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

Hacker News Discussion ↗

Side projects in this Ask HN thread range from indie video games, educational tools and personal productivity apps to hardware hacks, search engines and city dashboards, but a dominant motif is using AI and agents as “co‑developers” or embedded features. Many developers describe leveraging LLMs to scaffold code, orchestrate multi‑agent workflows, or build domain‑specific assistants, while simultaneously worrying about reliability, privacy, lock‑in and the erosion of “human‑only” spaces. Alongside the AI wave, there is a strong undercurrent of local‑first, open‑source, privacy‑preserving and niche, values‑driven products designed for small but highly engaged audiences.

AI agents and infrastructure

  • Many projects focus on building, hosting, and orchestrating agents: coding agents, multi-agent systems, MCP servers, Temporal-based workflows, and agent sandboxes (microVMs, containers, jailed environments).
  • Common goals: reliable “hands-off” code implementation, tool calling, token efficiency, durable long-running workflows, and safe secret handling (runtime security, PoP tokens, vault proxies).
  • Several tools wrap local models (llama.cpp, Apple local models) or expose heterogeneous local GPUs over networks, aiming to reduce dependence on cloud LLMs.
  • Some contributors report strong results from agentic optimization (e.g., code speedups) but anticipate community skepticism about “AI slop.”

Local‑first, privacy, and search

  • Strong thread of local-first and privacy-respecting apps: note-taking, personal CRMs, journaling, baby trackers, fitness dashboards, and system-wide ad blockers.
  • New alternative search engines and search layers (Uruky, Astrolabe for Nextcloud, self-hosted web/file search like Hister, local search tools) emphasize EU hosting, transparency, and no tracking.
  • Several “no-AI” or “human-only” platforms push back on generative content; some commenters suggest optional agent access could aid adoption.

Developer tools, infra, and languages

  • Many devtools: Kubernetes and Docker GUIs, high‑performance logging and regex search, apt caches resilient to upstream outages, git-based deployment systems, CLI utilities, and customizable WMS.
  • New languages and compilers: TypeScript checker in Rust, C package manager “like Cargo,” tensor DSLs, MLIR/Lean/Idris experiments, SAT solvers, and interpreters focused on provable correctness.
  • Multiple attempts to make web dev saner: opinionated stacks, typed DSLs for ML, repo-structure linters, and “browser for agents” or “text editor for agents.”

Games, education, and creative projects

  • Substantial indie game activity: city builders, puzzle games, party platforms, MMO tools, and educational titles for kids (math, chess, programming).
  • Creative tech includes audio languages, generative music tools, plotter art, jewelry based on math/CS, interactive books, and camera/animation tools.
  • Discoverability is a recurring pain point; many builders mention that coding is easier than getting players or users.

Health, wellbeing, and self‑tracking

  • Projects span asthma tracking, calorie/TDEE estimation, tinnitus logging, elder monitoring, stroke-recovery support, fitness and running coaches, sobriety tracking, and “soul cultivation” writing.
  • Some founders seek feedback on failed crowdfunding, questioning whether ideas were weak or just poorly marketed.

Debate and skepticism

  • Notable sub‑discussion around universal basic income in a city‑builder sim: disagreement over likely tax impacts, inflation, housing costs, and whether financial security improves or worsens social outcomes.
  • HR tech based on automated CV scoring draws concern about bias, legality (especially in Europe), and over‑reliance on keyword matching versus experience.
  • Several people explicitly step back from projects to manage burnout, focus on offline skills, or re‑evaluate AI’s role in their work.