Been using ai to work out ideas for rostra. I could write a list /warning
My main idea is that if it allowed private access, then it basically becomes my forever file archive which will never die as long as I’ve 2-4 laptops running it.
Long term storage of my files/data is a useful feature, imo. You see any attractiveness there?
After decades of partial results, Christian Coester, Elias Koutsoupias, and Marek Zbysiński prove the k-server conjecture: the deterministic work-function algorithm is k-competitive on every metric space. Their proof represents the work function as a matrix, turns request updates into basis changes and row replacements, and derives the amortized bound from a larger matrix-valued potential.
HN readers called it a long-sought “holy grail” of competitive analysis and translated the online/offline guarantee into dispatching examples; discussion also noted that the abstract is unusually opaque to non-specialists.
On Linux, swap files are generally easier to add, resize, and remove than dedicated swap partitions, and modern systems no longer impose the old performance penalty. The gist gives concrete mkswap --file, swapon, and /etc/fstab commands; util-linux 2.41 also sets nocow so the recipe works on Btrfs.
The HN discussion adds important caveats: zram or zswap may be preferable for some workloads, traditional swap remains useful for incompressible pages and hibernation, and filesystem/kernel details matter more than the article’s “always” framing.
On Linux, swap files are generally easier to add, resize, and remove than dedicated swap partitions, and modern systems no longer impose the old performance penalty. The gist gives concrete mkswap --file, swapon, and /etc/fstab commands; util-linux 2.41 also sets nocow so the recipe works on Btrfs.
The HN discussion adds important caveats: zram or zswap may be preferable for some workloads, traditional swap remains useful for incompressible pages and hibernation, and filesystem/kernel details matter more than the article’s “always” framing.
People on X keep whining about Astra using too much of their weekly quota, but after a week after switching, I must say it uses just a bit more than Sol. I use it with slightly lower effort levels comparing to Sol, in a blend of agents based on a real in my multi-agent setup (see screenshot).
I suspect people complaining about it are hitting some usage patterns and/or software bugs on harness level. Being deep into harness dev has a lot of perks - I have complete understanding and rich stats to judge and optimize efficiency.
https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash
DeepSeek-V4.1-Flash supports a continuously controllable reasoning effort from 1 to 100. All instruct results below use the maximum effort setting (reasoning_effort=100).
I just changed effort in Tau to be configured as a value between 0.0 and 1.0. I can see the future, ha!
trynix.dev boots a Linux machine in WebAssembly and runs arbitrary Nix store paths in a browser tab—over 310,000 historical nixpkgs package versions, including packages that are not in a public cache.
The system combines nixpkgs-multiverse’s package-history index, cached store paths, an in-memory Nix store, a WebAssembly Linux kernel, and a terminal emulator. Because it can consume any cache with permissive CORS headers, even a store path hosted on GitHub Pages can become runnable in the browser. Packages added to a running VM appear immediately on its shell PATH.
The three-comment Lobsters discussion was enthusiastic and suggested using the same machinery for replayable CI failures or building a nixpkgs pull request; one commenter connected that idea to Pernos.co and rr.
Filtered highlights from issue 667:
I skipped the routine merged-PR list, event calendar, and contributor-task sections.
