yarlinghe

Happy new year from China !!

One thing I’ve noticed building a small “use it and leave” tool is that the models tend to self-select into roles.

For me, Codex ended up being the reliable executor for actual logic and glue code, while Claude Code (especially the frontend design skill) was much better at overall UI structure, spacing, and interaction flow. I stopped trying to make one model do everything and just let them do what they seemed naturally good at.

This only worked because the project was tiny and mostly stateless — once scope or dependencies grow, the coordination cost goes up fast. Curious if others have seen a similar split in practice, or if this is just project-type dependent.

24h update after the Show HN — quick recap.

Based on the feedback here, I spent the last day tightening a few core things:

- Navigation / mobile: tapping a card now opens the symbol page directly; copy is a separate, larger button. The tiny expand target is gone.

- Search: alias + lightweight fuzzy matching is in (e.g. “section”, “angstrom”, “greater than”), plus a /codepoint/... fallback so pasting a literal character or U+XXXX always lands on a usable page, even outside the curated index.

- Missing glyphs: added a small “Font tips (if you see □)” section with system font notes and Noto suggestions.

- Content: HTML/CSS/JS guidance now recommends writing the literal character + UTF-8 by default; entities/codes are framed as legacy/reference only. Added Notes and improved Related (pairs, confusables, same block).

- Privacy: removed session recording; analytics are opt-in with a real reject-all (everything works fine with tracking off).

Direction going forward: fewer generic “how-to” instructions, more symbol-specific context — confusables, paired glyphs, rendering quirks, and references people actually look for.

Thanks again for the detailed feedback — it materially improved the project.

Great example — thanks for the context. I added explicit linking for paired glyph parts (left/right halves) and reference links in Notes. This kind of symbol history is exactly the gap I want to fill.

Good point. I added a Notes section and reworked “Related” to focus on things people actually get confused by (pairs, confusables, same block), not just category adjacency.

Agreed. Updated the HTML section to recommend writing the literal character + UTF-8 by default. Numeric refs/entities are now explicitly framed as legacy / edge-case references.

You’re right — the main index is curated. I added a /codepoint/… fallback so pasting a literal character or U+XXXX always lands you on a dev-friendly encoding page, even outside the curated sets. I also clarified this in the UI copy.

Added a small “Font tips (if you see □)” section with system font notes and Noto suggestions. Also added a codepoint fallback page so you can still inspect encodings even when the glyph doesn’t render locally.

Agreed — the small expand target was a mistake, especially on mobile. I switched to “tap card = open details” and moved copy into a dedicated, larger button. Much clearer now.

Noted.

Fixed.

You’re right.

The abbreviations (arrows in particular) made scanning worse than it should be. I’ve pushed a quick update: full names on hover/focus, plus a Compact / Readable toggle for the grid.

Thanks for the nudge.

Yep — fully agree.

For modern HTML/CSS/JS, you should just write the character and serve UTF-8. The entities / codes are there purely as reference for legacy cases, debugging, or when you only have a code point and no rendered glyph — not as a recommendation for normal authoring.

Yeah, mobile is rougher than desktop right now. Thanks for flagging it.

Good points. The hover hint could definitely do more useful work, and the repeated “Click to Copy” is probably overexplaining. UI polish still very much in progress

Purely cosmetic. In hindsight, probably unnecessary — but once you add color, it’s hard to stop

Fair point — you’re right to call this out. I’ve now removed Microsoft Clarity (no session recording/heatmaps), and GA4 is opt‑in with a real “Reject all” option (default off). This is already deployed.

Thanks — glad the copy buttons are useful. That was exactly the goal.

Good catch — I just pushed an update based on this.

The symbols search now supports code-point lookup, so you can search by:

U+2FFF

0x2FFF

plain hex (2fff, 4–6 digits)

This makes it possible to jump straight to a symbol page even when the glyph doesn’t render locally and you only have the code point (like the Emacs/BSD case you described).

One limitation to note: some symbols aren’t covered by common or default system fonts, so they may still appear as tofu boxes depending on the font stack. That’s a font coverage issue rather than Unicode itself.

Appreciate you calling this out — this was exactly the kind of workflow gap I wanted to catch.

Planning to add more symbol-specific notes (confusables, common pitfalls, rendering quirks). Curious what details people usually look for but can’t find.