S2Agentic head-to-head — terse-brain vs Karpathy llm-wiki
The stronger claim isn't token counts — it's what happens when an agent has to work the state. We ran the same knowledge-base workload through terse-brain and through a faithful out-of-the-box Karpathy llm-wiki (markdown) setup, and measured everything.
Setup
| Parameter | Value |
| Arms | T = terse-brain · K = Karpathy llm-wiki (markdown), out of the box |
| Workload | 15 documents ingested · 24-question schedule · 3 reps = 72 graded answers per arm |
| Model | Same model for both arms; graded by an independent LLM judge (Grok 4.5) |
| Question tiers | T1 direct recall → T4 abstention & contradiction handling |
Accuracy — parity
| Arm | Pass | Rate |
| T — terse-brain | 66 / 72 | 91.7% |
| K — Karpathy wiki | 67 / 72 | 93.1% |
One question of difference at n = 72 — statistical parity. Both arms shared the same weak spots (one authored-docs tier question and one abstention case where a regex check failed an answer the judge liked). The point is not that TERSE answers better; it's that it answers as well for a fraction of the machinery below.
Efficiency — where TERSE wins
| Metric (mean per rep) | terse-brain | Karpathy wiki | Advantage |
| Input tokens | ~1.81M | ~5.17M | 2.9× |
| Output tokens | ~59k | ~409k | 6.9× |
| Tool calls | ~109 | ~709 | 6.5× |
| Wall time | ~23 min | ~80 min | 3.5× |
| Est. API cost / rep | ~$2.10 | ~$6.09 | 2.9× |
| Est. cost, 3-rep total | ~$6.31 | ~$18.27 | 2.9× |
Why the gap: markdown can't enumerate or query itself, so the K arm greps, re-reads, and rewrites whole files — 709 tool calls of it. The T arm asks ? [DEPTH 2; CONTAINERS] for its index and patches by declaration, so most of the bookkeeping simply never happens.
Caveats — read before quoting
| # | Caveat |
| 1 | n = 72 questions per arm; single workload at personal-knowledge-base scale (~15 sources). Not a claim about terabyte corpora. |
| 2 | One model family per run; an LLM judge (Grok 4.5) graded answers. Judge choice can shift absolute pass rates. |
| 3 | Costs estimated at $2/M fresh input, $0.30/M cached, $6/M output. Your rates will differ. |
| 4 | TERSE 0.2.8 is pre-1.0; the harness and results (report.md / report.json) live in the repo so you can rerun as the language evolves. |
Reproduce it: the benchmark harness ships in the
terse-lang/terse repo — same docs, same question schedule, your model and judge of choice.