Promptic

All 25 shown comments (of 109 total) are from HackerNews, upvoted +546 to +915, and NONE mentions 'Promptic' by name - the conversation orbits the problem space: prompt-as-code, context management, and AI-assisted development. Strongest theme (+915): context is the real bottleneck - 'if you aren't managing your context, you are losing all of the context', with users describing speech-to-text prompts piped into continue.dev, Aider, and Cursor, and one user wanting to build a 'Context Management System'. Second theme is open skepticism of terminology: a +915 comment calls the prompt-vs-context distinction 'puffery to generate new hype'. Success stories exist: a +915 commenter 'vibe coded' a complete OAuth token-expiry test with mocks and stubs in Cursor on a stranger's unfamiliar codebase, concluding 'vibe coding is real and context is the biggest hurdle'. Counterweights: a +546 'stone soup' critique argues much of the value comes from the human asking the right questions, not the tooling ('the interesting part is knowing what questions to ask'); others report ChatGPT prioritizing its training corpus over supplied context (failing on newer tools like vitest), token-limit hallucination after long back-and-forth sessions ('he would have easily hit the token limit... The LLM will start hallucinating'), forced context triage due to context-length limits, and diminishing value on maintenance work ('I spend maybe ~15% of my time working on something new'). Heavy thread drift (C compiler assembly output, RLHF vs Waymo, Rembrandt edge-contrast painting technique) confirms this is broad AI-engineering discourse, not product-specific feedback.

No aggregate ratings were found for this product during the last harvest.