Feed aggregator

Article URL: https://www.cse.chalmers.se/~rjmh/Combinators/

Comments URL: https://news.ycombinator.com/item?id=46954159

Points: 1

# Comments: 0

I built an autonomous research agent in C# that runs entirely on local LLMs via Ollama.

Give it a topic and it: generates diverse search queries, searches the web (Brave Search API, free tier), fetches and reads relevant sources, analyzes each source for key findings, and synthesizes a structured markdown report with citations.

Everything runs locally — no OpenAI/Anthropic API needed. Just Ollama + llama3.1:8b.

It takes about 15 minutes per research run on a mid-range CPU (Ryzen 5 5500, no GPU needed). Not fast, but it does the research while you do other things.

Tech stack: C#/.NET 8, Ollama, SQLite for semantic memory, Brave Search API.

Why C# instead of Python? Because .NET developers want AI tools too, and the ecosystem is underserved compared to LangChain/LlamaIndex.

Known limitations: CPU inference is slow (~15min/run), 8B models occasionally produce malformed tool calls (handled with retries), and research quality depends on Brave Search results for your topic.

I also made a starter kit if you want to build your own agent from scratch: https://github.com/DynamicCSharp/agentkit

Comments URL: https://news.ycombinator.com/item?id=46954141

Points: 1

# Comments: 0

Article URL: https://github.com/TrevorS/voxtral-mini-realtime-rs

Comments URL: https://news.ycombinator.com/item?id=46954136

Points: 1

# Comments: 0

Article URL: https://k2xl.substack.com/p/the-first-time-i-visited-metas-headquarters

Comments URL: https://news.ycombinator.com/item?id=46954127

Points: 1

# Comments: 0

- Live demo: https://compsult.github.io/Quickpick-ui/ - GitHub: https://github.com/compsult/Quickpick-ui

Comments URL: https://news.ycombinator.com/item?id=46954126

Points: 1

# Comments: 0

Article URL: https://www.youtube.com/watch?v=Ru7BIqXQZns

Comments URL: https://news.ycombinator.com/item?id=46954123

Points: 2

# Comments: 0

Article URL: https://extended.reading.sh/ai-dot-com-crashes-on-superbowl

Comments URL: https://news.ycombinator.com/item?id=46954096

Points: 1

# Comments: 1

Article URL: https://jollycoder.com/posts/rise-of-the-cowboy-coder

Comments URL: https://news.ycombinator.com/item?id=46954080

Points: 1

# Comments: 0

With GitHub's recent instability, I'm seriously looking into alternatives. But the one piece of the puzzle I haven't figured out yet is GitHub Actions. I've used both TeamCity and Jenkins in the past, and have no desire to go back to either one; GHA, for all its flaws, was just so much better of an experience than TeamCity or Jenkins, for me at least. I've read about multiple alternatives to GHA for CI, but it's hard for me to tell just from the documentation what they feel like to use.

So, for those of you who have used left GitHub Actions and are using a different CI solution, I'd like to ask: what did you settle on? Why? Are you happy with it? And how does it feel to use it? Do you feel like it's invisible and gets out of the way, or do you feel like you're constantly having to massage its configuration all the time?

Comments URL: https://news.ycombinator.com/item?id=46954054

Points: 1

# Comments: 1

AI chip firm’s £100m expansion and facility in Bristol is being used to showcase the government’s AI opportunity plan in action

Article URL: https://github.com/AGDNoob/squish

Comments URL: https://news.ycombinator.com/item?id=46953779

Points: 1

# Comments: 0

Article URL: https://www.japantimes.co.jp/culture/2026/01/23/film/black-ox-tetsuichiro-tsuta/

Comments URL: https://news.ycombinator.com/item?id=46953742

Points: 1

# Comments: 0

Hi HN, I'm new here, wanted to share a little lib I've been working on. It's a cache library that lets you declare dependencies instead of classic TTL invalidation.

Basic usage:

from zoocache import cacheable, invalidate, add_deps @cacheable def get_product_page(pid: int, sid: int): product = db.get_product(pid) add_deps([f"product:{pid}"]) stock = db.get_stock(sid, pid) add_deps([f"store:{sid}:stock"]) return render(product, stock) get_product_page(42, 1) # computes and caches get_product_page(42, 1) # cache hit invalidate("product:42") # product info changed invalidate("store:1") # clears ALL store:1:* (stock, prices, etc) Tags are hierarchical, invalidating a parent like "store:1" clears everything under it. No scanning or pattern matching needed.

Internally it's a prefix trie with HLC for distributed consistency, a SingleFlight to avoid thundering herds, and MsgPack+LZ4 serialization. The core is Rust via PyO3. Supports in-memory, LMDB, and Redis backends.

It's a fairly niche tool, if simple TTL or lru_cache covers your use case, you don't need this. It's more for when you have complex data relationships and need precise, immediate invalidation.

Source: https://github.com/albertobadia/zoocache

Hope you find it useful. Happy to receive feedback, PRs, or discuss design tradeoffs.

Comments URL: https://news.ycombinator.com/item?id=46953736

Points: 1

# Comments: 0

Article URL: https://www.reuters.com/science/musk-says-spacex-prioritise-building-self-growing-city-moon-2026-02-08/

Comments URL: https://news.ycombinator.com/item?id=46953730

Points: 1

# Comments: 0

Article URL: https://flaneur.ink

Comments URL: https://news.ycombinator.com/item?id=46953632

Points: 1

# Comments: 1

Article URL: https://github.com/championswimmer/env.sync.local

Comments URL: https://news.ycombinator.com/item?id=46953617

Points: 1

# Comments: 0

Article URL: https://princeea.com/mexican-scientist-eva-ramon-gallegos-is-the-first-to-eradicate-hpv-in-29-women/

Comments URL: https://news.ycombinator.com/item?id=46953616

Points: 1

# Comments: 0

Article URL: https://apocryphos.com/post/windows_wlan-report/

Comments URL: https://news.ycombinator.com/item?id=46953593

Points: 1

# Comments: 0

Article URL: https://zenodo.org/records/18571935

Comments URL: https://news.ycombinator.com/item?id=46953585

Points: 1

# Comments: 1

Adobe recently announced the end-of-life for Adobe Animate, then walked it back after community backlash.

Regardless of what Adobe decides next, the message was clear: animators who depend on proprietary tools are one corporate decision away from losing their workflow.

2D animation deserves an open-source option that isn't a toy. We've been working with a professional animator to guide feature priorities and ensure we're building something that actually fits real production workflows - not just a tech demo.

Github Repo: https://github.com/17twenty/inamate

We're at the stage where community feedback shapes the direction. If you're an animator, motion designer, or just someone who's been frustrated by the state of 2D animation tools — we'd love to hear:

- What features would make you switch from your current tool?

- What's the biggest pain point in your animation workflow?

- Is real-time collaboration actually useful for animation, or is it a gimmick?

Try it out, break it, and tell us what you think.

Built with Go, TS & React, WebAssembly, PostgreSQL, WebSocket, ffmpeg (for video exports).

Comments URL: https://news.ycombinator.com/item?id=46953567

Points: 1

# Comments: 1