Ask HN: What Are You Working On? (June 2025)

Hacker News Discussion ↗

AI, Agents & Automation

  • Many are building AI-powered decision or planning tools: spending guides, A/B test planners, stock/crypto backtesting, phone-insurance risk sharing, and “ChatGPT for email” search across inboxes.
  • Several “AI for X” assistants: customer support, legal drafting, accounting alerts, LLM observability/monitoring, AI sales co-pilots, AI debugging/IDE integrations, and no‑backend “AI app platforms.”
  • Multiple agentic projects: browser automation via Chrome extensions and MCP, local “personal agents” on small models, AI interview simulators, coding copilots for niche stacks (Flutter, n8n, etc.).
  • Skepticism appears around AI substitution vs. tooling (eg. some resist AI-heavy learning tools, others worry about AI hallucinations in safety‑critical domains like pill-checking or tenancy agreements).

Developer Tools, Infra & Data

  • Strong cluster around dev tooling: uptime monitoring + Terraform, CLI utilities, IPC/RPC layers in extensions, Rust debuggers, deterministic runtimes, improved tokenizers, code history “event sourcing,” and multi-language IoT clouds.
  • Several storage/infra experiments: time-series DBs, Parquet compactors, Zig-based editors, F# source generators, static site platforms, custom object stores, and HTTP/spec conformance test suites.
  • Data-heavy projects: stock forums, ETF prospectus analyzers, QuickBooks add-ons, GDELT political event analysis, healthcare price transparency catalogs, and full‑history arbitrage or arbitrage‑like trackers.

Consumer, Productivity & Finance Apps

  • Many personal finance tools: budgeting apps (local-first, spreadsheet replacements, rule-based), net‑worth trackers, invoice generators, and banking‑style dashboards.
  • Numerous small productivity apps: task managers, semantic desktop search, mobile IDEs, email aliasing/hide-my-email clones, note/PKM tools, and Chrome extensions for learning jq, Slack export grooming, or MCP tracing.
  • Some projects aim to reshape notifications and scheduling (AI schedulers, weather SMS alerts, crisis‑resilient backups, cron-as-a-service).

Learning, Language & Education

  • Language-learning tools are prominent: parallel reading bots, sentence-click translation→flashcards, accent trainers, speaking SRS with LLM grading, journaling‑as‑language‑practice, and curriculum/AI course generators.
  • Education content platforms: daily puzzle sites, math apps for kids, live‑coded music tools, and podcast-topic visualizers. Debate around how AI fits into serious learning vs. books and formal curricula.

Games, Media & Creative Tools

  • Many game/engine projects: voxel engines, NES rhythm games, roguelike experiments in multiple languages, mixed-reality horror ports, and MMO/engine tech “labors of love” spanning a decade.
  • Creative tooling: AI sprite animators, font generators, interior design imagers, time‑lapse photography apps, diagram editors, cable‑harness CAD, and indie photography suites.
  • Content experiments: AI podcasts reading HN, audience‑driven GenAI rom‑coms, mail‑based Cold War history subscriptions, and niche movie/podcast discovery tools.

Hardware, Embedded & “Real World”

  • Hardware/embedded work includes repairable e‑bike batteries, CNC-built injection molds, drone OS frameworks, sleep neurostimulation devices, Nest thermostat replacement boards, digital pens that sync handwriting, tide clocks, and FPGA/PhysX integrations.
  • Several homelab and self‑hosting efforts: custom media servers, static hosting platforms via SSH, retro emulators, and on‑prem “AI in a box.”

Security, Privacy & Safety

  • Projects target email privacy, anti‑spam filters (eg. young domains, newsletter aliases), Slack scraping MCPs, steganographic containers, and machine‑friendly product security policies.
  • Debate emerges where harm reduction meets moralizing (eg. pill‑scanner for MDMA pills criticized as “don’t take random pills,” countered as pragmatic harm reduction).

Meta: Side‑Projects, LLM‑Driven Dev & Motivation

  • Many describe “vibe coding” with Cursor/Claude, using LLMs as accelerants while still wrestling with architecture, testing and maintenance. Some find it liberating; others find it depressing or chaotic.
  • Recurring themes: burnout from repeated startup failure, joy in long‑running “Moby‑Dick” projects, the struggle to find ideas or open‑source projects to contribute to, and a number of people frankly saying they’re working on “nothing” — and enjoying the break.