Phind 2: AI search with visual answers and multi-step reasoning

Original Article ↗ Hacker News Discussion ↗

Overall reception & positioning

  • Many commenters are impressed; several paying users say Phind now rivals or beats Perplexity, ChatGPT, and mainstream search for programming and some research/finance tasks.
  • Others still find Perplexity or ChatGPT better for certain queries (e.g., product search, playful/creative responses, vague memes).
  • Some feel Phind disappeared from view for a while and needs more visible presence to stay top-of-mind.

Visual answers, diagrams & UX

  • Strong enthusiasm for on‑the‑fly diagrams (via Mermaid), flowcharts, and image-rich explanations; people highlight them as the standout value vs. competitors.
  • Others find diagrams verbose, distracting, or slow to render, especially when they restate a simple question or clutter code answers. They want an easy per-query toggle or a “plain text only” profile that doesn’t require sign-up.
  • UI is widely praised: side-panel sources, rich layout, “article-like” responses, and tree-structured conversations with “zoom into a part of the answer” follow‑ups. Some want denser layouts to see more content at once.

Developer & power‑user use cases

  • Heavy use for coding help, SQL query construction, complex API setups, architecture diagrams, LangChain examples, and IDE integration.
  • Some miss the deprecated VS Code extension; others want APIs and broader IDE support (e.g., IntelliJ, Continue plugin).
  • URL ingestion sometimes fails (e.g., resume + job posting), which users expect a “search engine LLM” to handle reliably.

Model behavior, quality & trust

  • Reports of excellent, fast answers on technical topics (tax AMT/LTCG, specific stocks, programming).
  • Other queries show glaring errors: hallucinated economic calendars, wrong claims about GameStop in Canada, confusion over “should I?” vs. cited sources, and weaker handling of internet memes.
  • Several users note models are overly agreeable and not sufficiently grounded in the actual retrieved sources.

Business model, access & privacy

  • Pricing criticized for forcing a $20/month subscription just to try premium models; many request usage-based or short paid trials.
  • Free tier limits and countdown behavior are described as buggy/confusing.
  • Concerns that “new threads are public” by default and that training on user data is opt‑out; some only want anonymous, no-account use.
  • Region-based unavailability frustrates some.

Web ecosystem, copyright & infrastructure issues

  • Strong debate over Phind displaying third-party images in answers: critics see copyright violation and reduced incentives for creators; suggestions include using only CC-licensed media or revenue-sharing schemes.
  • Defenders compare this to Google Images and invoke fair use, but others note prior lawsuits against image search engines.
  • A serious bug is reported: shared conversation links behave like mutable sessions—anyone can edit the content and it persists for others—creating confusion and potential security/abuse concerns.