F3 · as chosen tangent contact
5 logs · aspect 0.97 · touches at one point
F3, in context
A mark is almost never seen alone. The "IO" reading is strongest in isolation and weakest anywhere the word is nearby — a tab title, a lockup, a bar you have already scrolled past twice. This is the test that decides it, not the 104px hero above.
alone
worst case
worst case
no word anywhere · this is where it reads as I◇
browser tab
16px + title
16px + title
PhotoPool — every camera in the room
in the bar
replacing the wordmark
replacing the wordmark
lockup
mark + wordmark
mark + wordmark
F3, tuned
F3's diamond meets the stem at a single point — a tangent, which can read as resting against rather than attached to. Two ways to firm that up without adding the full shoulder bar that would turn it back into F5.
F3a · Bitten diamond set into the stem
The diamond moves 3 units left so its point is buried in the stem rather than kissing it. Reads as one piece of geometry instead of two objects in contact.
5 logs · aspect 0.89 · overlapped
F3b · Stub a short shoulder, not a full bar
Eight units of shoulder instead of sixteen. Enough to say P, short enough that the diamond still dominates — the midpoint between F3 and F5.
6 logs · aspect 0.97 · stub shoulder
internal — not linked, not indexed · round 2
Letterform play
Round 1 only subsetted letters, which gives you two letters standing next to each other and never one glyph. These four each make a specific change to a letterform, and the change is the thing doing the work.
Fixing the "IO" read
F2 reads as IO, and the diagnosis is specific: it dropped the P's bowl, and a bare vertical stem is an I. A P's identity lives in the bowl, not the stem. All three below put a bowl back while keeping what F2 got right — the diamond fused rather than adjacent, and a compact aspect.
F5 · Shoulder one bar restores the letter
A single horizontal log across the top. That bar is the P's shoulder, and it is the whole fix: a stem with a shoulder cannot be read as an I, and the diamond now hangs beneath it as the bowl rather than floating beside it.
6 logs · aspect 0.92 · cheapest fix
F6 · Closed shoulder and waist, diamond as terminal
Both of the P's horizontals come back, so the bowl is genuinely closed and the letter is unambiguous. The diamond shrinks to a solid and becomes the bowl's outer terminal — the aperture is now a detail inside a real P.
5 logs · aspect 0.89 · solid diamond
F7 · Crossed full P, diamond over it, seam between
The whole P survives, unmistakable, and the diamond crosses it. Where they overlap the diamond carries a gap in the ground colour — which is the site's own seam, doing the job it does everywhere else: marking where two separate things meet.
Cost: the gap is painted in the ground colour, so this one has to know what it is sitting on. Fine on the site, awkward on a photo or a stranger's slide. The other two work anywhere.
8 logs · aspect 1.28 · knockout seam
F, refined — earlier pass
F is chosen. Two things are wrong with the sketch: it is 1.72 aspect, which letterboxes in any square slot, and the P and the diamond are two objects with a gap between them — they sit near each other rather than fusing. In the reference case the two letters nearly touch, and that contact is what makes them one glyph. Each variant below closes one or both.
F1 · As sketched baseline
8 logs · aspect 1.72 · 1-unit gap
F2 · Bowl the diamond IS the P's bowl
The diamond moves onto the stem until its left point meets it. Not a P next to an aperture — a P whose bowl is one.
5 logs · aspect 1.14 · touching
F3 · Bowl + tail raised, so the stem descends past it
The same fusion with the diamond set high, so the stem runs on beneath it. That descender is what keeps it reading as a letter rather than a symbol on a stick — and it squares the whole thing up.
5 logs · aspect 0.97 · touching
F4 · Held the diamond inside a squared bowl
G's closed frame with F's diamond set inside it. The aperture is held by the frame rather than sitting beside it, so the two ideas stop competing: a frame, and something in it.
5 logs · aspect 0.83 · solid diamond
Round 2, for reference
E · Rotor P + P rotated 180°, sharing a stem
The second P is turned upside down and pushed into the first until their stems land in the same column. The join is exact: the first P's short stem runs y0–18 and the second's runs y18–36, so they add up to one continuous full-height stroke.
Why the rotation earns itself: the result has exact 2-fold rotational symmetry — turn it upside down and it is identical. That is what stops it reading as "PP" and starts it reading as one object.
8 logs · PhotoPool · 180° symmetry
F · Aperture Po, with the o turned 45°
The closest relative of the reference case: one letter is rotated so the pair fuses. The o becomes a diamond, which stops being a letter and starts being an aperture — and "Po" is short, sayable, and nobody else's.
Watch: it is the most directly borrowed move of the four. The tilt has to be load-bearing here — a diamond that reads as an aperture — or it is just someone else's trick wearing our stroke.
8 logs · Photo · 45° rotation
G · Shutter the P's bowl closed into a square
One letter, one change: the P's bowl is squared off and closed into a true frame, and the stem runs past it. It stops being a letter with a hole and becomes a frame on a post — the sign on the table with the code on it.
Cheapest change, biggest shift in meaning. The counter is a true 18×18 square, so the hole in the letter is literally a frame. It is also the only one here that stays a single readable letter.
4 logs · PhotoPool · squared counter
H · Skew oo, the second frame dropped and tilted
The double-o of pool, but the second frame is knocked out of true and overlapped with the first. Two photographs put down on a table, one askew. The tilt does the same job as the rotation in E: it stops the pair being two of a thing and makes it one arrangement.
Risk: overlapping strokes at small sizes fill in. The 14° is deliberately shallow — past about 20° it stops reading as "placed down" and starts reading as "falling over".
8 logs · pool · −14° tilt, overlapped
Round 1, rejected: PP / o / oo / oo-stacked. All four subsetted the lockup without altering a single letterform, which is exactly why none of them cohered — they stayed letters sitting next to each other. Kept in git history rather than on the page.