julius

I always heard of GB/s as most important number ... AI told me their 8800 MT/s on 8 Byte, 12 DDR5-Channels means 845 GB/s. A Nvidia RTX 4090 has 1008 GB/s. Nvidia B200 has 8000 GB/s. Is this the right way of looking at it?

The best kept secret in AI training. Tianyi, since you are our insider. When you walk by the training teams office - how often do you see Pelicans on their screens?

CLIs need installs. What if your customer uses Web-based Agent? What if your customer has IT that does not easily allow install/updates of your CLI?

CLIs are working better with current harnesses in my experience, but no idea how I would ship them.

Maybe there is a smart trick to get them to do the right thing, but the things I tried did not work.

At one point I had some smaller model draw bounding boxes around everything that looked interactable and labels like "e3" ... then asked the model to tell me "click on e3". Did not work in my tests was pretty much as bad as x,y.

Click coordinates. Agentic GUI is really annoying when the multi-modal agent cannot click on x,y coordinates.

I tested Qwen3.6, Gemma4, Nemotron3-nano-omni. They fully hallucinate x,y coords. (did not try GLM-5V yet)

GPT-5.5 can easily do it. But also Vocaela, a tiny 500M model, is quite good at it. Hope they improve the training for x,y clicking soon on the smallish multi-modals.

Recently slopped a http service together just so my local models can click, instead of relying on all the wild ways agents currently hack into the browser (browser-use, browser-harness, agent-browser, dev-browser etc) https://github.com/julius/vocaela-click-coords-http

Super cool. Brave support by any chance? Using Linux, it found my Chrome, but thats not my primary browser.

Less information loss -> Less params? Please correct me if I got this wrong. The Intro claims:

"The dot product itself is a geometrically impoverished measure, primarily capturing alignment while conflating magnitude with direction and often obscuring more complex structural and spatial relationships [10, 11, 4, 61, 17]. Furthermore, the way current activation functions achieve non-linearity can exacerbate this issue. For instance, ReLU (f (x) = max(0, x)) maps all negative pre-activations, which can signify a spectrum of relationships from weak dissimilarity to strong anti-alignment, to a single zero output. This thresholding, while promoting sparsity, means the network treats diverse inputs as uniformly orthogonal or linearly independent for onward signal propagation. Such a coarse-graining of geometric relationships leads to a tangible loss of information regarding the degree and nature of anti-alignment or other neg- ative linear dependencies. This information loss, coupled with the inherent limitations of the dot product, highlights a fundamental challenge."

Lots of people who have relatively stable currencies (EUR, USD..) do not want to use bitcoin. What if bitcoin price goes down? How many extra steps is it to convert my USD to bitcoin and then back to USD? Do I only convert the 19.99 USD for my current purchase into bitcoin or do I put in more?

Do you solve these issues for customers? Or are you only targeting people who already are happy bitcoin wallet users? Are stablecoins part of your strategy?

Given how Visa,Mastercard,Paypal are seen as bad actors. Do you think you can capitalize on that, possibly partnering with Valve or something of that sort?

Chrome/Brave: https://chromewebstore.google.com/detail/youtube-no-translat...

Anyone with recent real-world experience?

From talking to AI, it seems the main issues would be:

- SEO (googlebot)

- Social Media Sharing

- CSP heavy envs could be trouble

Is this right?

Video of the ship and visualization of the technology: https://www.youtube.com/watch?v=wOK4TGd_l_Q

Thanks for making this playable. I have seen videos of it, but thought I had to wait for years until I can experience it.

The future will be wild. "Hey ChatGPT, lets play Counterstrike on the Enterprise-D. Counter-Terrorists agains Spongebobs"

At first I thought you made a website that gives me an empty Markdown file. But I am glad I downloaded it its actually a pretty nice template.

What are you personally doing with the yearly goals in that file. Are you copy and pasting them from last week, or are you typing them down everytime to re-iterate them (and possibly even modify) ?

Mercury sounds interesting. Requires a certain scale though (gravity is a bitch).

Considering just the initial mining and construction, bodies with low gravity and proximity to the earth feel like an efficient starting point, right? I always thought the moon would be a good place to bootstrap the first few thousand space habitats.

Your point about energy will probably be the biggest deal. Wondering how complicated it would be to ship a bunch of nuclear reactors to the moon. There seems to be quite a few companies working on small, "mass produced" reactors currently.

- Demo made me want this instantly -- looks better than tabs

- Realized its VSCode -- started downloading

- Stopped the download when I saw AI -- cannot work on company code

Will definately try this when AI is disabled.

I also just used ElasticMQ in a docker environment. Did you see any specific downsides, when you looked at it? In which scenarios should I consider replacing it with your solution?

Oh wow TIL ESP32 can run TensorFlowLite. Person detection in 54ms! https://github.com/espressif/esp-tflite-micro?tab=readme-ov-...

From their homepage it seems they do screensharing + AI summary of meeting.

Paper: https://arxiv.org/pdf/2406.05629v1

Effects wore off after 18 months I think. He explains in the video why it was expected.

He took the pills 6 years ago still not killed by cancer. And of course this is n=1. The first thing he says after taking the pills is "Lets hope this does not kill me" :D

Reminds me of the Youtube Guy who homebrewed a gene-therapy to treat his Lactose Intolerance, successfully

https://www.youtube.com/watch?v=aoczYXJeMY4

Thanks for sharing your product journey. Making hardware products is very interesting, when a single guy does it.

Would you mind sharing some of your prototype to production journey?

- How did you find companies that make the parts (pcb, enclosure)?

- How do you assemble all the parts (do you do that yourself or did you find a company to do that for you)?

Feels like science papers need a comment section. Replace peer-review with public-review. A way for authors to interact with the larger (science) community.

https://www.papertalk.xyz/ was on HN Frontpage but seems to not have gained any traction (yet).

Maybe arxiv should consider implementing it or integrating with some 3rd party?

One of the sites has a fantastic landing page: https://needle.tools/ (you scroll through 3D scenes)

And I hope we will have good AIs running on our own hardware. I obviously want to use AI to create lots of porn for me. And that is really no one else s business.

Speaking of - I already heard of a project that is fine-tuning StableDiffusion to produce porn.

How do they have sooo many employees. They are running a few websites and a bunch of apps. I can understand how they would need a hundred people for that. They have a massive userbase. So I can understand how they would need a thousand people just to deal with tech at scale and support.

But how do you manage to have so many people that you can fire 10k ?! If I googled well, they appear to have 72k employees. Where did you find the management that thought having 72k employees is reasonable to operate a bunch of websites and apps ?!

ChatGPT was my first choice just because I want to know if it is any good at helping me with my work.

But for this problem (CF-Distribution lost OAC settings when updating the root file), all the google fu in the world did not help me. It turned out that I had to update my aws-cli and my problem went away. Apparently no one else on the internet had that problem, so only my gut could help me figure it out.

So they are using sound waves to form very exact force-fields in 3D. But when you throw matter in there, would that not interact and reflect all the waves in unpredictable ways, destroying the very exact force-field pattern they are trying to achieve?

I asked it how to do something with an AWS cli tool. ChatGPT invented a new parameter that looked like a programmer would come up with it and it would do exactly what I was looking for (I assumed many people had my problem before).

aws cloudfront update-distribution --id <distribution-id> --distribution-config <new-config> --no-reset-origin-access-identity

Took me a while to figure out that the parameter --no-reset-origin-access-identity was not only not working. But it did never exist on any version of the cli tool.

Computer Science?

If you just want a career in making computer software. Build something. Then do it again. Google all the stuff you need to build your current thing. After a few month it will be obvious to you that you can build software. Then talk to companies about junior positions. Tell them you are self-taught and show them the stuff you have built already.

Best to talk to a company (usually small companies are easier), where you are allowed to talk to a programmer early on. The fact that you learned everything by yourself means they do not have to hold your hand all the time. Hand-holding time is the most important consideration for me, when taking on junior devs.