diff --git a/README.md b/README.md index faba763..5bfba32 100644 --- a/README.md +++ b/README.md @@ -73,7 +73,15 @@ You know exactly why. ## Numbers -5 coding tasks, same agent with and without ponytail: **−16% tokens, ~4× faster, 293 → 47 lines.** The 246 lines nobody wrote have never caused an incident. Data: [benchmarks/](benchmarks/). +5 coding tasks, same agent with and without ponytail: **−16% tokens, ~4× faster, 293 → 47 lines.** The 246 lines nobody wrote have never caused an incident. + +Then six harder tasks — streaming parsers, atomic file sync, auth, a concurrent money ledger — against a no-skill control and the [caveman](https://github.com/JuliusBrussee/caveman) skill, with adversarial security and concurrency probes: + +

+ Lines of code per task: ponytail 490 total vs caveman 1,440 vs control 2,943 — all passing the same adversarial probes +

+ +Everyone passes every probe. Ponytail does it with a third of caveman's code, a sixth of control's, and was ~5× cheaper when the surprise feature request landed (96 lines changed vs 487 for control). Every shortcut is marked in-code with its upgrade path. Data: [benchmarks/](benchmarks/). ## License diff --git a/assets/benchmark-loc.svg b/assets/benchmark-loc.svg new file mode 100644 index 0000000..5da8e8b --- /dev/null +++ b/assets/benchmark-loc.svg @@ -0,0 +1,69 @@ + + Non-blank lines of code per task — control vs caveman vs ponytail + + Six tasks. Adversarial probes: everyone passes. Lines of code: not everyone. + + + + Control (no skill) · 2,943 total + + Caveman · 1,440 + + Ponytail · 490 + + + log-analysis CLI + + 970 + + 283 + + 145 + + + file sync + + 587 + + 228 + + 99 + + + notification dispatcher + + 726 + + 396 + + 73 + + + validation engine + + 343 + + 218 + + 70 + + + auth module + + 155 + + 148 + + 49 + + + money ledger + + 162 + + 167 + + 54 + + Non-blank LOC, AST-counted. Every arm passes the same security (8/8) and concurrency (6/6) probes. 2026-06-12. +