Ask HN, Sept 2025: Builders Share Pragmatic AI, Open‑Source Tools, and Real‑World Projects

Read Articleadded Sep 30, 2025

HN’s September 2025 “What are you working on?” thread showcases hundreds of projects, from AI-powered developer tools and data infra to consumer apps, games, mapping, biotech, and hardware. A clear pattern emerges: pragmatic AI paired with conventional, reliable systems; strong preferences for local-first, open-source, and privacy-preserving designs. Makers actively solicit feedback, testers, and collaborators, and get immediate, practical responses.

Key Points

  • AI is everywhere but pragmatic: builders layer LLMs on top of conventional stacks to automate mundane tasks, summarize, classify, and score—while keeping reliability through traditional infra (databases, queues, search, CRDTs, and schemas).
  • Strong maker trends: local-first, privacy-first, self-hostable, and open-source solutions to avoid lock-in and to invite community collaboration.
  • Developer tooling boom: new data layers for Postgres/TypeScript, observability and log UIs, deployment/orchestration alternatives to heavy platforms, Git-for-data workflows, and testing/PR-like flows for spreadsheets.
  • Diverse domains beyond software: medical robotics, RF sensors, e-bike batteries, clocks, mapping/drone imagery, and niche consumer utilities demonstrate that solo and small teams are tackling real-world problems.
  • Community dynamics: posters actively seek feedback, beta users, contributors, and integrations; commenters provide candid product critiques, performance notes, and feature ideas.

Sentiment

The overall sentiment of the discussion is overwhelmingly positive and collaborative, strongly aligning with the article's observation of a "vibrant indie and OSS ecosystem." Users are enthusiastic about their projects, eager for feedback, and supportive of others. The emphasis on practical AI use as a default ingredient is clearly reflected, with many openly discussing their use of LLMs for coding, content generation, and automation, often with positive outcomes. While some nuanced discussions arise regarding AI's limitations or potential for over-reliance, these do not detract from the general consensus that AI is a powerful and integrated tool in current development.

In Agreement

  • **Pervasive AI Integration**: Many projects demonstrate AI as a "default ingredient," used for specific tasks like summarization, content generation, data structuring, automation, and code generation (e.g., in Laboratory.love, Petmoving.site, LLMRing.ai, DocSpring, Doctly.ai, Vyprsec.ai, X11.Social, Socratify). Numerous developers also report using LLMs to write code, often citing significant productivity gains (e.g., cjflog's "99% of the code" and paul_manias's 25x speed increase).
  • **Vibrant Indie & OSS Ecosystem for Niche Solutions**: The thread is replete with projects from solo builders and small teams addressing very specific problems across diverse domains, reinforcing the article's description of a thriving indie and open-source scene. Many projects emphasize local-first, privacy-preserving, or self-hostable models (e.g., Fallinorg, Colanode, EasyInvoicePDF.com, Prisme Analytics, DailySelfTrack).
  • **Active Community Engagement**: Founders frequently seek feedback, testers, and collaborators, and the community actively responds with suggestions, offers of help, and expressions of interest, demonstrating a strong collaborative spirit.
  • **Diverse Project Categories**: The projects span all categories mentioned in the summary, including AI-native tools, developer infrastructure (e.g., Typegres, Canine, RDX, Berk), hardware/biotech (e.g., robotic surgical aids, stem cell eggs, e-bike batteries), creative apps/games (e.g., voxel engines, puzzle games, retro titles), mapping, and consumer apps (e.g., health journals, financial tools, language learning).

Opposed

  • **Skepticism and Intentional Avoidance of AI in Products**: Despite AI's prevalence, some developers consciously choose not to integrate AI into their final products for reasons like maintaining human control or avoiding AI-generated content (e.g., seanwilson's color palette editor explicitly "no AI or auto generation!", geuis building an audiobook service to "Ignore AI voiced books", and zongheng using AI to build "AI-less tools").
  • **Challenges and Limitations of AI-Assisted Development**: While many praise AI for productivity, others highlight its imperfections and the continued need for human oversight and refinement. Examples include AI generating incorrect code (MangoCoffee), requiring substantial human effort to add depth (maltezev for storylines), or creating "vibe coded apps" that need fixing to be production-ready (jborden13).
  • **Concerns about Over-reliance on LLMs**: A cautionary perspective emerges regarding potential over-dependence on LLM-assisted coding, with one user advocating for "human + AI assistance" as the superior long-term approach compared to fully automated AI agents (yggdrasil_ai). Confidentiality in AI services is also questioned, prompting discussion on realistic guarantees (jacquesm on Dreamstate).
Ask HN, Sept 2025: Builders Share Pragmatic AI, Open‑Source Tools, and Real‑World Projects