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

Hacker News Discussion ↗

AI agents, coding tools & orchestration

  • Many projects focus on making AI coding agents usable in real workflows: guardrails for what agents can touch, DAG-based task planners, Kanban/PR-driven orchestrators, and IDE‑adjacent tools that coordinate multiple agents against a codebase.
  • Shared pain points: context limits, hallucinations, fragile GUIs for computer-use agents, and lack of stable, testable outputs. Several tools add regression tests, coverage-guided input generation, or “ratchet” budgets for code smells.
  • Common patterns: MCP-based skill/package managers, shared memory layers across tools, cost-budget controllers, and hosted runtimes for frameworks like OpenClaw.

Data, infra & devtools

  • Numerous infra and devtools: Postgres-centric backends (auth/permissions/queues in SQL), multi-cloud governance with YAML rules, internal ticket routers for Zendesk, ClickHouse consoles, object-storage movers, and cloud deployment abstractions targeting AI-driven dev loops.
  • Several projects emphasize local-first or self-hosted designs (file sync, error monitoring, observability, config sync, document extraction, Git-like data stores).

Productivity, collaboration & work practices

  • Time-tracking and journaling tools range from AI‑reconstructed workdays to agent‑augmented personal CRMs and weekly retrospectives.
  • One time-tracker sparked concern about corporate surveillance; the creator stressed built‑in privacy (no raw screenshots, review-before-share, opt‑in transparency, relative-time reporting).
  • Multiple task/plan tools target ops-heavy work (large deployments, retros, migrations) with human-centric dashboards rather than pure automation.

Education, reasoning & language

  • Projects for SQL, language learning, and critical thinking include exploratory canvases, daily coding/AI logs, argument-graph systems, word and kanji games, and AI‑assisted behaviour-change agents.
  • Some tools aim to formalize reasoning (graph-based arguments, proof assistants) or turn domain knowledge into reusable teaching flows.

Creative, games & hardware

  • Many hobbyist efforts: Godot/Unity games, realistic sports sims, no‑code 2D engines, beatmaking TUIs, raytracers, CAD tools, 3D keyboards, analog computers, PCB workflows, and custom controllers.
  • Strong interest in “old web”‑style personal projects: solitaire and word games, metaverse experiments, offline audio tools, and home‑lab networking/printing setups.

Privacy, ethics & experimentation

  • Debates appear around self‑experimentation with microplastics (praised as bold, criticized as unsafe/uncontrolled), AI “expert” debate sites (seen as fun vs. misleading), and surveillance‑adjacent tools (DNS, kids’ browsers, monitoring agents).
  • Several builders explicitly foreground encryption, local‑only processing, or minimal data collection as differentiators.