Posted by david927 1 day ago
Ask HN: What Are You Working On? (March 2026)
Interesting findings include Mistral doing better than Gemini 3 Pro in certain usescases, cross-LLM works better than one LLM to another, oh and - the cost all of of this. So, so expensive.
Applications on the public cloud raise strong concerns about data protection. As an architect, I spend a meaningful part of my time ensuring the security of customers’ data in the cloud.
Bao introduces an innovative approach where data remains on local devices while the cloud provides encrypted storage for synchronization and peer exchange. Because cloud providers cannot access the data, the need for due diligence is reduced.
Any feedback is welcome.
Also used the new Navigation API (and some Shadow DOM) to build a cheap, custom client-side rendering (sort of) into my site (https://taro.codes), and some other minor refactors and cleanup (finally migrated away from Sass to just native CSS, improved encapsulation of some things with Shadow roots, etc).
I've been wanting to write a simple AI agent with JS and Ollama just for fun and learning, but haven't started, yet...
https://github.com/RedbackThomson/nix-tasks
I started this project because at my company, we're still relying on ancient Makefiles as our build system and build tool versioning. I initially looked at using other task runners but they all use some sort of DSL that I think limits their functionality and/or doesn't allow for sharing and extending templates across repos. Nix-tasks lets you use Nix flakes to share common configuration - like your company-wide build scripts - and then import it and add repo specific tasks on top of them.
The project is still very much in alpha but I am using it every day and trying to find any annoyances or bugs before I share it further.
I was stuck on this conversation problem. First version had a dead-end search box: six starter prompts, one referencing a tool that didn't exist. No follow-ups. No guided flows. Users got an answer and had to invent the next question from scratch.
Now the assistant explores your library with you. Tag discovery, color browsing, weekly digests, smart collections that auto-curate as you save.
Semantic search runs hybrid, keyword matching plus pgvector cosine similarity on 768-dim embeddings. Streaming responses.
Almost there. https://bookmarker.cc/
I do wonder if the problem is not so much having a place to find LAN events but actually just having enough people put on LAN events in the first place. It feels like a thing of the past with how much less people interact in person these days. It's a shame because LANs are awesome!
Have you thought about ways to make it easier for people to host LAN events? Or does this solve that as well? I guess a solution would require matching random people together. Happy to discuss more - nick at onthe.town
> I do wonder if the problem is not so much having a place to find LAN events but actually just having enough people put on LAN events in the first place.
Sort of! I did a lot of research on this before I built lan.events. There are more gamers than ever, but LANs dropped off during COVID lockdowns despite surveys showing an increasing interest in in-person events. More or less, it's actually a venue problem. Running events has incredibly thin and risky margins for something that by its very nature needs to be planned out months in advance. Everything around the events are becoming prohibitively expensive: venues, vendors, equipment rentals, etc are all eating away at the ceiling gamers will pay and the floor that organizations can charge from.
LAN.events helps tackle this by decreasing the cost per ticket and shifting that cost to the customer rather than the event manager. We don't introduce minimum event costs or percentage based pricing which lets event managers keep or give back more profits. There is more I can do in this space, but that's the biggest way I can contribute right now.
Most of the work as of today is in a branch, can see the language spec at https://github.com/cretz/duralade/blob/initial-runtime/docs/..., and some samples at https://github.com/cretz/duralade/tree/initial-runtime/sampl....
May not amount to anything, but the ideas/concepts of this durable language are quite nice.
https://github.com/dbos-inc/dbos-demo-apps/blob/main/python/...
Admittedly the lang spec doesn't do a great job at the justification side, but the engine spec adjacent to it at https://github.com/cretz/duralade/blob/initial-runtime/docs/... that has sections showing CLI/API commands can help make it clearer where this runtime is unique.