Why? I thought it's a cool type of dataset (short/high density news summaries) to test some local RAG approaches. Gemma 4 gives some impressive results, but could probably use some more tweaking on the system prompt.

Comments URL: https://news.ycombinator.com/item?id=47624199

Points: 1

# Comments: 0

Why? I thought it's a cool type of dataset (short/high density news summaries) to test some local RAG approaches.

Comments URL: https://news.ycombinator.com/item?id=47605708

Points: 2

# Comments: 0

Teletext is a TV broadcast system that delivers highly concise, space-constrained news texts. That conciseness turns out to be a real advantage for semantic search and retrieval: dense information, little fluff, and consistent structure over decades.

The project runs fully offline and uses embeddings + Postgres/pgvector to explore, search, and query teletext news from the last 25 years with modern LLM techniques.

Repo: https://github.com/r-follador/TeletextSignals

Comments URL: https://news.ycombinator.com/item?id=46639122

Points: 1

# Comments: 0

I'm positively surprised on how well it works, especially if you also connect it to an LLM.

Another thing: although not strictly metric, but European recipes also use tablespoon and teaspoon as measurements for smaller volumes, so no need to convert this.

Just my two cents, other than that very nice work!!

I still believe it's the way forward.

It would be interesting to overlay TESSERA data there, although the resolution is of course very different.

One issue is however that the actual costs are not so much in early R&D (what the publicly funded universities and hospitals are doing), but in the later stage (clinical trials) which needs deep pockets and appetite for risk, which only big pharma has, because they see a potential big payout.

There's some interesting examples in the Readme.

Of course, nothing so exciting to be discovered in Switzerland anymore ;)