Yesterday a full workday of agent tasks — reading, summarizing, triaging piles of material — cost R$0 in API tokens. Not a promo credit, not a free-tier trick. One architectural decision: routing.
The common mistake is using AI like a single brilliant employee: throw everything at the most impressive model. But roughly 80% of an agent workday's volume isn't hard. Reading, summarizing, classifying, extracting fields, formatting. Paying frontier-model prices for mechanical work is hiring a surgeon to change a lightbulb.
The ladder
Give each task the lowest rung that can actually do it:
- Rung 1 — raw volume (harvesting, extraction, mass classification): small local model on your own machine. Token cost: zero.
- Rung 2 — reading and synthesis (long material, comparing sources): mid-size local model. Still zero.
- Rung 3 — hard decisions (final verdicts, judgment-heavy analysis, critical code): the frontier API, and only there. You pay for the rung, not the ladder.
Who picks the rung? A simple classification rule is enough to start — or the local model labels the task itself before executing it.
What a R$0 day actually is
A zero-token day is a volume day: everything landed on rungs 1 and 2. The local model ate the material, agents logged everything to files, results came back ready for review. Honest energy number: under R$1 per long cycle on typical hardware — premises vary with machine and tariff. R$0 in tokens isn't R$0 absolutely; it's trading per-request cost for the fixed cost of a machine you already own.
The two warnings gurus skip
- Cheap that redoes the work three times is expensive. If the small model fails and retries, your cost per completed task climbs — price-per-token is the wrong metric to decide with. Measure cost per completed task, always.
- Routing quality down without a reviewer manufactures error in silence. Every step down the ladder needs quality sampling: periodically, a human or a second agent checks a sample of what the cheap model produced. And some things never go to rung 1 — final decisions, anything published under your name, anything you can't verify afterward.
Three rules to keep
- Mechanical and voluminous → local model.
- Hard decision → frontier API, only there. It's worth paying, because it's the price of being wrong less.
- Always measure cost per completed task, never per token.
Full disclosure: this is part of the working method behind a US$15 one-time template I sell — a second-brain vault starter with the agent rules, derived-index and search tools included, plus a short ebook on what NOT to do. The routing ladder works fine with nothing but your own setup.
Second Brain Starter — US$15 one-time
Questions about where routing breaks in practice are welcome — that's the interesting part.
Cross-posted from Oroboro Labs — this post is the canonical version; the original carries the full method notes.
Top comments (1)
The cost-per-completed-task metric is the right lens here. In practice, the primary failure mode of a tiered routing ladder is rarely a clean crash that triggers an obvious retry. It is silent schema corruption or partial hallucination at Rung 1 that downstream steps consume without questioning. Once an unverified local extraction feeds three subsequent frontier calls, the remediation cost completely erases the initial token savings. Guardrailing the transition boundaries between rungs ends up mattering more than the raw token tariff.