Ask HN: What are you working on? (May 2026)

Hacker News Discussion ↗

AI & Agent Tooling

  • Many projects center on AI agents and harnesses: coding agents, multi‑agent orchestrators, task runners, context managers, model routers, and safety/permissions layers.
  • Common goals: reduce babysitting, add durable state, sandboxed execution (often via VMs or NixOS), and better guardrails for destructive tools.
  • Several tools focus on structured outputs, memory, local‑first inference, or cross‑provider switching; others try to make agents safer via deterministic policies instead of expensive LLM “auto review.”

Developer Tools & Infrastructure

  • New CLIs, LSPs, and libraries: TypeScript checkers in Rust, Go retry/backoff libs, repo structure linters, Git caching compilers, deterministic .NET runtimes, SQL canvases, and alternative CI/build/deploy systems.
  • Infrastructure projects include apt‑cache replacements robust to upstream outages, Kubernetes‑based agent runtimes, high‑throughput Bitcoin transaction distribution, and image registries for disk images.
  • Some work targets niche but painful areas: RADIUS re‑implementations, database auditing, TLS automation, and Kafka protocol mockers.

Games & Creative Media

  • Many indie games in progress: NES and browser games, RPGs, idle clickers, tactical sims, puzzle games, and narrative/AI‑driven TTRPG experiences.
  • Several devs emphasize deterministic engines, custom physics, and tooling (e.g., mocap pipelines, VFX editors, music languages).
  • Enthusiasm is high; concern appears mainly around content quality (e.g., AI‑generated puzzles and stories often still feel weak vs. human design).

Productivity, Notes & Knowledge Tools

  • New note‑taking, spaced‑repetition, journaling, task, and calendar apps—often local‑first, Markdown‑based, and sync‑light.
  • Some explicitly target AI‑era workflows: integrating with coding agents, generating specs from MD files, or surfacing code/PR context.
  • There’s recurring interest in replacing heavier SaaS (Miro, Notion, Anki, etc.) with simpler, privacy‑respecting tools.

Hardware, Embedded & “Real World” Projects

  • Projects span barbell motion sensors, ESP‑based automation, mocap on a budget, holographic displays, GNSS‑guided lawnmowers, 3D printers, and e‑bike batteries.
  • One thread around lifting IMU metrics shows disagreement: some argue bar path is key to safety; others say injury risk is more nuanced and “velocity” is more useful.

Privacy, Self‑Hosting & Alternatives

  • Multiple efforts to build EU‑hosted or privacy‑centric replacements for search, PagerDuty, AirDrop, email marketing, analytics, and AI chat clients.
  • Skepticism appears around products requiring Google sign‑in just to view a landing page or around heavy browser tracking (e.g., Firefox bookmark metadata).

Learning, Writing & Personal Experiments

  • Many are learning formal methods, compiler construction, OS dev, DSP, woodworking, FreeBSD, and languages by building real tools.
  • Others focus on writing blogs, books, or sci‑fi, often reflecting on AI’s role (from “AI slop” suspicion to heavy everyday use).