--- title: Seats and Sunsets — Steve Yegge | daily.dev description: Steve Yegge shares an update on Wheelhouse, his agentic orchestrator built with Beads and used to develop his game Wyvern, running dozens of Claude... canonical: https://daily.dev/posts/seats-and-sunsets-steve-yegge-p1atopjde twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Seats and Sunsets — Steve Yegge | daily.dev og:description: Steve Yegge shares an update on Wheelhouse, his agentic orchestrator built with Beads and used to develop his game Wyvern, running dozens of Claude... og:url: https://daily.dev/posts/seats-and-sunsets-steve-yegge-p1atopjde og:image: https://api.daily.dev/og/posts/P1aToPjDe.png og:image:alt: Seats and Sunsets — Steve Yegge og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Seats and Sunsets — Steve Yegge **[Steve Yegge](https://daily.dev/sources/yegge)** · 22 min read · 0 upvotes · 0 comments ## Summary Steve Yegge shares an update on Wheelhouse, his agentic orchestrator built with Beads and used to develop his game Wyvern, running dozens of Claude Fable/Opus/Sol instances. He describes a worsening 'fuel apocalypse' as Fable pricing rises (needing ~55 Claude Max accounts, ~$12k/month, to keep the factory running 24x7), forcing him to focus on token efficiency, cut fences from 650+ down to 14, and park/merge several agent 'seats.' He also relays a story from Brendan Hopper about an agent colony that got stuck looping in a low-capability model (Haiku) during a 'honey-time' bug, and expands his 'seats' concept — persistent role-based agent identities that cache trust and reduce token spend from re-verification — arguing that trust and fuel costs are fundamentally the same problem in agentic systems. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### why is my Claude Fable-based agent orchestrator suddenly running out of budget even though usage seems similar to before Repeated Fable pricing cuts framed as discounts have actually raised effective consumption costs, since burning through a full week's worth of one Claude Max account now takes only 2 to 4 hours instead of a full week. Sustaining a factory of around 25 Fable and 25 Opus/Sol instances running continuously would require roughly 55 Claude Max accounts, costing about $12,000 per month at current pricing. _Anyone tuning multi-agent token spend can follow ongoing pricing and orchestration shifts on daily.dev._ ### what causes an AI agent factory to stall out even when everything looks fine Excessive accumulation of mechanically-enforced refusal rules, called fences, can paralyze an agent system: one orchestrator built up over 400 ruling beads, 185 rule rows in a config file, and 650 distinct refusal sites across 173 scripts before work stopped being possible at all. The fix was cutting fences down to just 14 and requiring personal approval for any new one. _Developers debugging stalled agent pipelines can track these failure patterns as they surface on daily.dev._ ### why do AI agents burn extra tokens verifying things instead of just acting Without a persistent role or identity, a model booting into a fresh session must derive from scratch whether it can trust the actor, environment, and consequences of an action, and that derivation costs context on every single session. Giving models a persistent 'seat' with pre-established authority, history, and blame-free postmortem policies turns that costly derivation into a cheap lookup, cutting token spend on re-verification. _Teams designing agent trust models can follow emerging orchestration patterns like this on daily.dev._ ## Similar posts on daily.dev - [Gas Town’s Agent Patterns, Design Bottlenecks, and Vibecoding at Scale](https://daily.dev/posts/gas-town-s-agent-patterns-design-bottlenecks-and-vibecoding-at-scale-bte6tnb5e) · Hacker News · 1 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#claude](https://daily.dev/tags/claude) [View this post on daily.dev](https://daily.dev/posts/seats-and-sunsets-steve-yegge-p1atopjde) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Seats and Sunsets — Steve Yegge","url":"https://daily.dev/posts/seats-and-sunsets-steve-yegge-p1atopjde","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/seats-and-sunsets-steve-yegge-p1atopjde"},"datePublished":"2026-09-16T00:14:36.694Z","dateModified":"2026-09-16T06:07:26.257Z","description":"Steve Yegge shares an update on Wheelhouse, his agentic orchestrator built with Beads and used to develop his game Wyvern, running dozens of Claude...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/353281dc082f9284226ef680d8908bdb?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/353281dc082f9284226ef680d8908bdb?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Steve Yegge","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Steve Yegge","logo":"https://media.daily.dev/image/upload/s---P6HxL-S--/f_auto,q_auto/v1787723832/logos/yegge?_a=BAMAMicg0","url":"https://daily.dev/sources/yegge"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/seats-and-sunsets-steve-yegge-p1atopjde","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,ai-agents,claude","timeRequired":"PT22M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Steve Yegge","item":"https://daily.dev/sources/yegge"},{"@type":"ListItem","position":3,"name":"Seats and Sunsets — Steve Yegge"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/seats-and-sunsets-steve-yegge-p1atopjde#faq","mainEntity":[{"@type":"Question","name":"why is my Claude Fable-based agent orchestrator suddenly running out of budget even though usage seems similar to before","acceptedAnswer":{"@type":"Answer","text":"Repeated Fable pricing cuts framed as discounts have actually raised effective consumption costs, since burning through a full week's worth of one Claude Max account now takes only 2 to 4 hours instead of a full week. Sustaining a factory of around 25 Fable and 25 Opus/Sol instances running continuously would require roughly 55 Claude Max accounts, costing about $12,000 per month at current pricing. Anyone tuning multi-agent token spend can follow ongoing pricing and orchestration shifts on daily.dev."}},{"@type":"Question","name":"what causes an AI agent factory to stall out even when everything looks fine","acceptedAnswer":{"@type":"Answer","text":"Excessive accumulation of mechanically-enforced refusal rules, called fences, can paralyze an agent system: one orchestrator built up over 400 ruling beads, 185 rule rows in a config file, and 650 distinct refusal sites across 173 scripts before work stopped being possible at all. The fix was cutting fences down to just 14 and requiring personal approval for any new one. Developers debugging stalled agent pipelines can track these failure patterns as they surface on daily.dev."}},{"@type":"Question","name":"why do AI agents burn extra tokens verifying things instead of just acting","acceptedAnswer":{"@type":"Answer","text":"Without a persistent role or identity, a model booting into a fresh session must derive from scratch whether it can trust the actor, environment, and consequences of an action, and that derivation costs context on every single session. Giving models a persistent 'seat' with pre-established authority, history, and blame-free postmortem policies turns that costly derivation into a cheap lookup, cutting token spend on re-verification. Teams designing agent trust models can follow emerging orchestration patterns like this on daily.dev."}}]} ```