sean_pedersen

Good project but this one also exists https://huggingface.co/spaces/multimodalart/jev-decision-ind... and the results do not seem to add up and also model sets are different... still needs time to mature likely

Try a MoE model like Qwen3.6 35B-A3B for better tok/s

There are LLM models which compute only using addition no?

Just gave it a spin and feels very rough / unpolished. - Why is the default scroll wheel action panning vertically and not zooming in and out (the main interaction)? - Why is the app size 1GB? - Everything feels clunky / not native - these are not my real apps.

This would need a lot of work to feel frictionless but I like the concept.

Looks very interesting - is it open-source yet?

"Any attempt to build AI generated content (deep fake) detection systems is flawed, since the outputs of such a system may be used to train an even better fake data generator. This leads to an equilibrium state of digital uncertainty: nothing in the digital realm can be deemed as real anymore - only as digital. I do not care if a digital artifact is human or AI made - I only care if it is useful to me. Useful content is on point, factual and at best surprising (teaches something new)." - https://seanpedersen.github.io/posts/digital-uncertainty/

It is not about AI generated or not, it is and always has been about quality of the content. So just let the people up- / downvote.

Digger Solo - a smart file explorer with semantic search and maps for your files (images, videos, text, audio). All running locally on your machine.

https://digger.so/o

Please explain in simple words what it is and does. Is user data stored unencrypted on your servers?

as long as the text is not slop (no hallucinations / factual) and useful to me, I do not care about the author of it

very cool! would be useful if headings where linkable using anchor

There is a Rust port: https://github.com/Dicklesworthstone/pi_agent_rust

The test acc. UX flow is shit IMHO: I do not want to see the first user tips (just annoying flashes) and I can not directly edit the first doc I see.

Like the idea but it spawns a terminal on startup on Mac and is not WYSIWYG (like Obsidian). Hope this project develops into usable state soon.

Would this kind of bug have been catched by the Rust compiler?

By this reasoning aphantasiacs should be incapable of drawing anything from their mind.

Link to the OG paper: https://arxiv.org/abs/2503.01781

Maybe fight fire with fire and respond as default with a non-sensical question and see if the bug reporter responds genuinely confused or happily tries to engage in a non-sense convo…

I am working on a visual search & exploration engine: https://digger.lol

The goal is to create beautiful and useful maps of interesting data, empowering the user to explore more intuitively guided by semantic similarity. No user data needs to be tracked for this to work, the data speaks for itself.

This roughly works by translating semantic (visual or textual) similarity into spatial proximity. Diggers major features are: semantic mapping, text search and image search. The text and image search works bidirectionally, allowing to search for images (e.g. product images) using text and for text (e.g. books) using images.

I thought import names and PyPI names are not always equal, thus this can not work reliably, right?

I agree a loading indicator (spinner f.e.) is needed, I was confused when nothing happened. Also big yes on for a reset / delete feature.

I agree in that a perfectly consistent dataset won't completely stop statistical language models from hallucinating but it will reduce it. I think it is established that data quality is more important than quantity. Bullshit in -> bullshit out, so a focus on data quality is good and needed IMO.

I am also saying LMs output should cite sources and give confidence scores (which reflects how much the output is in or out of the training distrtibution).

I wrote up this blog post in 30 mins, that's why it reads a little rough. I could not find explicit research on the impact of contradicting training data, only on the general need for high-quality training data.

May be it is a pipe dream to drastically improve on hallucinations by curating a self-consistent data set but I am still interested in how much it actually impacts the quality of the final model.

I described one possible way to create such a self-consistent data set in this very blog post.

I like pixi (https://pixi.sh/latest/). Let's me pin python version, install packages from conda and PyPI. And also written in Rust.

This article screams LLM generated...

A lib generating 2D artifacts might benefit from showcasing these artifacts using image technology for the curious minds.

The reverse is a "Quantum" tree (DAG): 4->1,4->8

"That inverse doesn’t exist." That is my new math contribution apparently then. I am happy to explain my proof.

I can inverse f(x)=f(x+1) easily as g(x)=f(x)-1 since g(f(x)) = x do you understand my functional inversion approach now better?