From b518fdcf9dc9f03322c11546f053ddd7b0268f09 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ahmet=20Bar=C4=B1=C5=9F=20G=C3=BCnayd=C4=B1n?=
 <abgunaydin94@gmail.com>
Date: Mon, 3 Aug 2026 17:49:17 +0700
Subject: [PATCH] fix(landing): say that the body is moved directly, not
 carried by its legs
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The landing page said the connectome's output is a walking speed and a turn
and that the stepping rhythm is a hand-written tripod gait. Both true. But it
never mentioned the kinematic assist — the word appears nowhere on the page —
which leaves a reader to assume the legs still produce the motion. They do
not: in the default view the drive scalar is written straight into the body's
velocity, and the legs animate alongside.

Measured, and this is what makes it worth saying: with the assist on the body
covers ~0.80 cm per simulated second regardless of what is in control — the
CPG, the trained policy, or nothing at all. One run translated at the
commanded speed while the fly was upside down. With the assist off the
connectome-driven fly pivots on the spot: 3.43 cm of path for 0.17 cm of net
travel, -245 degrees of yaw.

So "control the fly with its brain" stays — the cascade genuinely decides
where it goes, and every link in that chain is real. What was overstated is
"walks". The page now says the body is moved by the command rather than
carried by its legs, that a switch turns the shortcut off, and what happens
when it is off.

Also softened the og:description, which said "watch the fly walk" with no
qualifier — that is the text that appears in link previews.

Found by the repo's author reading the page, after four hostile expert
reviews missed it: they concentrated on the brain-to-spine wording and on the
trained walker, and nobody checked whether the landing page disclosed the
assist at all.

e2e 34/34 (two browser-crash failures from concurrent memory pressure re-run
clean on an idle machine).

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
---
 index.html | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/index.html b/index.html
index ed2afd2..b1009fe 100644
--- a/index.html
+++ b/index.html
@@ -11,7 +11,7 @@
     <meta property="og:title" content="webgpu-fly — a real fly brain you can play" />
     <meta
       property="og:description"
-      content="A real fruit-fly brain, spine, and body running live in a browser tab. Press a key, fire real neurons, watch the fly walk. No install."
+      content="A real fruit-fly brain, spine, and body running live in a browser tab. Press a key, fire real neurons, and the cascade decides where the fly goes. The connectome simulation is real; the walking on top of it is an approximation we document. No install."
     />
     <meta name="twitter:card" content="summary_large_image" />
     <meta name="author" content="Ahmet Barış Günaydın" />
@@ -446,6 +446,15 @@ <h2 class="title reveal">A whole animal nervous system, end to end, running on y
           itself is a hand-written tripod gait, not neuron-by-neuron muscle
           control.
         </p>
+        <p class="body reveal">
+          And in this default view the body is <em>moved</em> by that speed and
+          turn directly, rather than being pushed along by its own legs. So the
+          brain really does decide where the fly goes; the legs you see stepping
+          are not what carries it there. There is a switch that turns the
+          shortcut off — with it off, the connectome-driven fly pivots on the
+          spot and covers almost no ground. The one thing that does walk on its
+          legs is the trained policy below, and even that leans on a stabiliser.
+        </p>
       </section>
 
       <!-- what's inside -->