How Hiipal Works: A Complete Step-by-Step Walkthrough

01 Before you start

There is no sign-up screen. When you open hiipal.com, the app assigns you a random display name (you can change it any time by tapping the name in the top corner of the home screen). That name is stored only in your browser — nothing is sent to a server until you actually search, chat, or play.

02 Running your first search, step by step

1. Find the search bar. On desktop it sits in the center chat panel; on mobile it is pinned to the bottom of the screen with a paperclip icon on the left for attachments.
2. Type a phrase — not a single keyword. "quiet cafes for reading" works better than "cafe," because Hiipal is matching meaning, not just words (more on that below).
3. Press enter or tap the arrow icon. Your message posts instantly, and a moment later a result card appears underneath it with a type-count row at the top — something like 3 IMAGES · 4 TEXTS · 1 DOC — telling you what kinds of content matched before you even open them.
4. Page through results with the arrows at the bottom of the card. Each result shows a match-percentage pill so you can see at a glance how relevant it is.
5. To search by image instead of text, tap the paperclip/attachment icon and choose a photo, paste an image directly into the input, or drag one in from another browser tab. The AI captions the image automatically and searches using both the picture itself and its generated caption.
6. To search a document, attach a PDF, Word file, or text file the same way. Hiipal reads the content and matches on meaning, not just filenames.

03 Under the hood: how the AI decides what matches

This part is not covered anywhere else on the site, so here is the short version of what actually happens between you pressing enter and a result appearing.

Every piece of content on Hiipal — a search query, an uploaded photo, a shared document — gets converted into a vector embedding: a list of roughly a thousand numbers that represents its meaning as a point in space. Two pieces of content that mean similar things end up as points that sit close together in that space, even if they do not share a single word in common. That is why searching "puppy" can surface a result whose caption only says "young dog" — the words differ, but the meaning-points are close.

To measure "close," Hiipal computes the cosine similarity between your query's point and every candidate's point — a number from -1 to 1 where 1 means identical meaning. Results below a minimum similarity threshold are discarded, and everything else is ranked by score, which is exactly the percentage you see on each result card.

For text queries, Hiipal also generates a few alternate phrasings of your question behind the scenes (a technique sometimes called query expansion) and searches with all of them in parallel, then merges the ranked lists into one. That is part of why a vague or oddly-worded query can still surface the right result.

04 Finding and messaging a nearby pal

1. Search anything. The instant your query is processed, Hiipal checks who else has searched something semantically close in roughly the last 45 minutes.
2. Watch the pals panel. Matches appear as small cards with a name and a rough similarity score — no refreshing needed, they show up live.
3. Tap a pal's card to open a lightweight profile popup with two options: send a quick one-tap greeting, or open a full chat thread.
4. Inside the chat thread, you can send text, drop an image or voice clip, or tap the phone/video icon in the header to start a peer-to-peer call — the conversation itself never leaves your two browsers.

05 Tips for getting better results

• Describe, do not label. "A cluttered desk with a laptop and coffee" beats "desk" — the AI has more meaning to match against.
• Use well-lit, single-subject photos for image search. Busy or dark images produce weaker captions, which weakens the match.
• If a search comes back empty, rephrase rather than repeat. A different wording of the same idea often crosses the similarity threshold when the original phrasing did not.