--- title: The Three Money-Saving Machines I Built — Did Nothing.... description: A developer measures four self-built tools meant to work with AI coding assistants: a model picker that routes tasks to cheaper models, a local laptop model, a... canonical: https://daily.dev/posts/the-three-money-saving-machines-i-built-did-nothing-the-boring-one-i-ignored-saved-everything--z2dfg3bjs twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: The Three Money-Saving Machines I Built — Did Nothing. The Boring One I Ignored — Saved Everything. | daily.dev og:description: A developer measures four self-built tools meant to work with AI coding assistants: a model picker that routes tasks to cheaper models, a local laptop model, a... og:url: https://daily.dev/posts/the-three-money-saving-machines-i-built-did-nothing-the-boring-one-i-ignored-saved-everything--z2dfg3bjs og:image: https://api.daily.dev/og/posts/Z2DfG3bjS.png og:image:alt: The Three Money-Saving Machines I Built — Did Nothing. The Boring One I Ignored — Saved Everything. 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. # The Three Money-Saving Machines I Built — Did Nothing. The Boring One I Ignored — Saved Everything. **[Medium](https://daily.dev/sources/medium_js)** · 12 min read · 0 upvotes · 0 comments ## Summary A developer measures four self-built tools meant to work with AI coding assistants: a model picker that routes tasks to cheaper models, a local laptop model, a data-compression tool, and a codebase question-answering tool with link verification. Testing revealed the three explicitly built as 'money-savers' saved almost nothing this week — the model picker's rule never triggers on a tie even when a cheap model matches an expensive one's results, the local model mostly double-checks itself against the paid cloud model rather than replacing it, and the compression tool remains switched off pending safety validation. Meanwhile the fourth tool, a plain codebase Q&A system originally seen as background plumbing, answered 86 questions with 431 code links (92% verified accurate) and caught 7 stale notes via a mostly-free local staleness check. The author also notes an uncounted caching/reuse mechanism hit a 99.8% reuse rate, likely saving more money than any of the three intentional cost-saving tools. The takeaway: measuring revealed which savings mechanisms were structurally capable of paying off versus which just looked clever. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Why would a rule that only switches to a cheaper AI model when it produces better results fail to save money even when the cheap model performs just as well? Because a tie does not count as a win under a strict 'only switch for better quality' rule, so when an expensive model already scores perfectly, a cheaper model that matches it can never trigger a switch, even though matching performance at lower cost is exactly the savings opportunity being missed. The fix is to default to the cheap model first and only fall back to the expensive one on failure. _Developers designing model-routing logic can compare real-world approaches and pitfalls like this on daily.dev._ ### How accurate were the code line links generated by an AI codebase question-answering tool in a real usage test? About 92% of generated links pointed to code that actually existed at the stated location, based on 431 links produced across 86 answered questions in one week, with 36 links found to be fabricated references to nonexistent code. A verified link only confirms the reference is real, not that the accompanying explanation is correct. _Anyone evaluating AI-assisted code navigation tools can track findings like this on daily.dev._ ### Why might running an AI model locally on a laptop fail to reduce cloud AI costs even when it handles many tasks? A local model can process a high volume of tasks yet barely reduce cloud costs if its output is routed onward to a paid cloud model for final verification rather than replacing that call outright. In one measured week, a local model ran 20 background jobs but only 1 actually substituted for a paid cloud call, because most jobs were reviews that still deferred to the cloud model. _Teams weighing local versus cloud AI inference tradeoffs can follow practical breakdowns like this on daily.dev._ ## Similar posts on daily.dev - [Instead of paying $40 a month for cloud AI tools, I run everything through a local model on my gaming PC](https://daily.dev/posts/instead-of-paying-40-a-month-for-cloud-ai-tools-i-run-everything-through-a-local-model-on-my-gamin-gyzicbsx7) · XDA Developers · 4 upvotes · 1 comments - [I Gave Five AI Coding Models the Same Ticket. A Blind Judge Picked the One That Cost Nothing.](https://daily.dev/posts/i-gave-five-ai-coding-models-the-same-ticket-a-blind-judge-picked-the-one-that-cost-nothing--b1xt8pif3) · Medium · 2 upvotes · 1 comments - [Do you really need the latest AI model? Or are you just upgrading your tech ego?](https://daily.dev/posts/do-you-really-need-the-latest-ai-model-or-are-you-just-upgrading-your-tech-ego--o2wyzwvun) · Codemotion · 0 upvotes · 0 comments - [Can a Local LLM Run My AI Assistant?](https://daily.dev/posts/can-a-local-llm-run-my-ai-assistant--k8pjpamza) · Towards Data Science · 0 upvotes · 0 comments - [Cut AI-coding costs without slowing down](https://daily.dev/posts/cut-ai-coding-costs-without-slowing-down-psm55sanb) · LeadDev · 0 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#ai-coding](https://daily.dev/tags/ai-coding), [#rag](https://daily.dev/tags/rag), [#local-ai](https://daily.dev/tags/local-ai) [View this post on daily.dev](https://daily.dev/posts/the-three-money-saving-machines-i-built-did-nothing-the-boring-one-i-ignored-saved-everything--z2dfg3bjs) ```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":"The Three Money-Saving Machines I Built — Did Nothing. The Boring One I Ignored — Saved Everything.","url":"https://daily.dev/posts/the-three-money-saving-machines-i-built-did-nothing-the-boring-one-i-ignored-saved-everything--z2dfg3bjs","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/the-three-money-saving-machines-i-built-did-nothing-the-boring-one-i-ignored-saved-everything--z2dfg3bjs"},"datePublished":"2026-08-18T00:49:12.356Z","dateModified":"2026-09-14T06:54:29.983Z","description":"A developer measures four self-built tools meant to work with AI coding assistants: a model picker that routes tasks to cheaper models, a local laptop model, a...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/bc3673d8b9379bd602b87ec9b65150bc?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/bc3673d8b9379bd602b87ec9b65150bc?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Medium","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":"Medium","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/medium","url":"https://daily.dev/sources/medium_js"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/the-three-money-saving-machines-i-built-did-nothing-the-boring-one-i-ignored-saved-everything--z2dfg3bjs","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,ai-coding,rag,local-ai","timeRequired":"PT12M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Medium","item":"https://daily.dev/sources/medium_js"},{"@type":"ListItem","position":3,"name":"The Three Money-Saving Machines I Built — Did Nothing. The Boring One I Ignored — Saved Everything."}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/the-three-money-saving-machines-i-built-did-nothing-the-boring-one-i-ignored-saved-everything--z2dfg3bjs#faq","mainEntity":[{"@type":"Question","name":"Why would a rule that only switches to a cheaper AI model when it produces better results fail to save money even when the cheap model performs just as well?","acceptedAnswer":{"@type":"Answer","text":"Because a tie does not count as a win under a strict 'only switch for better quality' rule, so when an expensive model already scores perfectly, a cheaper model that matches it can never trigger a switch, even though matching performance at lower cost is exactly the savings opportunity being missed. The fix is to default to the cheap model first and only fall back to the expensive one on failure. Developers designing model-routing logic can compare real-world approaches and pitfalls like this on daily.dev."}},{"@type":"Question","name":"How accurate were the code line links generated by an AI codebase question-answering tool in a real usage test?","acceptedAnswer":{"@type":"Answer","text":"About 92% of generated links pointed to code that actually existed at the stated location, based on 431 links produced across 86 answered questions in one week, with 36 links found to be fabricated references to nonexistent code. A verified link only confirms the reference is real, not that the accompanying explanation is correct. Anyone evaluating AI-assisted code navigation tools can track findings like this on daily.dev."}},{"@type":"Question","name":"Why might running an AI model locally on a laptop fail to reduce cloud AI costs even when it handles many tasks?","acceptedAnswer":{"@type":"Answer","text":"A local model can process a high volume of tasks yet barely reduce cloud costs if its output is routed onward to a paid cloud model for final verification rather than replacing that call outright. In one measured week, a local model ran 20 background jobs but only 1 actually substituted for a paid cloud call, because most jobs were reviews that still deferred to the cloud model. Teams weighing local versus cloud AI inference tradeoffs can follow practical breakdowns like this on daily.dev."}}]} ```