AI article
I Built Five Self-Improving Loops in One Evening. They All Had the Same Bug.
Community description: Quick disclosure up front, because it matters for how much weight to put on anything below: this was...
Dev.to | Sep 18, 2026 | rello
Automated excerpt
That's what five fast structural tests of one hypothesis actually looks like. All five ran the same structure and all five climbed over 10 rounds: 0. 45 → 0. 61 (stated max: 0. 80) Five for five, on paper. Frozen judge, held-out eval. sia optimizes a genome (versioned JSON: prompt/context/workflow/harness layers) against a fixed 8-task set, split 5 train / 3 holdout.
Selected automatically from source text; not independently written or fact-checked. Read the original for full context.