Hacker News

Article URL: https://forecasting.substack.com/p/forecasting-newsletter-32025-long

Comments URL: https://news.ycombinator.com/item?id=43292560

Points: 1

# Comments: 0

Want to train your own AI model? You've gotta start somewhere. If you've never tried it before, here is your chance!

Running an AI model has two parts, during the training phase you set weights, and during the inference phase you use the weights to predict words and other tokens (such as punctuation).

This minimal implementation in pure javascript runs right in your browser:

https://taonexus.com/mini-transformer-in-js.html

You can provide your own training data if you want: words.txt is a dictionary of punctuation and output tokens each on their own line and input.txt is a text corpus to train on.

If you load them from the server, you get words.txt from the top 2000 or so words that appear in in the Project Gutenberg free public domain corpus (as well as a few punctuation marks). For input.txt we've used a selection of 2000 lines of Shakespeare text.

After you finish training (you can stop early if you are impatient), on the Inference tab you can write a few words to have the model complete it for you.

Congratulations! After completing these steps you can now boast that you have trained your own AI model.

Comments URL: https://news.ycombinator.com/item?id=43292537

Points: 1

# Comments: 0

Article URL: https://www.theguardian.com/us-news/2025/mar/07/cdc-study-vaccines-autism

Comments URL: https://news.ycombinator.com/item?id=43292519

Points: 1

# Comments: 0

Article URL: https://parv.bearblog.dev/splines/

Comments URL: https://news.ycombinator.com/item?id=43292503

Points: 1

# Comments: 0

Mistral recently launched their OCR model, claiming to be the best in the world. We put that claim to the test against JigsawStack vOCR — our own OCR solution, built by a team of just three people.

We benchmarked both models across real-world multilingual documents, handwriting recognition, structured data extraction, and bounding box accuracy. Here’s a quick breakdown:

- Languages: Mistral supports 12 benchmarked languages, while JigsawStack vOCR handles 70+ including Telugu, Hindi, and lesser-used scripts.

- Handwriting Recognition: Mistral struggles with handwritten and distorted text, while JigsawStack vOCR accurately extracts text from printed materials, handwriting, and even text on walls.

- Structured Output: Mistral requires additional post-processing with an LLM for structured data, while JigsawStack natively returns structured JSON output.

- Bounding Boxes: Mistral OCR does not provide bounding box data, while JigsawStack supports both sentence and word-level positions.

Check out the full breakdown with examples, screenshots, and API comparisons in our blog post Mistral OCR vs. JigsawStack vOCR here: https://jigsawstack.com/blog/mistral-ocr-vs-jigsawstack-vocr

We'd love to hear feedback and answer any questions! If you’re building with OCR, try out JigsawStack vOCR and let us know your thoughts.

Comments URL: https://news.ycombinator.com/item?id=43292499

Points: 3

# Comments: 0

Article URL: https://benhouston3d.com/blog/github-mode-for-agentic-coding

Comments URL: https://news.ycombinator.com/item?id=43292491

Points: 1

# Comments: 0

Article URL: https://www.programmablemutter.com/p/coming-up-with-an-alternative-silicon

Comments URL: https://news.ycombinator.com/item?id=43292477

Points: 1

# Comments: 0

Article URL: https://www.theguardian.com/science/2025/mar/07/athena-spacecraft-mission-dead

Comments URL: https://news.ycombinator.com/item?id=43292471

Points: 4

# Comments: 0

Article URL: https://aider.chat/2024/12/03/qwq.html

Comments URL: https://news.ycombinator.com/item?id=43292462

Points: 1

# Comments: 0

Article URL: https://www.cdc.gov/acip/disclosures/by-member.html

Comments URL: https://news.ycombinator.com/item?id=43292455

Points: 1

# Comments: 0

After writing about Meta’s highly misguided $50 million bounty for developers to create content in Horizon Worlds, a former software engineer at Horizon Worlds reached out to me. They had read my latest book, where I devote a chapter to Meta’s many missteps in trying to build the Metaverse.

But this engineer had even more surprising details to share.

I’ve always believed the fundamental problem is that Meta leadership never truly understood the metaverse concept, and simply treated it like a 3D version of Facebook. In interviews for the book, it also became clear to me that most of the people working on Horizon Worlds weren’t themselves experienced or passionate about virtual worlds.

Indeed, in 2022, Meta leadership sent out an internal memo requiring employees to dogfood Horizon Worlds more (i.e. actually play it).

It was actually worse than that, this ex-developer tells me. Required to dogfood their own virtual world, the engineer tells me, many Meta staffers automated their dogfooding:

"Before I left they were mandating that employees spend a certain number of hours per week in the game actively playing it. So therein started an automation war where all the people with 200 hours a week never actually played the game once. People just had to launch the game with an Android command over USB, then make sure the proximity sensor on the headset was taped to keep it on."

Yes: Instead of playing Horizon Worlds, developers of Horizon Worlds at Meta figured out a hack where they could just pretend to do so.

Meta’s assumptions were evident even on the code level, with Meta treating the Metaverse as a 3D version of a mobile app:

"Horizon Worlds / Workrooms, etc. is a pretty awful codebase with hundreds of people working on it. They grabbed a bunch of people from the Facebook/Instagram half of the company because they knew React. Horizon Worlds uses a VR version of that called ReactVR.

"What this effectively means is that most of the people developing Horizon Worlds (HW) are 2D app developers driven by engagement metrics and retention numbers. So... HW became flooded with a ton of 2D developers who never put on the headset even test their menus, all competing to try to create the most 'engaging' menu that would sell microtransactions, or drive social engagement, or make some other number look good - because that's WHAT THEY DO at Facebook/Instagram."

I go into much more detail on my Patreon, free to read here: https://www.patreon.com/posts/what-went-wrong-123796458

Comments URL: https://news.ycombinator.com/item?id=43292444

Points: 2

# Comments: 1

Article URL: https://www.bbc.com/news/articles/cy83r93l208o

Comments URL: https://news.ycombinator.com/item?id=43292442

Points: 2

# Comments: 1

Article URL: https://www.nytimes.com/2025/03/07/world/canada/trump-trudeau-canada-51st-state.html

Comments URL: https://news.ycombinator.com/item?id=43292433

Points: 3

# Comments: 1