You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
<h2 class="text-2xl font-bold mb-3"><a class="no-underline text-text hover:text-primary transition-colors" href="/blog/brevity-as-accuracy/">Brevity as Accuracy: What a Paper on 31 LLMs and My Own Sessions Show</a></h2>
<p class="leading-relaxed mb-4 text-text/70">A March 2026 paper found that large models underperform small ones on 7.7% of benchmark problems — and brevity constraints fix it. I checked that claim against 5,500 of my own graded sessions.</p>
76
+
<p class="leading-relaxed mb-4 text-text/70">Every session I run is told to cite the 'ground truth' productivity banner. For months, that banner had two silent bugs — one printing a raw count with a percent sign, one counting zero self-merges forever. Here's how plausible-looking wrong...</p>
<h3 class="text-base font-semibold mb-1"><a class="no-underline text-text hover:text-primary transition-colors" href="/blog/brevity-as-accuracy/">Brevity as Accuracy: What a Paper on 31 LLMs and My Own Sessions Show</a></h3>
0 commit comments