beklein

Very cool tool, and always amazing what Emacs can do...

If this isn’t your _daily_ use case and you only need to edit video from time to time, just ask your preferred LLM to give you the FFmpeg commands to cut, speed up, mute, flip, add text, etc. This has worked quite well for me with simple use cases, and well no need to learn Emacs.

- Blog post: https://ai.meta.com/blog/dinov3-self-supervised-vision-model... - Paper: https://ai.meta.com/research/publications/dinov3/ - Hugging Face: https://huggingface.co/collections/facebook/dinov3-68924841b...

I guess it is not very likely but I would suggest a solo scientists/innovator as the best candidate...

Near future AI systems, novel research methods and unique datasets (and a lot of luck) could let a single researcher make breakthrough advances. I am thinking of a room-temperature superconductor in materials science or a better transformer architecture in computer science, potentially yielding a billion-dollar patent or company, even if that’s beyond today’s AI system's capabilities.

The specific idea here is not important, the leverage of a new scientific advancement could be though.

Not every company needs to be a SaaS with focus on B2B that won't scale sufficiently for investors with one employee...

Maybe light’s insanely fast and space is just huge. It’s all relative ;)

If you want to make something, have a look at https://huggingface.co/lerobot For hardware check: https://github.com/TheRobotStudio/SO-ARM100

There will be a world wide hackathon in two weeks time, no better way to get started and get to know some people: https://huggingface.co/LeRobot-worldwide-hackathon

On the announcement page they are saying that "...introducing updates to the file search tool that allow developers to perform searches across multiple vector stores...". On the docs, I still find this limitation: "At the moment, you can search in only one vector store at a time, so you can include only one vector store ID when calling the file search tool."

Anybody knows how searching multiple vector stores is implemented? The obvious plan would be to allow something like:

  "vector_store_ids": ["<vector_store_id1>", "<vector_store_id2>", ...]

You can also grab the paper as a PDF here: https://arxiv.org/pdf/2505.05469.

Keep in mind that these sites are run by AI researchers, not dedicated UX teams at major tech companies—so the interface can feel a bit rough around the edges. That said, your critique is still valid; it’s just fair to cut them a little slack given their priorities.

I could find these two resources: Hugging Face: https://huggingface.co/collections/ibm-ai-platform/bamba-674... GitHub: https://github.com/foundation-model-stack/bamba

Great point, but you need to have a good understanding in how LLMs work to understand this limitation. If you don't have an intuitive understanding think like it's one of these draw on my back games, just each new token is a new human in the loop, known words are like simple shapes you felt before 100 times on your back and are easy for you to reproduce and change, random ASCII strings are harder to _grasp_ and will produce a fuzzy output... all models are wrong, but some are useful.

https://www.youtube.com/watch?v=bA_DQHoxj34

I just want to second this. Your prompt asks for a description, and you get a description. If you instead ask something like, "Do or don't you know about the unspoken etiquette ..." you'll get an answer about whether that specific thing exists.

https://chatgpt.com/share/680b32bc-5854-8000-a1c7-cdf388eeb0...

It's easy to blame the models, but often the issue lies in how we write our prompts. No personal criticism here—I fall short in this way too. A good tip is to ask the model again, with the prompt + reply and the expected reply why this didn't work... we all will get better over time (humans and models)

This is amazing! As someone working with industrial robots, normally under strict environmental constraints and control, witnessing such real-world robotics progress truly excites me about the future!

By the way, they’ve open-sourced their π0 model (code and model weights). More information can be found here: https://github.com/Physical-Intelligence/openpi

- Expanded Usage: 5x more usage than Pro => $100 per month

- Maximum Flexibility: 20x more usage than Pro => $200 per month

Older and related article from one of the authors titled "What 2026 looks like", that is holding up very well against time. Written in mid 2021 (pre ChatGPT)

https://www.alignmentforum.org/posts/6Xgy6CAf2jqHhynHL/what-...

//edit: remove the referral tags from URL

Here's the link to the full playlist with 20 video demonstrations (around 1min each) on YouTube: https://www.youtube.com/watch?v=4MvGnmmP3c0&list=PLqYmG7hTra...

You can check the example code on the Mistral documentation, you would _only_ have to change the value of the variable `document_url` to the URL of your uploaded PDF... and you need to change the `MISTRAL_API_KEY` to the value of your specific key that you can get from the Le Platforme webpage.

https://docs.mistral.ai/capabilities/document/#ocr-with-pdf

Thanks for the corrections, that’s what I wanted to say!

They tried a while ago: https://news.ycombinator.com/item?id=40373284

Sadly little people noticed...

If you are only interested in the neat table UI check: https://github.com/githubocto/flat-ui

Tip: If you ever need to do this on a public GitHub repository you can use "gitingest".

This will open a website that creates a copy of all the file contents of the repo (code, docs, ...) It's a great tool to use when using new/obscure code with LLMs in my opinion.

The UX is so just easy and great, change the URL from <https://github.com/user_name/repo_name> to <https://gitingest.com/user_name/repo_name>

//edit: fixed URLs

Great article, I couldn't find any details about the prompt... only the snippets of the `CHUNKING_PROMPT` and the `GET_NODE_BOUNDING_BOXES_PROMPT`.

Is there is any code example with a full prompt available from OP, or are there any references (such as similar GitHub repos) for those looking to get started within this topic?

Your insights would be highly appreciated.

Thank you for all the effort you put into this tool and keeping it up to date!

Very interesting project, thank you for sharing.

Are you supporting the Batch API from OpenAI? This would lower costs by 50%. Many OCR tasks are not time-sensitive, so this might be a very good tradeoff.

Really love your idea and the app, congratulations!

I will probably start referring to my longer checklists as "steplists" — what a great name!

I noticed that abandoned items in Focus Mode are not specially marked as abandoned in List Mode. I would expect these items to sync in both modes for this state. I also prefer a half-checked state for items in progress; it gives you some instant gratification for starting to work on the next item... but that's just personal preference.

I can only recommend visiting any industrial trade shows. New and upcoming products and services will be shown there in plenty. Skip the sales people there and ask for an engineer. They are normally happy to talk to you openly and answer any question. I know this might be the solution involving the most work (and cost) on your part but it’s also one of the fastest ways to build your network in this space.

More like a cousin of LLMs are Vision-Language-Action (VLA) models like RT-2 [1]. Additionally to text and vision data they also include data from robot actions as "another language" as tokens to output movement actions for robots.

[1]: https://robotics-transformer2.github.io

I can recommend redbean[1], it's basically a webserver and your static files in a zip file that is executable on basically any computer. It's also super small, includes TLS and reasonably fast.

[1] https://redbean.dev