Suspend positive scaling claims; state the negative result and the pre-registered protocol (user doctrine)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn

1 files changed, 19 insertions, 16 deletions

diff --git a/index.html b/index.html
index 5870fb8..b15f65f 100644
--- a/index.html
+++ b/index.html
@@ -252,8 +252,8 @@
             <thead><tr><th>Model</th><th>Data</th><th>EP (val CE)</th><th>BP twin</th><th>Gap</th></tr></thead>
             <tbody>
               <tr><td>72M transformer LM</td><td>FineWeb-Edu, 1.4B tok</td><td>3.33</td><td>3.29</td><td>+4-5% ppl</td></tr>
-              <tr><td>135M transformer LM</td><td>FineWeb-Edu, 2.7B tok</td><td colspan="3">trained end-to-end, zero instability
-                  events; scaling analysis below</td></tr>
+              <tr><td>135M transformer LM</td><td>FineWeb-Edu, 2.7B tok</td><td colspan="3">trains stably end to end; the frozen
+                  72M recipe does not transfer (gap grows to roughly 30% ppl). See the scaling note below.</td></tr>
             </tbody>
           </table>
           <p class="is-size-6 has-text-grey">Twin discipline: identical architecture, tokenizer, data order,
@@ -276,16 +276,18 @@
   <div class="container is-max-desktop">
     <div class="columns is-centered">
       <div class="column is-four-fifths">
-        <h2 class="title is-3">The scaling science</h2>
+        <h2 class="title is-3">The scaling note</h2>
         <div class="content has-text-justified">
-          <p>Scaling a physical learning rule surfaces phenomena backprop never meets. Between widths 512 and 768
-          we identified a width-scaling loss in the EP gradient: localized to the top half of the network,
-          invisible to every per-step alignment metric, and traced to response components that finite nudge
-          displacement under-reaches. We built a screening instrument that measures this leak in 90 minutes per
-          candidate recipe, mapped its dose-response law (logarithmic across two decades of displacement
-          amplification), and demonstrated a pure estimator-side treatment that closes 97% of it: no change to
-          the model, its inference path, or the cost budget. The same instruments provide the go/no-go protocol
-          for each next rung of the ladder.</p>
+          <p>Scaling the frozen 72M recipe one width step (512 to 768) fails: the EP to BP gap grows from
+          4-5% to roughly 30% perplexity. We treat this as the central result of the current phase and the
+          central research object. It has sharp structure: the loss is fully localized to the top half of the
+          network, is invisible to every per-step alignment metric, is untouched by estimator symmetrization,
+          precision, relaxation depth, and output-head interventions, and is removed completely by
+          substituting true gradients in the top half. A mechanism-identification campaign across four widths
+          is in progress, with a pre-registered transfer law and a blind holdout at the next width. Positive
+          scaling claims are suspended until the law predicts an unseen width with zero tuning.</p>
+        </div>
+        <div class="content" style="display:none">
         </div>
         <h2 class="title is-3">Hardware line</h2>
         <div class="content">
@@ -301,11 +303,12 @@
         </div>
         <h2 class="title is-3">Roadmap</h2>
         <div class="content has-text-justified">
-          <p>Staged scaling with matched BP controls and hardware-relevant ablations at every rung: a 150M-600M
-          ladder (the gap is measured against a backprop twin at every rung), then 1B-3B; each stage
-          gated on the previous stage&rsquo;s loss, alignment, and throughput numbers. In parallel: the
-          algorithm&rarr;regime map across the activity-difference family (contrastive / coupled-learning arms on
-          the same harness), and a bounded single-column hardware feasibility study.</p>
+          <p>Mechanism first: identify the width-scaling term across four measured widths, freeze a transfer
+          law with its thresholds and budgets, then validate blind at 270M with from-scratch runs and two
+          seeds. The scaling ladder resumes only on a zero-tuning hit at the holdout width. In parallel: the
+          algorithm-to-regime map across the activity-difference family, and a bounded single-column hardware
+          feasibility study.</p>
+        </div>
         </div>
       </div>
     </div>