Daily issue / 2026-05-17
Useful things that surfaced in Hacker News discussions on 2026-05-17. Open the context to see why each one caught our attention—or jump straight to the original thread.
What we found
Code search for agents that cuts grep-heavy token usage.
Source discussion: Show HN: Semble – Code search for agents that uses 98% fewer tokens than grep
Editorial paraphrase
The thread highlighted CPU indexing, zero-key setup, and the claim that it beats grep+read on token use and retrieval quality.
Read the original threadAn $80 Android tablet converted into a usable Debian workstation.
Source discussion: I turned a $80 RK3562 Android tablet into a Debian Linux workstation
Editorial paraphrase
One commenter said Debian with most devices functional was great; others discussed what runs in 4 GB RAM and suggested lightweight stacks.
Read the original threadCross-platform terminal emulator suggested for a low-RAM tablet setup.
Source discussion: I turned a $80 RK3562 Android tablet into a Debian Linux workstation
Editorial paraphrase
A commenter suggested running WezTerm + tmux on 4 GB RAM instead of a heavier desktop environment.
Read the original threadTerminal multiplexer recommended for the tablet workstation stack.
Source discussion: I turned a $80 RK3562 Android tablet into a Debian Linux workstation
Editorial paraphrase
The comment recommending WezTerm + tmux explicitly tied it to fitting development tools into 4 GB of RAM.
Read the original threadWayland compositor praised as responsive on small ARM hardware.
Source discussion: I turned a $80 RK3562 Android tablet into a Debian Linux workstation
Editorial paraphrase
The same tablet discussion said sway runs well on a PinePhone Pro and feels responsive enough to replace a heavier desktop.
Read the original threadVersion control system still remembered as pleasant and user-friendly.
Source discussion: Mercurial, 20 years and counting: how are we still alive and kicking? [video]
Editorial paraphrase
One commenter said Mercurial was more user-friendly while still offering the needed features and performance; another still uses it for personal projects.
Read the original threadCurated CUDA reading list with book-by-book recommendations.
Source discussion: CUDA Books
Editorial paraphrase
A commenter recommended CUDA Programming: A Developer’s Guide to Parallel Computing with GPUs and criticized some other intro books as too shallow or error-prone.
Read the original threadRust Markdown linter that commenters said is faster than rumdl.
Source discussion: Mado: Fast Markdown linter written in Rust
Editorial paraphrase
The author said Mado is older and faster per rumdl’s benchmark graph, while also noting that Mado stays focused on linting.
Read the original threadLocal LLM runtime used on a Mac Studio over tailnet.
Source discussion: Apple Silicon costs more than OpenRouter
Editorial paraphrase
One comment said they can run Ollama on a Mac Studio over the tailnet “for free,” which anchored the local-inference discussion.
Read the original threadRetro meter-style clock build with strong maker appeal.
Source discussion: A nicer voltmeter clock
Editorial paraphrase
One commenter said they always enjoy projects like this because they are artistically neat and spark ideas; another shared previous voltmeter-clock references.
Read the original threadTiny web server project running on an 8-bit MCU.
Source discussion: Hosting a website on an 8-bit microcontroller
Editorial paraphrase
Commenters pointed to earlier smallest-web-server projects and said the ACE1101 version was great fun to reproduce.
Read the original thread