--- title: The new core skills: agentic engineering at Granola description: A Granola iOS engineer recounts a year building with AI coding agents, from late 2025 through late 2026, as frontier models like GPT-5.2 through GPT-6 Astra... canonical: https://daily.dev/posts/the-new-core-skills-agentic-engineering-at-granola-xv0kzbdgv twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: The new core skills: agentic engineering at Granola | daily.dev og:description: A Granola iOS engineer recounts a year building with AI coding agents, from late 2025 through late 2026, as frontier models like GPT-5.2 through GPT-6 Astra... og:url: https://daily.dev/posts/the-new-core-skills-agentic-engineering-at-granola-xv0kzbdgv og:image: https://api.daily.dev/og/posts/xV0kZbdGV.png og:image:alt: The new core skills: agentic engineering at Granola 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 new core skills: agentic engineering at Granola **[Jacob’s Tech Tavern](https://daily.dev/sources/jacobbartlett)** · 25 min read · 1 upvotes · 2 comments ## Summary A Granola iOS engineer recounts a year building with AI coding agents, from late 2025 through late 2026, as frontier models like GPT-5.2 through GPT-6 Astra and Opus 4.5 through 5.5 rapidly advanced. He runs up to 11 parallel git checkouts, talks to agents via voice dictation instead of careful prompts, and has agents verify their own work with screenshots and self-written specs. On a Watch app rewrite, a model called Fable 5 cut the codebase from about 13k to 8k lines before being shut down days later. He cut latency on a system by 70% and argues the core new skills are managing attention budget, deciding what to delegate, and system design, since code itself is now cheap. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### why do engineers run multiple AI coding agents in parallel instead of just one Running agents in parallel lets you delegate execution for hours at a time, because higher-effort models make fewer mistakes but work more slowly, and agents can compile, run tests, read logs, and fix their own work autonomously. One approach uses a shared file like ARGUMENT.md where multiple agents debate root causes and reach consensus before implementing a fix. _daily.dev surfaces workflow writeups like this for engineers scaling their own multi-agent setups._ ### how can I get an AI coding agent to verify its own work for a UI feature Ask the agent to take screenshots of its work and display them on a local HTML page, comparing a Figma design side by side with the implementation screenshot. Variations include requesting screen recordings for animations, generating an Excalidraw diagram to explain unfamiliar systems, or compiling a full almanac of every screen with text feedback boxes. _developers refining agent verification loops can find more techniques like this through daily.dev._ ### what happened to the Fable 5 AI model after it shipped Fable 5 was used to redesign a Granola Apple Watch app's architecture, traversing PRs, tests, commit history and tickets to cut the codebase from roughly 13,000 to 8,000 lines with the same functionality better structured. Three days after its release, it was pulled, with the team narrowly avoiding a deadline problem by having rearchitected just in time. _daily.dev helps developers track fast-moving model releases before a dependency gets pulled overnight._ ## Community discussion Top comments from developers on daily.dev. **@kartiknvj** · 0 upvotes > Making the GPT and Claude agents argue in [ARGUMENT.md](http://ARGUMENT.md) until they agree on a fix is a neat trick. You said you mostly let go and dig into the root cause once the fix checks out. I'd still keep a short list of where they disagreed before converging, since that's where my limited review attention would go first. **@bickov** · 0 upvotes > How does the agent take the shots for the HTML page on iOS? xcrun simctl io booted screenshot, or something inside the app? A tall screen captured in one piece gets shrunk before the model reads it. Claude 4.7+ fits images into 2576 px on the long edge, so small text near the bottom of a long list can drop below readable size. Splitting the capture by screen height avoids that. ## Similar posts on daily.dev - [My old job feels distant now](https://daily.dev/posts/my-old-job-feels-distant-now-p6vygtvnp) · Hacker News · 1 upvotes · 0 comments - [Weekly Dose \#11 - AI Agents Are Getting Easier to Build, and Harder to Control](https://daily.dev/posts/weekly-dose-11---ai-agents-are-getting-easier-to-build-and-harder-to-control-ynvfl6cn9) · Machine Learning Pills · 1 upvotes · 0 comments - [The big players in AI coding: models, tools, and who to follow](https://daily.dev/posts/the-big-players-in-ai-coding-models-tools-and-who-to-follow-oqpb5a6tf) · Flavio Copes · 0 upvotes · 0 comments --- Tags: [#architecture](https://daily.dev/tags/architecture), [#ios](https://daily.dev/tags/ios), [#ai-coding](https://daily.dev/tags/ai-coding), [#claude-code](https://daily.dev/tags/claude-code) [View this post on daily.dev](https://daily.dev/posts/the-new-core-skills-agentic-engineering-at-granola-xv0kzbdgv) ```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 new core skills: agentic engineering at Granola","url":"https://daily.dev/posts/the-new-core-skills-agentic-engineering-at-granola-xv0kzbdgv","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/the-new-core-skills-agentic-engineering-at-granola-xv0kzbdgv"},"datePublished":"2026-10-28T16:24:31.503Z","dateModified":"2026-10-28T16:34:50.074Z","description":"A Granola iOS engineer recounts a year building with AI coding agents, from late 2025 through late 2026, as frontier models like GPT-5.2 through GPT-6 Astra...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/bcd2078927c783032a603313f4a0c8f4?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/bcd2078927c783032a603313f4a0c8f4?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Jacob’s Tech Tavern","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":"Jacob’s Tech Tavern","logo":"https://media.daily.dev/image/upload/s--c3Fx6cgz--/f_auto/v1714799103/logos/jacobbartlett","url":"https://daily.dev/sources/jacobbartlett"},"commentCount":2,"discussionUrl":"https://daily.dev/posts/the-new-core-skills-agentic-engineering-at-granola-xv0kzbdgv","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":2}],"keywords":"architecture,ios,ai-coding,claude-code","timeRequired":"PT25M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Jacob’s Tech Tavern","item":"https://daily.dev/sources/jacobbartlett"},{"@type":"ListItem","position":3,"name":"The new core skills: agentic engineering at Granola"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/the-new-core-skills-agentic-engineering-at-granola-xv0kzbdgv","comment":[{"@type":"Comment","text":"Making the GPT and Claude agents argue in ARGUMENT.md until they agree on a fix is a neat trick. You said you mostly let go and dig into the root cause once the fix checks out. I’d still keep a short list of where they disagreed before converging, since that’s where my limited review attention would go first.","datePublished":"2026-10-29T13:51:24.129Z","url":"https://daily.dev/posts/xV0kZbdGV#c-X5mnPMFaM","author":{"@type":"Person","name":"kartik-nvjk","url":"https://daily.dev/kartiknvj","image":"https://media.daily.dev/image/upload/s--3gGgsVCw--/f_auto/v1781456774/avatars/avatar_TvTVeiMdkRCqWUDullFmy?_a=BAMAMiWQ0"}},{"@type":"Comment","text":"How does the agent take the shots for the HTML page on iOS? xcrun simctl io booted screenshot, or something inside the app? A tall screen captured in one piece gets shrunk before the model reads it. Claude 4.7+ fits images into 2576 px on the long edge, so small text near the bottom of a long list can drop below readable size. Splitting the capture by screen height avoids that.","datePublished":"2026-10-29T07:05:36.178Z","url":"https://daily.dev/posts/xV0kZbdGV#c-KlMtXeuUn","author":{"@type":"Person","name":"ab","url":"https://daily.dev/bickov","image":"https://lh3.googleusercontent.com/a/ACg8ocKwOmJ31UWqM4ZXB5ms-yRdFR4BYzyb7FPBRpg1O6DV0MIAF2EP=s96-c"}}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/the-new-core-skills-agentic-engineering-at-granola-xv0kzbdgv#faq","mainEntity":[{"@type":"Question","name":"why do engineers run multiple AI coding agents in parallel instead of just one","acceptedAnswer":{"@type":"Answer","text":"Running agents in parallel lets you delegate execution for hours at a time, because higher-effort models make fewer mistakes but work more slowly, and agents can compile, run tests, read logs, and fix their own work autonomously. One approach uses a shared file like ARGUMENT.md where multiple agents debate root causes and reach consensus before implementing a fix. daily.dev surfaces workflow writeups like this for engineers scaling their own multi-agent setups."}},{"@type":"Question","name":"how can I get an AI coding agent to verify its own work for a UI feature","acceptedAnswer":{"@type":"Answer","text":"Ask the agent to take screenshots of its work and display them on a local HTML page, comparing a Figma design side by side with the implementation screenshot. Variations include requesting screen recordings for animations, generating an Excalidraw diagram to explain unfamiliar systems, or compiling a full almanac of every screen with text feedback boxes. developers refining agent verification loops can find more techniques like this through daily.dev."}},{"@type":"Question","name":"what happened to the Fable 5 AI model after it shipped","acceptedAnswer":{"@type":"Answer","text":"Fable 5 was used to redesign a Granola Apple Watch app's architecture, traversing PRs, tests, commit history and tickets to cut the codebase from roughly 13,000 to 8,000 lines with the same functionality better structured. Three days after its release, it was pulled, with the team narrowly avoiding a deadline problem by having rearchitected just in time. daily.dev helps developers track fast-moving model releases before a dependency gets pulled overnight."}}]} ```