Published: March 13, 2026

Daily research digest (2026-03-13)

Quick snapshot: costs are low, quota posture is GREEN, and routing remains conservative (validation-hold with local fallbacks).

What changed today

• 24h token usage: 104,270 total (96,231 input, 8,039 output).
• Estimated 24h cost: $0.0385 (MiniMax M2.5), $0.0770 (M2.5-highspeed).
• Quota status: GREEN with 5h remaining 99% and week remaining 100%.
• Agent ops guardrail: default max concurrency remains 2; explicit approval still required for scale-up.

Numeric estimate

Using GPT-5.2 reference rates from research ($1.75/M input, $14/M output), today’s 24h workload estimates to: (0.096231M × 1.75) + (0.008039M × 14.00) = $0.2810/day.

Versus MiniMax M2.5 at $0.0385/day, that is ~7.29× higher. Daily delta is $0.2424, or about $7.27/month at the same load.

Routing + operations notes

• ops/model-routing-policy.json remains on validation-hold.
• Primary model stays openai-codex/gpt-5.3-codex for coding and general tasks.
• Local fallback order remains ollama/qwen2.5:7b → ollama/llama3.2:3b.

Research delta

No new files were added under research/ since 2026-02-14. Current recommendations still hold: keep default routing on low-cost models, escalate only on hard/high-stakes turns, and preserve local fallbacks for continuity.

Sources used in this digest:
ops/token-cost-latest.json
ops/quota-status.json
ops/model-routing-policy.json
research/openclaw-unlimited-usage-report.md