# Ask Fable to make codex the workhorse (codex-first) > [!question] Lingering Questions > - What are the actual routing rules in the linked codex-first SKILL.md — this tweet only states the principle, not the mechanism. > - Does the planner/workhorse split hold up when the workhorse (Codex) needs judgment mid-execution, not just at hand-off? > [!tip] Article Ideas > - None — this is a pointer to an external skill, not original analysis worth writing up on its own. > [!link] Cross Reference > - [[3. Resources/Permanent Notes/Route Judgment to the Expensive Model and Bulk Execution to the Cheap One]] — this thread's claim is the same principle, restated for the Fable-plans/Codex-executes pairing; already captured > - `codex-first` skill (`~/.claude/skills/codex-first/`) — the referenced external artifact this tweet is really pointing at > [!quote] Vocabulary > - *Workhorse* — the cheaper model doing high-volume execution once a capable planner has set direction ## Metadata - Source: [[3. Resources/Literature Notes/Ask Fable to make codex the workhorse (codex-first)]] - Author: @steipete (Peter Steinberger) - URL: https://twitter.com/steipete/status/2074638582418231495#thread ## Author's thesis A capable planner (Fable 5) should delegate the bulk of coding execution to a cheaper "workhorse" model (Codex), keeping judgment and planning on the expensive tier and routing token-heavy implementation to the cheap one — the same cost-arbitrage principle as the advisor/orchestrator patterns, applied specifically to Codex as executor. ## Key arguments 1. The core pattern: a capable planner delegates bulk coding execution to Codex as the workhorse. 2. The rationale is general: keep planning/judgment on the expensive model, route token-heavy implementation to the cheaper executor. 3. The author flags that the tweet body is a pointer — the real content lives in the linked codex-first SKILL.md, not in the 280 characters here. ## Claim atoms ### Route bulk coding execution to a cheaper workhorse model > [!quote] ID:rw_01kx4nmew84ck7s5w5y3engcx0 > a capable planner (Fable 5) delegates the bulk of coding execution to Codex as the "workhorse," Where/why: tweet body, main pattern statement; names the division of labor — planner reasons, workhorse executes. #xref ### Keep judgment on the expensive model, volume on the cheap one > [!quote] ID:rw_01kx4nmew84ck7s5w5y3engcx0 > keeping planning/judgment on the expensive model and routing token-heavy implementation to the cheaper executor Where/why: tweet body, rationale clause; the cost-arbitrage principle generalizable beyond Fable/Codex. #xref #aphorisms ### The tweet's substance is the external skill, not its body > [!quote] ID:rw_01kx4nmew84ck7s5w5y3engcx0 > the substance of this tweet is the external codex-first SKILL.md, not the tweet body itself Where/why: author's own closing note; a reading-discipline flag — chase the linked artifact for real depth. #xref ## Permanent candidates - **Already captured**: [[3. Resources/Permanent Notes/Route Judgment to the Expensive Model and Bulk Execution to the Cheap One]] already states this exact routing principle and explicitly names the codex-first "workhorse" pattern in its evidence text. No new Permanent created — connected instead of duplicated. ## Processing checklist - [ ] Read the codex-first SKILL.md directly and capture its actual routing rules (this vault already has a `codex-first` skill — compare against it) - [x] Connect into the relevant MOC — [[3. Resources/MOCs/AI Cost & Tokenomics MOC]] ## Links - [[3. Resources/MOCs/AI Cost & Tokenomics MOC]] — same cost-tier routing question as the Fable 5 advisor/orchestrator thread - [[3. Resources/Permanent Notes/Route Judgment to the Expensive Model and Bulk Execution to the Cheap One]] — evidence-for; restates the same claim for the Codex pairing