who has struggled with getting CSS to do normal layout stuff that had clear precise semantics but required weird CSS trickery, it's actually more scary than lucky that stuff like container queries have arrived 30 years after CSS was introduced.

I agree with GP that CSS should be scrapped.

"The classical tradition of "accurate" painting (Raphael and Michelangelo and Rembrandt) is not exciting.

But we're not ready to go full on free jazz/postmodernist/de-constructionist. You're not ready for it yet, but your kids are going to love it."

I made an eink rpi display for staying in touch with my parents, inspired by the poetry clock (https://www.theverge.com/23669343/ai-clock-chatgpt-poems-rhy...).

My dad didn't like poetry clock, but he does like image gen. So we got a (color) Inky Impression 7.3 and hooked it up to an RPi.

I made a basic telegram bot that you could send a verbal prompt to ("snowy day"). It would then ask which of your favorite artist styles it should create an image in. I found that presenting a list of two styles combined had cooler results. The prompt would be used to fetch a random quote on the topic, and quote and style would then be feed to stable diffusion, and maybe 30 seconds later you have fresh art and a quote on the display.

My dad then asked if we just could forward images directly there. He prefers, each day, to post an image of whatever the day is (November 13 is "World Kindness Day") and occasionally share a family photo. My mom looks forward to seeing what day he picks every day.

I didn't know I wanted this until I read the Show HN description. Suggest to update the description on the site for people who don't know what Donkey Konga is.

TBH, I've really want a passthrough one-button 8-channel recorder for a long time. It's great for rehearsals of small groups.

One thing I've been looking for, and was a bit tricky implementing myself to be very fast, is this:

I have a particular max token length in mind, and I have a tokenizer like tiktoken. I have a string and I want to quickly find the maximum length truncation of the string that is <= target max token length.

Does chonkie handle this?

I know the reported scores of MT3 are very good, but have you had success with using it yourself?

https://replicate.com/turian/multi-task-music-transcription

I ported their colab to runtime so I could use it more easily.

The MIDI output is... puzzling?

I've tried feeding it even simple stems and found the output unusable for some tracks, i.e. the MIDI output and audio were not well aligned and there were timing issues. On other audio it seemed to work fine.

Regarding keeping the test set private to avoid contamination, the comments about leakage are spot on. The real test set should always be the future.

We should evaluate LLMs on text from beyond their knowledge cutoff date, by computing their per-byte perplexity or per-byte compression ratio. There's a deep theoretical connection between compression and learning.

The intuition here is that being able to predict the future of science (or any topic, really) is indicative of true understanding. Slightly more formally: When ICLR 2025 announces and publishes the accepted papers, Yoshua Bengio is less surprised/perplexed by what's new than a fresh PhD student. And Terence Tao is less surprised/perplexed by what will be proven in math in the next 10 years than a graduate student in a related field.

This work has it right: https://ar5iv.labs.arxiv.org/html//2402.00861

My advisor always used to say: "If the wisest people in your field say that something is possible, they are probably right. If they say that something is not possible, they may very well be wrong."

IANAL, but my understanding is that EU Digital Services Act (DSA) protects EU consumers against this lack of transparency for moderation/bans.

https://freedomhouse.org/article/eu-digital-services-act-win... "Providers of hosting services, including online platforms, now have an express legal obligation to provide clear and specific statements of reasons for their content moderation decisions. The DSA also empowers users to challenge such decisions through an out-of-court dispute settlement mechanism."

https://www.theverge.com/2024/2/16/24074933/european-union-d... "When it comes to content moderation, sites will have to provide a reason to users when their content or account has been moderated, and offer them a way of complaining and challenging the decision. There are also rules around giving users the ability to flag illegal goods and services found on a platform."

"Wear a tuxedo for your ID photos." - Glenn O'Brien (TV Party, Downtown 81, the Style Guy)

One theory is that "goofy-footed" came from the way Goofy surfed in an old Walt Disney cartoon: https://www.pacificlongboarder.com/news/Did-you-know-The-ter...

Pretty cool. So it greedily constructs an ensemble of trees from scratch. At each step, it takes the one decision that reduces the loss the most. That's a nice approximation for the most compact tree ensemble.

It would be amazing if you:

a) There were a test / eval suite to determine which model works best for what. It could be divided into a training suite and test suite. (Training tasks can be used for training, test tasks only for evaluation.) Possibly a combination of unit tests against known xpaths, and integration tests that are multi-step and end in a measurable result. I know the web is constantly changing, so I'm not 100% sure how this should work.

b) There were some sort of wiki, or perhaps another repo or discussion board, of community-generated prompt recipes for particular actions.

"Give a man a reputation as an early riser and he can sleep 'til noon." - Mark Twin

Do you mind sharing why you chose SPLADE-esque sparse embeddings?

I have been working on embeddings for a while.

For different reasons I have recently become very interested in learned sparse embeddings. So I am curious what led you to choose them for your application, and why?

Some of the best performing embedding models (https://huggingface.co/spaces/mteb/leaderboard) are LLMs. Have you tried them?

It's quite a smear to write "He's probably unemployed at the moment or just brushing off before the next interview.". Cheap smears are a form of casual dismissal and devalue this site.

I was stoked on D1 when it was announced.

But the maximum DB size is 10GB (https://developers.cloudflare.com/d1/platform/limits/).

Really?

That's called a Telegram bot and you can write one yourself.

Don't call me crazy (I am actually), but sometimes I will keep both ChatGPT and Claude open side-by-side and use them to audit each other.

I'll give them the same prompt.

When they respond, re-prompt with: "What are your thoughts on this approach? Pros and cons. Integrate the best ideas from both: [answer from the other model]"

Repeat until total satisfaction or frustration is achieved.

Touch screen I think is cool because then you could interact with the art.

I did a little research and these Waveshare displays look really beautiful. For a particular size display, they have models in different resolutions (from lower to higher res). But they're only for rpis and windows, not like mac or ipad. The QLED ones are the best.

https://www.waveshare.com/product/ai/displays/10.1hp-capqled...

FYI, the link to the frame used to mount the display just blocks you:

https://www.leroymerlin.fr/produits/decoration-eclairage/dec...

For those curious, it's a MILO 21 x 29.7 cm black frame. These links work for me:

https://www.leroymerlin.pt/produtos/decoracao-e-tapetes/mold...

Also, the screen he uses (HMTECH Raspberry Pi Screen 10.1) is pretty hard to find. Do people have other good recommendations for screens with similar quality and specs?

Any idea why this uses infrared light and an infrared camera?

Also, there is at least one synth (Serum) where the developer declined to document the file format of the presets:

"I reached out to Steve through Xfer's forum and he was prompt and helpful. Unfortunately, the .fxp file format is completely dependent on the source code, and he can't release a spec for it without making the code open source. Which probably isn't happening any time soon."

https://www.reddit.com/r/edmproduction/comments/69hxa7/rever...

This looks great but... to work with a github repo I have to give:

* read and write access to ALL gists

* read and write access to ALL repos, public and private

* read and write access to SSH public keys

That's a no from me.

Is the realtime audio API now open to everyone?

Suggestion:

Add "passthrough" field with a float value, and the idea is that you are sub-awarding a percentage of your grant to projects that you rely upon or want to succeed.

Then, people who are not open source maintainers but do spend their time indexing high-quality open source can have a 1.0 pass through funding.json passthrough to repos they believe are worth supporting.

And people who do not spend their time indexing high-quality open source but trust other people that DO spend their time indexing high-quality open source can have an index fund of donation streams.

In which case we can welcome both people who want to endow but don't have time and the people who have time but can't endow as much.

If anyone is interested in talking to me about this problem, I'm down