Hacker News

LoongForge is an open-source training framework developed by the Baidu AI Cloud Baige team, built to deliver faster training for mainstream LLMs, VLMs, diffusion, and embodied models, thereby significantly reducing costs.

Comments URL: https://news.ycombinator.com/item?id=49713630

Points: 2

# Comments: 0

Article URL: https://borretti.me/article/the-other-road

Comments URL: https://news.ycombinator.com/item?id=49713628

Points: 2

# Comments: 0

Article URL: https://www.ksl.com/article/51622813/chatgpt-invented-fake-police-testimony-in-murder-appeal-new-mexico-high-court-says

Comments URL: https://news.ycombinator.com/item?id=49713621

Points: 3

# Comments: 0

Article URL: https://superdark.ai/

Comments URL: https://news.ycombinator.com/item?id=49713617

Points: 3

# Comments: 0

Article URL: https://princeton-introml.github.io/

Comments URL: https://news.ycombinator.com/item?id=49713599

Points: 2

# Comments: 0

I’m building FlareDB, a Rust-based streaming database with Apache Beam as programing interfere The latest release, v0.2.0, i launched FlareDbIO, which lets data processing pipelines write results into FlareDB, query them using SQL, and read the results back into another pipeline. The goal is to make processed data persistent and without having to maintain a external serving database

Docs: https://docs.flare-db.com/io

GitHub: https://github.com/flare-db/flare-db

Comments URL: https://news.ycombinator.com/item?id=49713583

Points: 2

# Comments: 0

This is an update to our old Trendradar (https://news.ycombinator.com/item?id=45332076)

# What changed * This trend radar focuses on trends in general (not only technological trends as the previous ones) * This trend radar splits individual trends into granular signals, and each signal is based one 3 or more individual sources.

The idea is to provide companies a data-driven scan of the environment (horizon scanning) as a basis. With this basis, they can derive conclusions about the direction of a market.

Although V2, there are still a few issues to fix, especially around self-reporting & PR sources which tend to overrepresent trends. Furthermore, we are planning on expanding it to include a more diverse set of sources (especially patents & GitHub), to catch changes even earlier.

Ironically, AI has made web scraping per se cheaper, but overall it seems that the total costs (as a consequence of complexity and the sheer number of content) have increased.

Comments URL: https://news.ycombinator.com/item?id=49713580

Points: 2

# Comments: 0

Article URL: https://github.com/mukundzha/hazzel

Comments URL: https://news.ycombinator.com/item?id=49713561

Points: 4

# Comments: 0

Article URL: https://www.aljazeera.com/news/2026/9/14/attacks-will-be-fully-autonomous-russia-ukraine-race-towards-ai-warfare

Comments URL: https://news.ycombinator.com/item?id=49713543

Points: 2

# Comments: 0

Article URL: https://www.theguardian.com/music/2026/sep/15/green-lung-druidess-and-the-return-of-folk-horror-to-british-heavy-metal

Comments URL: https://news.ycombinator.com/item?id=49710612

Points: 1

# Comments: 0

Article URL: https://apps.apple.com/us/app/icao-epassport-inspector/id6762305723

Comments URL: https://news.ycombinator.com/item?id=49710592

Points: 1

# Comments: 0

Article URL: https://fastpaced.com/articles/boring-lfs/

Comments URL: https://news.ycombinator.com/item?id=49710591

Points: 1

# Comments: 0

Article URL: https://www.technologyreview.com/2026/09/15/1144028/ai-infrastructure-boom-investment-bubble-risk/

Comments URL: https://news.ycombinator.com/item?id=49710588

Points: 1

# Comments: 0

Article URL: https://hardcaml.org/

Comments URL: https://news.ycombinator.com/item?id=49710585

Points: 1

# Comments: 0

Article URL: https://emsh.cat/en/the-people-are-still-there/

Comments URL: https://news.ycombinator.com/item?id=49710584

Points: 1

# Comments: 0

Article URL: https://projectgoldscript.com/development/becoming-a-time-mage/

Comments URL: https://news.ycombinator.com/item?id=49710558

Points: 2

# Comments: 0

Article URL: https://coinmarketcap.com/community/post/379296157/

Comments URL: https://news.ycombinator.com/item?id=49710525

Points: 1

# Comments: 0

I built GSL because I got fed up of building architecture diagrams that couldn't quite capture all of the nuance of the real world. I liked the idea of using a system like Mermaid or PlantUML to represent my systems but there was never really a satisfactory way of describing what was happening.

I liked the easy-to-follow syntax of those diagramming tools but they very quickly ran into limitations. When you describe an API gateway component, you can't really say that it belongs in a specific region but also that it belonged in a certain zone of the architecture or which team owns it. And when I described what systems interacted with each other I only had the ability to use a text field to describe all of the metadata about the relationship. Is it HTTPS vs an async lambda invocation? What auth do we need?

Something else I wanted was the ability to query the architecture. Surely when you have the full graph of what talks to what, you should be able to pose questions like: * "which customer actions trigger a call to vendor X's API?" * "what will be affected if database Y goes down?"

I couldn't really find anything that did that in a way that felt natural. Diagramming tools were either completely visual or too presentation focussed to express what I wanted. Existing graph languages felt too clunky or cumbersome. I struggled to find any that worked with sets and querying in a way that felt right.

So one rainy afternoon, I decided to see how far I could get with ChatGPT and some ideas - GSL was born!

In the repo there is: * A specification for GSL, a language for defining directed graphs with nodes and edges along with set annotations to declare cross-cutting relationships * A specification for GQL, the companion query language * gsl-query - a CLI tool for executing queries on your GSL models * gsl-diagram - a very primitive tool for turning GSL into Mermaid or PlantUML diagrams * A small VS Code plugin for syntax highlighting * Guides, tutorials, documentation * Rich examples (including some non-tech oriented ones)

I think GSL has turned into the language that I was looking for at the beginning of the year. I'm starting to use it in my day job. The guides and documentation have been written to make it easier for agents and LLMs to use and my casual use so far has shown some promise.

It's still very much an experiment but I'd like to get other people to take a look and see if it's useful for them too. Have I even made the right assumptions and picked the right abstractions? Have you tried to do something like this yourself?

You can begin with the "Getting Started" guide, the specs, or just jump right in to the flagship examples. "Example #3: Architecture Across Boundaries" is fairly close to what I wanted to do originally.

https://github.com/dnnrly/gsl-lang

Comments URL: https://news.ycombinator.com/item?id=49710515

Points: 1

# Comments: 0

I've been working on and off on this project to have a reasonably easy way to run a docker sandbox on my machine. Recently with claude "auto mode" not working as "automatically" as before I resurrected this. Hopefully it is of use to you too.

Usage: sclaude / scodex just like you'd call claude or codex.

To update: sclaude update (updates the script and tooling)

It runs locally in a container. Docker also works via Docker-in-Docker (but can be turned off).

Feedback welcome!

Comments URL: https://news.ycombinator.com/item?id=49710493

Points: 1

# Comments: 0

Article URL: https://github.com/rabbiveesh/composr

Comments URL: https://news.ycombinator.com/item?id=49710485

Points: 1

# Comments: 1