# Guides

[HTML](https://lukejbyrne.com/library/guides/)

- [Fish Audio Drama 3: give two game characters distinct voices](https://lukejbyrne.com/library/guides/qtbcofryvsg/index.md): Follow the Nora and Dex example: generate speech, direct delivery and connect replies to a small set of game actions.
- [Local AI on Omarchy: setup, exact model and example prompts](https://lukejbyrne.com/library/guides/9fdubxu7ziy/index.md): Install LM Studio, choose one Qwen GGUF and check a fresh offline answer. A documentation-based companion with copyable fictional prompts.
- [Five AI desktops: compare their approach before choosing one](https://lukejbyrne.com/library/guides/9gsycbo0gcc/index.md): Explore Omarchy, deepin, AnduinOS, Linux-AI OS and Marlow through their documented features and the examples discussed in Luke’s video.
- [Elephant–Goldfish: plan a feature and review it with fresh AI agents](https://lukejbyrne.com/library/guides/vm3_aieah8w/index.md): Install Google’s open-source EGM plugin, agree a signup form’s behaviour and review the written plan before implementation.
- [Claude Code + DaVinci Resolve: a guide to editable Fusion graphics](https://lukejbyrne.com/library/guides/iqjqjt8iyeo/index.md): Connect carefully, inspect the open project and revise named elements instead of asking for an unexplained finished clip.
- [Mistral Large 4: Ollama Cloud setup and Claude Code](https://lukejbyrne.com/library/guides/u3zuzz52jgm/index.md): Understand Le Chonk, try the documented cloud commands, and check access before treating a model connection as a working coding setup.
- [Tacit OS: desktop apps, AI agents and the USB trial](https://lukejbyrne.com/library/guides/p0vewtf5smi/index.md): Understand Tacit’s Linux desktop, agent accounts and the official USB trial and installation options.
- [Omarchy 4.5 preview: AI tools, desktop apps and how to explore it](https://lukejbyrne.com/library/guides/lim8jjtg8yg/index.md): Understand the 4.5 RC2 preview and the documented Mac and Windows virtual-machine trial routes.
- [FreeLLMAPI: connect provider models to Claude Code](https://lukejbyrne.com/library/guides/txy-dy0x1y4/index.md): A companion to the router setup, provider keys, fallback choices and experimental Claude Code connection.
- [Gemini 4 Carbon: coding claims, access and what to check](https://lukejbyrne.com/library/guides/jjkkaehioxk/index.md): A source-attributed explanation of Google’s announcement, with attention to access and verifiable work.
- [Jev: route, check and score with a structured decision model](https://lukejbyrne.com/library/guides/miyacu4cjtc/index.md): Follow the state-and-questions pattern, then inspect three small examples in the official Playground.
- [Softr client portal: connect records and check client access](https://lukejbyrne.com/library/guides/qqoccqc9uoq/index.md): Build a read-only portal with fictional records, then check each client’s view in Softr Preview.
- [Grok Bot: save a research skill and understand routines](https://lukejbyrne.com/library/guides/aw7ular5vi8/index.md): Follow the manual research-and-replay demonstration, with the existing prompts guide and scheduling limits.
- [GPT-6.1 Terra: choose a model for the work in front of you](https://lukejbyrne.com/library/guides/sgabcex_mgq/index.md): A companion to Luke’s model-choice explainer, its attributed launch examples and manual review workflow.
- [Canva Craft: make a poster and check what remains editable](https://lukejbyrne.com/library/guides/7qa_aixsdao/index.md): A companion to the fictional café-poster test and the Grab Text revision.
- [Build a source-linked second brain with Obsidian and Codex](https://lukejbyrne.com/library/guides/hixuiuranl0/index.md): Organise originals and summaries, trace answers back to notes, and resume work in a fresh Codex task.
- [From one robot sketch to a moving 3D character with Tripo](https://lukejbyrne.com/library/guides/ng2bvigxswc/index.md): Turn a robot sketch into a rigged character, export its walk and wave animations, then inspect the result in Blender and Unity.
- [Four Google AI workflows: email, Sheets, a handoff app and a persistence test](https://lukejbyrne.com/library/guides/1eekrq468d4/index.md): Four practical Google AI examples: email drafting, project spreadsheets, an editor handoff app and a test of its published behaviour.
- [ImagineArt: a product image, Brand Kit and short video](https://lukejbyrne.com/library/guides/gyq9n9netma/index.md): Build a fictional coffee campaign from a hero image through relighting, Brand Kit, Film Studio camera movement and music.
- [Apply five AI concepts to one job in your business](https://lukejbyrne.com/library/guides/oc71wzd7wxq/index.md): Five concepts behind useful AI work: context, connected knowledge, automation, agents and evaluation.
- [One post into five editable Venngage infographics](https://lukejbyrne.com/library/guides/j1wvtmhfo3y/index.md): Turn one written post into five editable infographics, then check wording, resize, restyle and export each design.
- [ImagineArt workflows: one product photo, three ad versions](https://lukejbyrne.com/library/guides/d-nkgyrk8n8/index.md): Turn one product photo into multiple ad variations using connected image and video nodes, with a targeted branch revision.
- [Build a Softr client portal around real records and permissions](https://lukejbyrne.com/library/guides/x3bnzw0db0e/index.md): Build a client portal with real records, login and permissions, then test separate client views and a request workflow.
- [Claude + After Effects: recreate and revise editable animation](https://lukejbyrne.com/library/guides/irgsj8h6dss/index.md): Recreate and revise editable motion graphics, then build a Scripti launch sequence using layers and keyframes.
- [Make an AI infographic you can edit in Venngage](https://lukejbyrne.com/library/guides/jff9y1kgulo/index.md): Create an infographic, correct individual words and numbers, fix the layout and export the finished design.
- [Choose an open-source AI tool for a specific job](https://lukejbyrne.com/library/guides/pdubw-i8ac4/index.md): A practical tour of five open-source projects for AI coworkers, custom tools, design, software research and speech.
- [Build the evidence for a better salary conversation](https://lukejbyrne.com/library/guides/84ha7y6bpkg/index.md): Luke shares the negotiation, evidence and career decisions behind his salary progression, alongside relevant engineering programmes.
- [Use Claude and Perceptis to turn an audit into an editable deck](https://lukejbyrne.com/library/guides/w3fbk8yvddk/index.md): Connect Perceptis to Claude, build a five-slide deck, refine one slide and inspect editable text, shapes and chart data in Keynote.
- [Tripo to Blender: build a Mars concept advert](https://lukejbyrne.com/library/guides/qnnwjbxtita/index.md): Generate matching 3D assets, compare retopology and build a fictional Mars advert with lighting and moving cameras in Blender.
- [OpenDots: set up a researcher that waits for approval](https://lukejbyrne.com/library/guides/dwalirqzh3o/index.md): Install OpenDots and build a Researcher Dot that drafts a brief for review before saving it as a page.
- [Violoop V3 review: test approval behaviour before delegating work](https://lukejbyrne.com/library/guides/m7c-lbv6cwm/index.md): A hands-on V3 review covering email drafting, approval controls, proactive suggestions and background research.
- [AntSeed + Codex: test an alternate provider on a real repair](https://lukejbyrne.com/library/guides/i35yu1w67qw/index.md): Connect Codex to another provider through AntSeed and test a free model on a broken mobile website menu.
- [Compare four AI video editors with the same footage](https://lukejbyrne.com/library/guides/q_dky8brzhy/index.md): Compare four models on the same raw footage and saved editing workflow, including cuts, graphics and follow-up revisions.
- [Macaly Cloud + Claude: build and verify a persistent idea board](https://lukejbyrne.com/library/guides/3n5bzm8o1mk/index.md): Connect Claude to Macaly Cloud, build a persistent video idea board, test its data and add publishing and sign-in.
- [Which Google AI course fits your goal?](https://lukejbyrne.com/library/guides/xbrzkbhj760/index.md): Compare seven Google programmes by goal, workload and practical AI content.
- [Hire three specialists to prepare an AI-built app for launch](https://lukejbyrne.com/library/guides/j3sv0cxt1jw/index.md): Follow three freelance briefs for a logo, landing-page copy and launch video, then review what changed.
- [Use a tutorial to build and test a HighLevel website and CRM](https://lukejbyrne.com/library/guides/kagxfxfzy7a/index.md): Use a tutorial and an existing business site to build and test a website, enquiry form, automation and CRM pipeline.
- [Incogni setup and early results: what to check](https://lukejbyrne.com/library/guides/cpbahmurdp4/index.md): Set up data-removal requests, read the early dashboard results and understand what completed requests mean.
- [Plan a shot in Blender before generating the AI video](https://lukejbyrne.com/library/guides/hepir8iermc/index.md): Plan the action in Blender, guide Seedance 2.5 with the exported reference and compare the result with the original text-only run.
- [Claude Code video editing: from the starter skill to an editable cut](https://lukejbyrne.com/library/guides/vysvvbdbswy/index.md): Set up the free starter skill, review cuts and graphics, then continue with editable Final Cut Pro sources.
- [One listing photo into a reusable Lovart campaign](https://lukejbyrne.com/library/guides/-ajsxjulgzw/index.md): Build a campaign kit, refine the first set, then reuse the instructions on a second property.
- [Codex + Klap: make, review and schedule three Shorts](https://lukejbyrne.com/library/guides/qbkllcevdis/index.md): Separate repeatable clip generation from the browser work needed to edit and schedule the results.
- [Databricks beginner workflow: bakery data to an AI/BI dashboard](https://lukejbyrne.com/library/guides/g3dv6dj3bpo/index.md): Check the data, rank sales, analyse a small review sample and build a dashboard without implying the data proves causation.
- [Build a reusable AI consultant with Claude and Perceptis](https://lukejbyrne.com/library/guides/rmvbsajybio/index.md): Turn a website brief into evidence-backed recommendations and an editable deck, with review before slide production.
- [Research and rebuild a TikTok ad with Maxfusion MCP](https://lukejbyrne.com/library/guides/smyxb2m0xyk/index.md): Move from reference research to a new creative, while keeping the source analysis and production decisions inspectable.
- [Softr MCP: build a project tracker, permissions and an alert](https://lukejbyrne.com/library/guides/vkdzhbv7jsk/index.md): Create real records and a live dashboard, then test the blocked-project workflow and team access.
- [Give a Codex monitoring task reliable web inputs with Decodo](https://lukejbyrne.com/library/guides/kfjbnfjld0m/index.md): Replace blocked page fetches with a defined collection route and test the report against every expected package.
- [Bifrost gateway: test tool permissions, budgets and failover](https://lukejbyrne.com/library/guides/fnnabknzwpo/index.md): Use synthetic support and finance tasks to verify access controls at the gateway.
- [defract + Claude Code: take one feature through a reviewable process](https://lukejbyrne.com/library/guides/h7h0c_u-lyg/index.md): Use scope, design and architecture to resolve decisions before implementation.
- [Editable AI motion graphics with Higgsfield and After Effects](https://lukejbyrne.com/library/guides/ofzuful0j1i/index.md): Ask for native layers and keyframes, then verify that the project can actually be changed.
- [Redesign a website with Codex, Astra and Higgsfield](https://lukejbyrne.com/library/guides/aufp_kgq8dq/index.md): Build a concept preview around the real site, then inspect the generated media and working pages.
- [Airbnb photos to Blender camera guides and five AI shots](https://lukejbyrne.com/library/guides/iqpkn271msy/index.md): Use a rough 3D model to plan movement, then compare guided generations against photo-only results.
- [A practical checklist from 3,500 hours of vibe coding](https://lukejbyrne.com/library/guides/48l4rrwbddc/index.md): Use Luke’s lessons to choose a smaller job, define reliability and get evidence from real users.
- [Compare Astra and Fable on the same AI property-video brief](https://lukejbyrne.com/library/guides/zm3jrlgbjjm/index.md): Control the inputs, record handoffs and judge the finished motion rather than declaring a universal model winner.
- [Plan AI video movement in Blender before generating it](https://lukejbyrne.com/library/guides/7bzeyhje53a/index.md): Create a simple motion reference, inspect the blocking and compare the same brief with and without that guide.
- [One filmed action into multiple Higgsfield ad directions](https://lukejbyrne.com/library/guides/lkjfyc5al7e/index.md): Choose Motion Transfer or Object Swap, then check that the original movement survives the transformation.
- [Use an AI visibility audit to write a better product answer](https://lukejbyrne.com/library/guides/uomrqkjyjd4/index.md): Record a baseline, choose one real customer question and monitor the published answer without claiming instant ranking success.
- [Clueso MCP: turn a rough recording into a product demo and SOP](https://lukejbyrne.com/library/guides/8dp7aoyvgie/index.md): Connect Claude to the editor, structure the scenes and check the voiceover and written guide before export.
- [Seven foundations for an AI-built app you can maintain](https://lukejbyrne.com/library/guides/ulky7oui148/index.md): Use Git, staging, tests, permission boundaries, monitoring, backups and decision notes as a practical release checklist.
- [How to evaluate a world-model demo: memory, actions and transfer](https://lukejbyrne.com/library/guides/bbo3wkk6agu/index.md): Read a convincing simulation more carefully by separating visible quality from whether its world behaves consistently.
- [Puter.js Decision Room: test sign-in, AI usage, history and export](https://lukejbyrne.com/library/guides/_trwyx3onry/index.md): Understand the user-pays model and verify each user’s saved work before deploying a multi-model tool.
- [Build a bounded overnight R&D loop from a daily.dev feed](https://lukejbyrne.com/library/guides/l16ljw_t144/index.md): Rank relevant stories, verify their claims and test small experiments with visible failures.
- [Test service demand before building another SaaS feature](https://lukejbyrne.com/library/guides/z-dgapimifm/index.md): Use Luke’s consulting experiment as a worksheet for collecting evidence about a real customer problem.
- [Tutti VM: hand a real task from Codex to Claude and a teammate](https://lukejbyrne.com/library/guides/2skbggocska/index.md): Carry the working context through a permissioned Room while keeping task selection and approval explicit.
- [Test an Android APK from a Mac with GeeLark](https://lukejbyrne.com/library/guides/lxkv2qlk5cu/index.md): Install your build on a cloud phone and compare real screens with the original app.
- [Build a daily.dev news site with rules you can actually test](https://lukejbyrne.com/library/guides/raj361e43k0/index.md): Keep the API token on the server and test whether each editorial section earns its label.
- [Airbnb photos to a cinematic video with Claude and Higgsfield](https://lukejbyrne.com/library/guides/y31bn4fedwc/index.md): Audit the images, specify camera moves and use deliberate cuts instead of fragile room-to-room transitions.
- [Deploy LinkDrop with livemy.app and test three accounts](https://lukejbyrne.com/library/guides/rwbfkvae3bw/index.md): Go from a public repository to a live app, then verify authentication and saved records on the deployed version.
- [Compare an AI side project using time, receipts and repeatability](https://lukejbyrne.com/library/guides/3zfgt7rbpqy/index.md): A worksheet for reading Luke’s four-project comparison without turning personal results into an income promise.
- [Descope + FastMCP: give two Claude connections different scopes](https://lukejbyrne.com/library/guides/rxksjq_prfw/index.md): Enforce tool permissions on the server and test workspace ownership separately.
- [Claude + Perceptis: compare, edit and export an eight-slide deck](https://lukejbyrne.com/library/guides/pfrrbhk2ofm/index.md): Use the same briefs, inspect real slide layouts and verify the PowerPoint remains editable in Canva.
- [One paying SaaS customer: a worksheet for testing the business](https://lukejbyrne.com/library/guides/bnjzaxfdzyy/index.md): Separate working software, signups, retained use and the priorities that have actually been tested.
- [Give Claude real developer news before asking for an app idea](https://lukejbyrne.com/library/guides/tpnau8damss/index.md): Compare an ungrounded idea list with a source-backed one, then test a small implementation of the selected idea.
- [SureThing: two content agents, shared memory and a reviewable routine](https://lukejbyrne.com/library/guides/vq9eyeadk60/index.md): Set up competitor research and YouTube repurposing, then inspect the reports and draft posts before expanding the system.
- [Pollo Agent 2.0: a landing page into five reviewed ad creatives](https://lukejbyrne.com/library/guides/ppmppq6ere4/index.md): Supply a coherent brief, inspect each output and carry approved brand choices into revisions.
- [Set up a daily.dev reading routine with fewer tabs](https://lukejbyrne.com/library/guides/p68nfsn4x0i/index.md): Personalise the feed, inspect summaries against original stories and decide what deserves a deeper read.
- [Claude Code + Higgsfield: plan and review a visual website upgrade](https://lukejbyrne.com/library/guides/lc3hgvjkfne/index.md): Give the agent a specific site brief and use browser feedback to check the generated media and working interface.
- [Build a coordinated Lovart campaign from a brand book](https://lukejbyrne.com/library/guides/sgrc7u5mc3u/index.md): Keep editable text and brand choices consistent, then simplify video generation when the first result breaks down.
- [Evaluate a consolidated HighLevel stack around one customer record](https://lukejbyrne.com/library/guides/stdlbtk0c3m/index.md): Map the workflows and usage charges before replacing a collection of specialist tools.
- [Before the next AI build: a focus and demand worksheet](https://lukejbyrne.com/library/guides/tkjzqutyncy/index.md): Use Luke’s app-graveyard lessons to define the user, delivery and evidence before adding another project.
- [Claude + Omnisend: audit, segment and draft an email-to-SMS campaign](https://lukejbyrne.com/library/guides/p51hla-uyby/index.md): Start read-only, inspect consent and audience records, then review the campaign and automation before sending.
- [Build a startup-idea app with Kimi K3 in Verdent](https://lukejbyrne.com/library/guides/1cxgelbxfmg/index.md): Use a managed model route, inspect the agents’ work and test real generation separately from a polished interface.
- [Claude Code + Telnyx: build and test a phone agent](https://lukejbyrne.com/library/guides/wp-l4o7gkkw/index.md): Use the original repository and master prompt, check the answers and only then test the assigned phone number.
- [Set up a reviewable X and LinkedIn workflow with York AI](https://lukejbyrne.com/library/guides/1g1p9midajs/index.md): Give the agent brand context, connect its work surface and inspect drafts and engagement suggestions before approval.
- [Unibase: organise saved context and move it between AI chats](https://lukejbyrne.com/library/guides/a888l_wokqy/index.md): Import selectively, label useful memories and check what is actually sent into the next assistant.
- [Review an autonomous Polsia business experiment](https://lukejbyrne.com/library/guides/w1n4ac-hf2i/index.md): Inspect the product, backlog and reported users without treating autonomous activity as a validated company.
- [Hacktron: connect code review and inspect a security-test report](https://lukejbyrne.com/library/guides/nnbu7lhp8ui/index.md): Use an authorised scope, review reproducible findings and verify fixes against the affected behaviour.
- [Test MiniMax M3 inside Claude Code with three real coding tasks](https://lukejbyrne.com/library/guides/dn_ntez-nym/index.md): Keep the inputs and checks visible when comparing a redesign, bug repair and new feature.
- [Build and test a custom booking app with Emergent](https://lukejbyrne.com/library/guides/ovdb2lg7ha4/index.md): Check durations, prices, timezones, confirmations and payments as separate parts of the booking flow.
- [Six Codex workflows: define the inputs, review stops and result checks](https://lukejbyrne.com/library/guides/q5p9krr4elu/index.md): Use the chapter sequence as a planning worksheet for email, billing, phone access, hosting, browser tests and task visibility.
- [Test Tencent HY3 through Cline and OpenRouter on a real app](https://lukejbyrne.com/library/guides/fhhm2ztkwtc/index.md): Use a page change, backlog item, bug investigation and growth brief to inspect more than one coding capability.
- [Use an Ollama model route with Claude Code and inspect token usage](https://lukejbyrne.com/library/guides/tpb5wa7wxqc/index.md): Separate local and cloud execution, test a small job and keep the recorded model-switching quirk visible.
- [apfel on a Mac: inspect a local model through prompts, chat and an API](https://lukejbyrne.com/library/guides/x87fiwoxifq/index.md): Follow the recorded command-line routes, verify compatibility and distinguish model-local processing from the rest of your workflow.
- [Redesign a SaaS interface with GLM 5.2 and Command Code](https://lukejbyrne.com/library/guides/8n3yradwu_s/index.md): Keep the setup, design passes and pre-ship checks separate so the result can be assessed as a working interface.
- [Apify MCP: choose Actors, inspect results and save usable JSON](https://lukejbyrne.com/library/guides/anj8bz2uimk/index.md): Connect the data tools, run a small test and check coverage and cost before expanding the workflow.
- [Choose an AI-engineering course for the job you want to understand](https://lukejbyrne.com/library/guides/79wx-pmqhli/index.md): Compare a beginner path, Claude Code practice and deeper agent concepts using the video’s three selections.
- [Anijam: build an animated story in reviewable stages](https://lukejbyrne.com/library/guides/6wh4ts1uxsm/index.md): Approve the outline and characters before producing the shots, then inspect the timeline and sound.
- [Define a stopping rule before giving an agent a loop](https://lukejbyrne.com/library/guides/uh_ctq1_bo0/index.md): Separate one prompt, a goal with a completion test and a repeated timed check.
- [AIsa + Claude Code: build a dated research brief from live data](https://lukejbyrne.com/library/guides/83gtmo3zdic/index.md): Keep the key private, test one data response and inspect the evidence behind the proposed topics and draft post.
- [Custom GoHighLevel pages: separate the design from the lead workflow](https://lukejbyrne.com/library/guides/cmuu9h_gk1m/index.md): Generate a custom page, move its code into GoHighLevel and check that a form creates the intended contact.
- [ElevenLabs voice agents: connect a conversation to a real SaaS task](https://lukejbyrne.com/library/guides/qefk8c01kvk/index.md): Use a master prompt to plan the voice infrastructure, then review keys, interface, guardrails and conversation behaviour.
- [Bloome: organise a repository task in an agent group chat](https://lukejbyrne.com/library/guides/zfcbfykwh5g/index.md): Create a team with explicit roles, hand it a real repository and inspect the code review and landing-page result before accepting the work.
- [Evaluate GLM 5.2 and Claude Opus 4.8 on a real codebase](https://lukejbyrne.com/library/guides/0qfqmfinfcg/index.md): Repeat a bounded code review and page task, verify the findings and keep historical pricing separate from measured quality.
- [Sintra helpers: from website onboarding to a reviewed recurring task](https://lukejbyrne.com/library/guides/wv_wgl6w6ta/index.md): Give named helpers business context, review individual and team outputs, then configure recurring work and integrations deliberately.
- [egolite: set up agent browsing and inspect a parallel test](https://lukejbyrne.com/library/guides/dpl4ruhrr9w/index.md): Set up egolite, inspect the imported login context and compare the actual browser tasks rather than expanding one timing result into a universal speed claim.
- [Obsidian and Claude Code: maintain a wiki without losing the sources](https://lukejbyrne.com/library/guides/zjrsoxtlunu/index.md): Create a local LLM wiki, ingest a small source collection and verify answers against the original material.
- [Five levels for turning an AI prototype into reliable software](https://lukejbyrne.com/library/guides/jen15ipsxgy/index.md): Use five levels to move from a working prototype to a bounded, observable application, with retained learning links.
- [Build a phone-receptionist demo with Claude Code and ElevenLabs](https://lukejbyrne.com/library/guides/ifegj4y0zxe/index.md): Build the website and voice agent, connect booking data and inspect the phone, calendar and guardrail behaviour before considering a client handoff.
- [Skywork 3.0: review a launch pack before using it](https://lukejbyrne.com/library/guides/cvnfitwopem/index.md): Brief one product, review the research and turn approved strategy into visual launch assets.
- [Command Code: compare DeepSeek V4 and Claude Opus 4.8 on the same job](https://lukejbyrne.com/library/guides/ooezg1ha37y/index.md): Compare four coding tasks from a saved starting point, with cost, scope and visible application behaviour recorded separately.
- [Decodo fallback: retrieve usable public-page context for Claude Code](https://lukejbyrne.com/library/guides/mdbyylfumom/index.md): Use the documented Decodo fallback for public-source extraction, keep the API token out of prompts and inspect the Markdown returned by a small test.
- [Multilogin Cloud Phones: check a build beyond the desktop preview](https://lukejbyrne.com/library/guides/5px_je2juxu/index.md): Run the same app flow across cloud Android profiles, document a mobile defect and retest the revised build.
- [Fable 5 and GPT 5.5: a worksheet for reviewing the Scripti code comparison](https://lukejbyrne.com/library/guides/yb2x8mn4dui/index.md): Use the Scripti comparison to organise bug, UI and security evidence without treating reported counts as a universal model ranking.
- [Fable 5 on Scripti: separate reported findings from a verified repair](https://lukejbyrne.com/library/guides/loigvdftq9m/index.md): Read the historical bug hunt, fixes, redesign and security-audit claims with the source’s model-unavailability notice attached.
- [Lessie AI: turn a distribution brief into a reviewed creator shortlist](https://lukejbyrne.com/library/guides/dezbgbwmkdq/index.md): Define the audience, inspect creator matches and prepare evidence-based outreach drafts before sending anything.
- [Pomelli: build a marketing workspace around checked brand and product inputs](https://lukejbyrne.com/library/guides/ghhhspinvbe/index.md): Review business DNA and catalog records, generate product imagery, then carry approved assets into a campaign brief.
- [Opus 4.8 and Codex 5.5: compare how they review one product](https://lukejbyrne.com/library/guides/gwoxojw2suc/index.md): Compare a landing-page redesign, bug review and infrastructure audit on the same SaaS, keeping the different findings and combined verdict visible.
- [ElevenLabs Speech Engine: add voice to an existing second brain](https://lukejbyrne.com/library/guides/brciizursua/index.md): Connect a channel knowledge base to a voice layer, then test grounding, interruptions and the finished answers.
- [Nimbalyst: organise coding agents around a visible project state](https://lukejbyrne.com/library/guides/9f09cwdm1gq/index.md): Bring sessions, files and tasks into one project workspace, then inspect diffs, commits and diagrams before accepting the work.
- [Bluedot and Claude Code: make meeting recordings useful beyond the call](https://lukejbyrne.com/library/guides/zpffmtenh2q/index.md): Record a meeting, verify its synced context and build a reviewed summary routine.
- [Build a game prototype by speaking to Codex through Typeless](https://lukejbyrne.com/library/guides/qhlqev_kzoo/index.md): Use voice dictation for a coding brief, test the resulting game prototype and feed specific corrections back to Codex.
- [Turn a SaaS landing page into ad concepts with Flova](https://lukejbyrne.com/library/guides/1cv1vmomyck/index.md): Develop a product brief into posters, storyboards and video, then check that the generated assets still represent the product.
- [Obsidian and Codex: build a source-backed second brain](https://lukejbyrne.com/library/guides/zzggjanbcjw/index.md): Build linked notes from transcripts and web pages, then check answers against their retained sources.
- [Tabbit: organise a browser-heavy research workflow](https://lukejbyrne.com/library/guides/aqz2ucv1dxk/index.md): Use Tabbit’s skills, agents and page context to organise one research task, then inspect the answers against the pages involved.
- [Build and inspect a multi-role app with AutoCoder](https://lukejbyrne.com/library/guides/akhrqrrgzl8/index.md): Turn a task-management brief into user flows, backend views and a published preview, checking permissions alongside the interface.
- [Pollo AI and Seedance 2.0: build a reviewable ad-testing workflow](https://lukejbyrne.com/library/guides/3j5igubopde/index.md): Plan seven fictional-product ad formats, inspect the generations and only then consider API scaling.
- [Codex Mobile: run a task from your phone and check its preview](https://lukejbyrne.com/library/guides/gtbdo2tth3i/index.md): Connect the phone workflow, start a small coding run and inspect preview access, skills and remote interruption separately.
- [Connect an AI coding agent to a complete SaaS stack](https://lukejbyrne.com/library/guides/q14zfbbhvlg/index.md): Map domains, hosting, authentication, email and payments, then verify how the connections work together.
- [Codex masterclass: organise a project, run a task and inspect the result](https://lukejbyrne.com/library/guides/9aqh8buatsc/index.md): Follow a complete setup-to-review route through projects, files, artifacts, automation, browser work and the CLI.
- [Organise Codex work through projects, plans and connected tools](https://lukejbyrne.com/library/guides/wzojy9mohn0/index.md): Use the retained workspace, planning and plugin sections to organise a task, with the source’s out-of-range automation chapter markers disclosed.
- [Build and review a SaaS knowledge base with Document360](https://lukejbyrne.com/library/guides/i7oegqxutbm/index.md): Connect the documentation workspace, generate source-based help articles and API pages, then inspect the public support experience.
- [Cuey: compare answers without confusing agreement with proof](https://lukejbyrne.com/library/guides/brypv41yrs0/index.md): Compare model answers, inspect disagreements and keep imported memory and pricing decisions separate.
- [Qwen: compare connected agent work, frontend building and local execution](https://lukejbyrne.com/library/guides/rrqlt8mm1oc/index.md): Follow four Qwen experiments: a packaging agent, landing page, TweetCaster app and local-model switch through Ollama.
- [Use Codex Goals for a bounded coding objective](https://lukejbyrne.com/library/guides/xivm9za3vmw/index.md): Enable the recorded CLI feature, give it a verifiable finish line and inspect the result after a longer session.
- [Long prompt or one line: inspect the racer and Lighthouse builds](https://lukejbyrne.com/library/guides/vxak-t_wboq/index.md): Compare prompt length and model choice separately using the same game and landing-page jobs.
- [Codex workspace basics: plan a task and inspect its output](https://lukejbyrne.com/library/guides/o11e0pmbhrg/index.md): Follow the workspace overview from project selection and planning to connected plugins and parallel-chat status.
- [Build a music-video pipeline with Claude Code, Suno and Freebeat](https://lukejbyrne.com/library/guides/qlmrojrn4ze/index.md): Connect the retained project, generate one song and inspect its storyboard and assembled video before repeating the workflow.
- [DeepSeek V4 comparison: verify findings before counting the winner](https://lukejbyrne.com/library/guides/0lawndmvgwq/index.md): Inspect design, bug reports and operating tradeoffs separately in a dated three-model comparison.
- [OpenClaw through ClawBro: test improvisation, heartbeats and teamwork](https://lukejbyrne.com/library/guides/kpukq97p0xk/index.md): Follow the improvisation, heartbeat and collaboration tests, keeping each observed result distinct from a promise of continuous autonomy.
- [Compare GPT-5.5 and Claude Opus 4.7 on three coding jobs](https://lukejbyrne.com/library/guides/6qnw3uauyo8/index.md): Use the same code and task criteria to review a landing page, bug findings and a competitor-search repair.
- [Kane AI: add a reviewable QA step to an AI-built application](https://lukejbyrne.com/library/guides/euxo8d6qcnm/index.md): Build a natural-language test, inspect its browser run and then configure parallel or recurring checks.
- [Softr: build a CRM around your deals, tasks and permissions](https://lukejbyrne.com/library/guides/yxfqw5rhdh0/index.md): Build a sponsor-management CRM in Softr, inspect role-specific views and follow the email workflow test.
- [Test TicNote Cloud’s Shadow Agent with three source collections](https://lukejbyrne.com/library/guides/a5zoafjasa4/index.md): Turn a transcript into slides, organise mixed notes and audio, then track decisions across multiple meetings.
- [Claude Design: review motion, slides, websites and app mockups](https://lukejbyrne.com/library/guides/i5yggqbq6pc/index.md): Compare four design outputs and apply a consistent system before treating any prototype as finished.
- [MCPize: turn a service idea into a testable MCP server](https://lukejbyrne.com/library/guides/da8lylzeoyu/index.md): Follow the MCPize idea, build and publish sequence, then inspect the deployed server from an agent client.
- [Compare Claude Opus 4.7 and ChatGPT 5.4 on design-heavy code](https://lukejbyrne.com/library/guides/de-0qlojvoq/index.md): Use a real landing-page redesign and a one-shot flight simulator to examine how the tools handle vague briefs and interface decisions.
- [Opus 4.7 workflow tips: permissions, session controls and a custom /go skill](https://lukejbyrne.com/library/guides/2vttgxxqqhc/index.md): Distinguish built-in controls from Luke’s custom /go skill, then review an end-to-end change.
- [Claude Desktop: inspect two agent sessions and their working context](https://lukejbyrne.com/library/guides/qragjiubkra/index.md): Inspect the Desktop update’s folders, two-agent test, notifications and live previews alongside the comparison with Luke’s custom setup.
- [Build a YouTube package and sponsor deck in HIX AI](https://lukejbyrne.com/library/guides/iqjltnqbirg/index.md): Use one project brief across agents, images and slides, then review exported files before comparing subscription value.
- [Ark: manage projects, sessions and running apps from one workspace](https://lukejbyrne.com/library/guides/zu5qnqmznx0/index.md): Organise real Claude Code sessions, preview their outputs and review suggested tasks in a local command centre.
- [Move a Lovable project into Claude Code without losing the starting point](https://lukejbyrne.com/library/guides/6yo0kzbrihu/index.md): Export a Lovable project through GitHub or the recorded browser-download route, then open and preview the local files in Claude Code.
- [Test Claude’s Advisor Strategy against a solo model](https://lukejbyrne.com/library/guides/ol2dqxjlcg4/index.md): Compare Haiku, Sonnet and Opus routes on the same task, tracking total cost, waiting time and useful output.
- [MiniMax Agent and MaxClaw: inspect three distinct workflows](https://lukejbyrne.com/library/guides/imboaigta_y/index.md): Separate website building, research-team instructions and local desktop work before evaluating autonomy and cost.
- [OpenClaw for beginners: memory, alerts and a small app](https://lukejbyrne.com/library/guides/ng1j0habfva/index.md): Follow a memory tracker, trending alerts and micro-app build through the hosted Inclaw AI setup.
- [Read the Mythos announcement claims with their evidence in view](https://lukejbyrne.com/library/guides/iafzwpqv2ou/index.md): Use the archived explainer to separate model claims, benchmark comparisons, access restrictions and forecasts.
- [Animated websites: connect Higgsfield assets to a Claude Code build](https://lukejbyrne.com/library/guides/t0gwftn6doo/index.md): Generate a video asset, implement a scroll effect and investigate jitter before accepting the site.
- [Claude Code Ultraplan: review a plan in a web session](https://lukejbyrne.com/library/guides/79e2l-6fvx8/index.md): Start an Ultraplan web session, select its repository, review the plan with comments and bring the work back to the local environment.
- [Connect Claude Code to Google Workspace with the GWS CLI](https://lukejbyrne.com/library/guides/bgesnjfeqde/index.md): Follow the installation and OAuth sequence, then check a branded presentation and a source-linked Gmail research task.
- [Sleek.design and Claude Code: take a SaaS concept toward a mobile app](https://lukejbyrne.com/library/guides/alqdy7lz540/index.md): Use the master-prompt route for design and backend work, then inspect screens, errors and production data assumptions.
- [Pollo Agent: turn a TikTok reference into a new, inspected video](https://lukejbyrne.com/library/guides/omcu3-2-_a0/index.md): Move from a scout dashboard to reviewed storyboards and video variants before considering API automation.
- [Claude Code versus Anti-Gravity: inspect three different build tasks](https://lukejbyrne.com/library/guides/kluadhnhjes/index.md): Compare mobile design, landing-page output and a Supabase-authenticated backend while keeping setup, skills and model choice visible.
- [Understand the reported Claude Code leak without confusing its scope](https://lukejbyrne.com/library/guides/zy9nfkfpm_w/index.md): Read Luke’s April account as a dated report, separating claims about source files, hidden features and unrelated supply-chain risk.
- [Claude Code internals: separate reported architecture from your usable workflow](https://lukejbyrne.com/library/guides/j-mnyrmk4ag/index.md): Use the historical capability discussion to audit instructions, context and settings without assuming internal features are available.
- [MyClaw and OpenClaw: connect a personal agent to Telegram](https://lukejbyrne.com/library/guides/x3r8l798ktg/index.md): Create the hosted agent, connect Telegram and inspect memory, image generation and a scheduled briefing as separate checks.
- [Compare Softr and Lovable by completing a client-portal workflow](https://lukejbyrne.com/library/guides/1pfahazh-2y/index.md): Use the same brief for login, projects, files, invoices and administration, then verify permissions and the email handoff.
- [OpenRouter as a Claude Code fallback: check the route and the cost](https://lukejbyrne.com/library/guides/mfjchkvj94c/index.md): Prepare a provider fallback, preserve project context and inspect usage before resuming a large task.
- [Claude CoWork: choose a skill, connector or plugin for the job](https://lukejbyrne.com/library/guides/zzk4kxec2l0/index.md): Distinguish reusable task instructions, external-tool connections and bundles of capabilities before building a specialised role.
- [Choose between Claude Cowork and Claude Code by the work you need](https://lukejbyrne.com/library/guides/skpiqbwrzng/index.md): Use Luke’s March comparison to map interface convenience, terminal access and project requirements before changing your routine.
- [Seven Claude Code workflows: organise the work before repeating it](https://lukejbyrne.com/library/guides/cfuoakjrm0w/index.md): Review seven everyday workflows with distinct inputs, permissions and evidence of completion.
- [Claude Code masterclass: build a reliable project workflow before adding teams](https://lukejbyrne.com/library/guides/s2hxdkthtey/index.md): Work through installation, project instructions, GitHub, MCP connections and skills before coordinating agent teams.
- [Build a reusable browser workflow with BrowserAct and Claude Code](https://lukejbyrne.com/library/guides/s0liryeyrfc/index.md): Run a small extraction in the remote browser, inspect the CSV and save the checked method as a skill or workflow.
- [Claude Code /voice: test the words and the workflow](https://lukejbyrne.com/library/guides/s3d1jyzop3i/index.md): Compare dictation accuracy and cleanup across terminal voice, ChatGPT and WhisperFlow using the same spoken task.
- [Claude, Rhine and GPT Zero: inspect the writing behind the score](https://lukejbyrne.com/library/guides/gz9lkwbeari/index.md): Compare detector scores with the actual meaning and readability of the rewritten text, using the two samples in the recording.
- [Monitor Claude Code sessions with See Claude](https://lukejbyrne.com/library/guides/qhe8jnvuldc/index.md): Use the local dashboard to identify active work, inspect history and send a task to the correct terminal.
- [Claude Code /btw: keep a quick question separate from the working conversation](https://lukejbyrne.com/library/guides/pj43iugturk/index.md): Ask a side question without confusing the temporary answer with a change to the main task.
- [Claude visualisations: turn application code into an explanation you can inspect](https://lukejbyrne.com/library/guides/va8pfn5wj1o/index.md): Use real application code to create a user-flow diagram, extend it with charts and interactive explanations, and inspect the saved output.
- [Read Luke’s move to an AI solo business through its work and numbers](https://lukejbyrne.com/library/guides/u0cqmucxaoy/index.md): Use the personal timeline to examine income, responsibilities and the systems behind a one-person business.
- [Claude Code loops versus scheduled tasks: identify what actually keeps running](https://lukejbyrne.com/library/guides/n-2l3dkaxea/index.md): Choose a session loop, desktop schedule or GitHub Actions by the job’s persistence and execution requirements.
- [Claude in Chrome: inspect a browser-controlled lesson workflow](https://lukejbyrne.com/library/guides/hgfxxn-lmg8/index.md): Follow the Chrome setup and Skool lesson test, comparing the control modes, editing result and navigation failures.
- [Compare Claude Opus 4.6 and ChatGPT 5.4 on design-heavy code](https://lukejbyrne.com/library/guides/vrvcj5a1bag/index.md): Use a real landing-page redesign and a one-shot flight simulator to examine how the tools handle vague briefs and interface decisions.
- [Pixel Agents: make Claude Code sessions easier to follow in VS Code](https://lukejbyrne.com/library/guides/ommqijx4pgm/index.md): Use the Pixel Agents extension to organise multiple sessions while checking the underlying work directly.
- [Blue Dot: turn a recorded meeting into usable notes and follow-up](https://lukejbyrne.com/library/guides/tyrhjmq_wy8/index.md): Follow Blue Dot from recording to notes, CRM output and follow-up drafts, then compare the mobile in-person route.
- [Build a first SaaS by checking each stage from idea to deployment](https://lukejbyrne.com/library/guides/3qv9vwovgo4/index.md): Follow the to-do app lesson through local development, version control, testing, voice input and a multi-user cloud deployment.
- [Claude Code on a phone: keep the session and project identifiable](https://lukejbyrne.com/library/guides/ltuqn86pndy/index.md): Connect a phone to the intended session and distinguish remote control from a separate cloud session.
- [WordPress.com or WordPress.org: choose around the work you will maintain](https://lukejbyrne.com/library/guides/aqf33danywe/index.md): Compare self-hosted and managed WordPress through hosting responsibility, plugins, AI building and the full operating cost.
- [Choose Sonnet or Opus by the cost of a checked coding result](https://lukejbyrne.com/library/guides/enxdwydq7ow/index.md): Use the February decision framework to compare task difficulty, output quality and the revisions needed to finish.
- [Gemini 3.1 Pro and Opus 4.6: test the brief as well as the model](https://lukejbyrne.com/library/guides/wiimeope2xu/index.md): Compare three levels of direction on the same SaaS page and separate visual judgement from integration work.
- [Claude Code and Insforge: build a link-in-bio app with a backend](https://lukejbyrne.com/library/guides/hh8yvswyygm/index.md): Use the Launchpad-generated brief, connect Insforge and inspect the link-in-bio application before checking its deployed version.
- [Apply the Google AI Professional Certificate to three real work tasks](https://lukejbyrne.com/library/guides/pewr6hhfavs/index.md): Use Luke’s course takeaways to test spreadsheet analysis, source-based research and a small reusable AI tool.
- [Gemini 3.1 Pro versus Opus 4.6: inspect two real coding tasks](https://lukejbyrne.com/library/guides/ztjni6fafbe/index.md): Review the flight simulator and SaaS feature separately, recording failures and usage beside completed work.
- [Rork Max: inspect a browser-built SwiftUI to-do app](https://lukejbyrne.com/library/guides/pqq_yafai9y/index.md): Follow the Rork Max to-do-app example through corrections, interface work, native iOS features, login and publishing.
- [Create and revise a code-based promo video with Remotion and Claude Code](https://lukejbyrne.com/library/guides/vc8ry3ij5h4/index.md): Set up the project, save the animation instructions and inspect the rendered video against the product brief.
- [Sonnet 4.6 versus Opus 4.6: compare task outcomes before averaging the verdict](https://lukejbyrne.com/library/guides/7tn-dznpwxk/index.md): Keep the failed build, successful redesign and historical usage comparison visible in the same review.
- [n8n or Claude Code: compare the maintenance of one real workflow](https://lukejbyrne.com/library/guides/80-czoke7ls/index.md): Use Luke’s argument about manual node configuration, code-first work and MCP to assess one existing automation rather than assuming every workflow should move.
- [Build a multi-shot concept in Higgsfield Cinema Studio 2.0](https://lukejbyrne.com/library/guides/1mfzn4qbo7k/index.md): Generate a reference image, explore the scene and review camera, character and speed choices across the finished sequence.
- [Seedance 2.0 preview: inspect the evidence before planning a workflow](https://lukejbyrne.com/library/guides/uqwuswsa7ck/index.md): Read the Seedance 2.0 preview as early coverage, then separate feature reports from tests you can actually run.
- [n8n and Atlas Cloud: turn a product row into an image and video](https://lukejbyrne.com/library/guides/pfadii7f7m8/index.md): Follow product data from Google Sheets through scene, image and video generation to Google Drive, inspecting the wait-for-completion step.
- [Choose an AI course by the skill you need to practise](https://lukejbyrne.com/library/guides/dlmhffka-fm/index.md): Compare four learning routes from the February overview, then judge progress through a real task rather than the certificate alone.
- [Kling 3.0 on Higgsfield: review a multi-shot sequence as a complete film](https://lukejbyrne.com/library/guides/9vmyiou8mtm/index.md): Plan a structured sequence from supported references, then inspect shot continuity, character details and audio.
- [ProProfs Training Maker: build a course around a checkable learning goal](https://lukejbyrne.com/library/guides/ytmaa4npsc4/index.md): Generate a training course, review the lessons and assessments, then inspect progress reporting and completion certificates.
- [Kling 3.0 preview: distinguish expected capabilities from a demonstrated workflow](https://lukejbyrne.com/library/guides/h0sja0swdyg/index.md): Read the pre-release expectations separately from the existing Kling demo and the comparison with Veo 3.1.
- [Build a prioritised sponsor shortlist with Lessie AI](https://lukejbyrne.com/library/guides/cellrzeij38/index.md): Describe the sponsor fit, inspect the returned contacts and review outreach separately from search speed.
- [n8n and Templated: generate a reviewable LinkedIn carousel package](https://lukejbyrne.com/library/guides/oyhxrnbnsiu/index.md): Connect research, carousel writing, image generation and PDF rendering, then review the emailed package before posting.
- [TestSprite and Claude: turn recorded failures into a repair-and-retest loop](https://lukejbyrne.com/library/guides/zk_syj1nxqg/index.md): Follow the sixteen failing tests, Claude’s repair pass and the rerun with six passing results, without collapsing that into a clean test suite.
- [Design the automation before asking Claude to build an n8n workflow](https://lukejbyrne.com/library/guides/_dpd2-2tfb8/index.md): Use the February account of failed workflows to clarify execution, deterministic steps and the points that need human judgement.
- [YouArt: keep an energy-drink campaign connected from stills to video](https://lukejbyrne.com/library/guides/jloffsxlxaw/index.md): Build product variations and lifestyle images in a reusable workflow, then inspect the video and campaign exports.
- [Atoms: inspect the handoff from competing builds to an AI team](https://lukejbyrne.com/library/guides/bf5rxfcam3y/index.md): Compare the race-mode builds, inspect the backend and follow the handoff into a team-produced landing page and deployed app.
- [Test Manus 1.6 across an app, browser research and thumbnails](https://lukejbyrne.com/library/guides/tvuwrkmo6sm/index.md): Use one small task in each category to check what the combined workspace can actually deliver for your project.
- [Verdent: make the code change and its rollback inspectable](https://lukejbyrne.com/library/guides/e_kxem9fwvy/index.md): Inspect an existing-codebase change through diffs, tests, source control and review before expanding to parallel agents.
- [Higgsfield AI Influencer Studio: build a consistent character and review its video](https://lukejbyrne.com/library/guides/dgp94sn1mw8/index.md): Create or modify an AI character, inspect its video and assess Higgsfield Earn separately from the generation result.
- [Build a source-linked video-trend briefing with n8n and Virlo](https://lukejbyrne.com/library/guides/kcppe4vfvsw/index.md): Collect trending records, preserve them in Sheets and use the same evidence for an email briefing and chat agent.
- [Flova: take a product ad from a prompt to a checked export](https://lukejbyrne.com/library/guides/hnejxfe2j64/index.md): Review the ad brief, storyboard, generated shots and exported sound as separate production stages.
- [Higgsfield Apps: choose an effect and inspect what it changes](https://lukejbyrne.com/library/guides/xutevqz9dlu/index.md): Compare the Viral Chaos ad example, character swaps and stylised Nano/Sims versions through the actual returned clips.
- [Compare Readdy AI and Lovable through a complete website brief](https://lukejbyrne.com/library/guides/6f1tb9daprc/index.md): Review reference-based generation, animation and integrations without confusing a polished preview with a functioning published site.
- [Google AI Studio Build Mode: from a Gemini app idea to a reviewed deployment](https://lukejbyrne.com/library/guides/fj2rg_cmibo/index.md): Choose an app task and model, check its AI inputs and inspect the deployed version separately from the builder preview.
- [Neo Browser: inspect the sources behind a contextual answer](https://lukejbyrne.com/library/guides/0tkramiiylk/index.md): Organise a browser task around imported context, summaries and reviewed email or shopping outputs, using Neo’s recorded examples.
- [Turn monitored YouTube channels into a checked idea briefing with rtrvr](https://lukejbyrne.com/library/guides/rchzfytom48/index.md): Collect channel data in Sheets, test the cloud workflow and inspect the evidence behind an agent’s suggested topics.
- [AI foundations with Brilliant: turn concepts into a checkable workflow](https://lukejbyrne.com/library/guides/42xf4t25g_o/index.md): Connect probabilistic model behaviour to problem decomposition and a small workflow with explicit boundaries.
- [Skywork: turn video content into a consulting website draft](https://lukejbyrne.com/library/guides/sqfvi9fd3m0/index.md): Give Skywork source video content, inspect the generated consulting page and portfolio, then check the editable HTML and export.
- [Compare ChatGPT Images and Gemini through five distinct image tasks](https://lukejbyrne.com/library/guides/ojrx0uxzaxg/index.md): Review instruction following, realism and consistency without turning one attractive output into an overall model ranking.
- [n8n thumbnail workflow: keep text, template and video destination aligned](https://lukejbyrne.com/library/guides/cbr3aklhgym/index.md): Connect video analysis, dynamic thumbnail rendering and a deliberate upload check before adding an RSS trigger.
- [GPT 5.2: inspect the real-code result alongside the launch claims](https://lukejbyrne.com/library/guides/6kfl5bufa9q/index.md): Keep launch benchmarks separate from the GPT 5.2 website-redesign example using RooCode and OpenRouter in VS Code.
- [Choose a Python data-engineering course by the work you need to practise](https://lukejbyrne.com/library/guides/0vlstob5dh4/index.md): Compare the four routes in Luke’s December overview, then test learning on a small data task.
- [Gemini 3 versus ChatGPT 5.1: a five-task review worksheet](https://lukejbyrne.com/library/guides/lzsg-mz8ud8/index.md): Compare five task types without turning one impressive output into a universal model ranking.
- [Nano Banana Pro: compare prompt following across several image tasks](https://lukejbyrne.com/library/guides/bgnaayy5gyk/index.md): Inspect photos, thumbnail edits, location-based scenes and infographics separately, then check any image-to-video result.
- [Build and inspect a quotes website with TRAE Solo Coder](https://lukejbyrne.com/library/guides/qxllkl777vu/index.md): Review the plan, parallel tasks and branch changes before testing the generated site and its deployed version.
- [AutoCoder: review the full-stack result behind a refined prompt](https://lukejbyrne.com/library/guides/ggq4r8gtr68/index.md): Refine the requirements, inspect a template or generated app, and verify role behaviour after deployment.
- [n8n and Templated: turn a tweet into a reviewed image-post workflow](https://lukejbyrne.com/library/guides/xw6i5uxpvfg/index.md): Follow a Google Sheets row through Templated design, ChatGPT 4.1 captions and Blotato delivery while keeping the selected account and content visible.
- [Inspect an editable 3D world in Marble by World Labs](https://lukejbyrne.com/library/guides/4q8d-n1c_pk/index.md): Follow the cathedral example to distinguish generated appearance, spatial navigation and what survives editing or export.
- [Riff: build a customer-insights workflow you can inspect](https://lukejbyrne.com/library/guides/nsmzu-4fwkg/index.md): Trace spreadsheet input through an AI analysis and Slack output, then check the scheduled run against the same data.
- [Atlas and Comet: compare four browser-agent tasks](https://lukejbyrne.com/library/guides/qgst6bovcnc/index.md): Compare the browsers on comments, an n8n email agent, technology research and a LinkedIn post, keeping task completion distinct from speed.
- [Evaluate Atlas through the browser tasks it actually attempts](https://lukejbyrne.com/library/guides/3fftzcvzk7c/index.md): Use the October review’s research, comment, workflow and grocery examples to inspect actions and outcomes separately.
- [n8n AI Workflow Builder: turn a natural-language request into a checked automation](https://lukejbyrne.com/library/guides/9tych4dj2pi/index.md): Inspect a generated workflow node by node, then test chapter output, email context and calendar behaviour separately.
- [Nextbrowser: turn a research question into a checkable report](https://lukejbyrne.com/library/guides/eqarmtcwjbq/index.md): Follow influencer discovery and Reddit research, then inspect the scheduled report destination before relying on recurring delivery.
- [Choose an AI workflow or agent by the decisions the job requires](https://lukejbyrne.com/library/guides/ub52ra38wxc/index.md): Use the three-level explainer to define model work, tool actions and the controls around a complete task.
- [Veo 3.1 on Higgsfield: review image, motion and sound separately](https://lukejbyrne.com/library/guides/md6ntid-ky4/index.md): Compare complete clips for visual detail, audio timing and prompt adherence, preserving glitches beside the strongest results.
- [Gemini 2.5 Computer Use: follow the demo, local setup and API resources](https://lukejbyrne.com/library/guides/-xmcblrtwh8/index.md): Compare the Browserbase demo, local preview repository and API route, keeping the place where the tools run distinct from model access and cost.
- [Read the Sora 2 Pro access demo as a dated Higgsfield workflow](https://lukejbyrne.com/library/guides/isdrsllafn8/index.md): Use the October example to explore the generator and presets, then inspect the actual clip and the current access conditions.
- [ChatGPT apps: inspect what happens between the request and the useful result](https://lukejbyrne.com/library/guides/3idmc0h2h-a/index.md): Review the original app-in-chat experience through connection, task completion and regional availability rather than launch promises alone.
- [GPT-5 Prompt Optimizer: test the revision rather than trust its label](https://lukejbyrne.com/library/guides/oiam2z5u5dw/index.md): Compare raw and optimised prompts on the same task, checking changed assumptions as well as the final answer.
- [Gemini Live API: distinguish a spoken response from a completed action](https://lukejbyrne.com/library/guides/x13ucqvjvni/index.md): Build the recorded Gemini voice assistant, then inspect function calls, language changes and the attempted email action separately.
- [Build a clickable app prototype with Sketchflow](https://lukejbyrne.com/library/guides/zt_3hs-wje8/index.md): Turn an app brief into screens and workflows, then inspect the HTML export as a demonstration rather than a finished service.
- [Dualite LinkedIn bio roaster: trace the prompt, model response and public app](https://lukejbyrne.com/library/guides/huzfb38xt8q/index.md): Refine the app rules, connect Gemini Flash and test the roast behaviour before deploying the reviewed build.
- [Rork: build a plant-identification app and test it with real images](https://lukejbyrne.com/library/guides/au3iximxyb0/index.md): Follow the Rork plant-identifier build through phone testing, Gemini image analysis and the final identification example.
- [Build an image-to-video ad workflow with n8n, Nano Banana and Veo3](https://lukejbyrne.com/library/guides/62trnmegav0/index.md): Connect the trigger, record and generation services, then inspect one complete product-ad run before scaling.
- [XPack: inspect the MCP service before offering access to it](https://lukejbyrne.com/library/guides/v_lhdy4cryq/index.md): Separate server deployment, editor connection and paid access before judging the marketplace example.
- [Agent Opus: review the decisions inside an automated video](https://lukejbyrne.com/library/guides/kmycolunucw/index.md): Inspect the proposed script-to-video chain and distinguish a finished generated clip from the prediction about short-form feeds.
- [Evaluate a small full-stack app built with Zoer](https://lukejbyrne.com/library/guides/ymeeqbwcsyk/index.md): Use the recorded all-in-one app-building idea to check interface, stored data and the live handoff as separate parts of a product.
- [Hera: keep motion graphics editable and their information accurate](https://lukejbyrne.com/library/guides/lqemzhajudq/index.md): Choose a template or prompt, verify imported media and data, then inspect the exported animation.
- [ElevenMusic: test a musical brief through lyrics, performance and mood](https://lukejbyrne.com/library/guides/ek1_b1obrj8/index.md): Follow the first listening tests, Brew jingle attempt and Scottish-accent song while separating technical success from musical judgement.
- [Automate short-video processing with Submagic, n8n and Drive](https://lukejbyrne.com/library/guides/drqh9khvyn0/index.md): Connect the source folder and editing API, then follow one video through generation, completion and the final file.
- [Venngage: turn a launch plan into a readable infographic](https://lukejbyrne.com/library/guides/i2djtxbgnui/index.md): Generate a Q4 launch timeline, then inspect content, brand edits, accessibility and the exported file.
- [Lindy 3.0 and GPT-5: inspect a lead agent and a news workflow](https://lukejbyrne.com/library/guides/4k7dyahezhu/index.md): Compare the lead-generation and tech-news flows with the recorded browser-login failure rather than treating every step as autonomous success.
- [Build an applied AI learning path for developers](https://lukejbyrne.com/library/guides/ncozelshv-0/index.md): Connect API work, retrieval and application design to a small project you can test and explain.
- [Google Opal: remix a mini-app and check how its steps connect](https://lukejbyrne.com/library/guides/s2hf3ut69t8/index.md): Pair the historical demo discussion with Google’s documented remix and step-editor route, then inspect one small app end to end.
- [BrowserAct: turn a public-page extraction into a usable CSV](https://lukejbyrne.com/library/guides/jifk0bpzwp8/index.md): Follow the public GitHub Trending extraction to CSV and inspect filtering and integration separately from broad access claims.
- [Use Kiro’s specification-first approach on one clear feature](https://lukejbyrne.com/library/guides/abejjw1isag/index.md): Review requirements, steering context and tasks before judging the generated implementation.
- [n8n and Blotato: inspect the distribution route before repeating it](https://lukejbyrne.com/library/guides/utopv6qemhu/index.md): Trace a video record from Airtable through upload to the intended social accounts, checking each delivery separately.
- [UPDF 2.0: inspect editing, conversion and OCR on the document you need](https://lukejbyrne.com/library/guides/6jgp48nloai/index.md): Use the retained UPDF editing, AI and OCR resources to assess a small document task without assuming the title proves a complete Adobe replacement.
- [Test ElevenLabs’ 11 AI assistant with a checked calendar task](https://lukejbyrne.com/library/guides/b4vc1wb3spu/index.md): Connect a voice assistant, inspect the event it creates and separate research answers from calendar changes.
- [Learning to code with AI: apply the eight rules to one real project](https://lukejbyrne.com/library/guides/uqgpaoemwgk/index.md): Use a small project to practise problem solving, then choose AI help, a stack and public feedback deliberately.
- [Gemini CLI: inspect a small terminal task before expanding the workflow](https://lukejbyrne.com/library/guides/ujqevspfiyo/index.md): Follow the terminal setup, Hello World example and API-key configuration while treating the original free-tier claims as dated.
- [Build a reviewable YouTube-description workflow with n8n](https://lukejbyrne.com/library/guides/hucdspan7bm/index.md): Extract the transcript, generate a draft and carry explicit feedback through to a checked metadata update.
- [n8n headshot generator: inspect the complete image lifecycle](https://lukejbyrne.com/library/guides/arhupmzawlu/index.md): Trace a submitted photo through storage, generation, email delivery and cleanup, checking identity and data handling at each stage.
- [n8n research agent: connect the tools and check the evidence](https://lukejbyrne.com/library/guides/6ca4aoqboao/index.md): Connect Perplexity and Gemini 2.5 in n8n, map the actual chat input and inspect the research output’s references.
- [Build a no-code Gemini email assistant in n8n](https://lukejbyrne.com/library/guides/unskufjptyk/index.md): Connect a model and Gmail tools, then test reading, drafting and an explicitly approved send as separate behaviours.
- [n8n human in the loop: connect a specific approval to a specific post](https://lukejbyrne.com/library/guides/u5wyyvwj4ea/index.md): Send the complete draft to Telegram, route feedback back into the workflow and publish only the reviewed version.
- [Veo 3 and n8n: trace the generation request through approval and storage](https://lukejbyrne.com/library/guides/jgu-p06mfrs/index.md): Follow Veo 3 generation through Fal, human review and Airtable records without confusing a completed clip with a publishable result.
- [Automate an avatar-video handoff with HeyGen and n8n](https://lukejbyrne.com/library/guides/ixwxfmqvip0/index.md): Keep the script, avatar and voice linked while testing video generation, status checks and the downloaded result.
- [ElevenLabs and n8n: inspect the voice-to-tool handoff](https://lukejbyrne.com/library/guides/dxylgsyrlxe/index.md): Connect the voice agent to a bounded n8n tool, checking recognised inputs, tool parameters and the spoken result.
- [Learn AI through one useful project, then deepen the parts it needs](https://lukejbyrne.com/library/guides/pqlsryjmvim/index.md): Follow the no-code-to-Python learning path through prompting, APIs, model concepts and one flagship project.
- [Ask useful database questions with AskYourDatabase](https://lukejbyrne.com/library/guides/_9uibheucny/index.md): Define a business metric, inspect the returned answer and carry the same checks into a dashboard or embedded chat.
- [n8n blogging pipeline: review the source, draft and published result](https://lukejbyrne.com/library/guides/c5vye-ea9-y/index.md): Trace the article from chatbot request through research, generation and WordPress delivery, keeping preferences separate from factual evidence.
- [Dart and RooCode: connect project management to a coding task](https://lukejbyrne.com/library/guides/ernzzm0fhtk/index.md): Connect Dart’s project context to the coding agent, inspect generated tasks and compare their implementation with the plan.
- [Turn Fathom meeting notes into a checked follow-up workflow](https://lukejbyrne.com/library/guides/i6jxri1_f5s/index.md): Review transcript, highlights and decisions before moving a summary through Zapier into Notion or another destination.
- [rtrvr.ai: inspect Act, Crawl and Extract through the actual browser result](https://lukejbyrne.com/library/guides/-evce18px4w/index.md): Compare simple browser actions, a multi-step research task and structured extraction without assuming every attempt succeeds.
- [YouTube to newsletter with n8n: define the source-to-draft path](https://lukejbyrne.com/library/guides/_tamhgfi8ke/index.md): Use the limited source as an orientation for a video-to-newsletter workflow, then inspect source fidelity and a controlled email handoff.
- [Self-host an n8n workflow and verify it stays usable](https://lukejbyrne.com/library/guides/kdtjbw_pfuq/index.md): Follow the recorded VPS and domain setup, then test the actual workflow independently of the server dashboard.
- [Python SQL AI API: separate query generation from database access](https://lukejbyrne.com/library/guides/ksrx06y69p4/index.md): Trace a natural-language request through generated SQL to a database result, inspecting the query before execution.
- [n8n MCP: inspect the trigger, client configuration and tool response](https://lukejbyrne.com/library/guides/fanu6r7qu8q/index.md): Distinguish MCP-triggered n8n workflows from community tool nodes, then inspect the supplied Claude Desktop configuration.
- [Review an AI-edited short from Gemoo against its source video](https://lukejbyrne.com/library/guides/amwhvmhvw2w/index.md): Use automated captions, effects and cutaways as editable choices, then check the complete clip before sharing it.
- [Dart AI: distinguish planning, scheduling and execution](https://lukejbyrne.com/library/guides/vf55bckzt_c/index.md): Turn a project into reviewable tasks and inspect the reports or plans generated before accepting task completion.
- [Chat2DB: use natural language to draft and explain a checkable query](https://lukejbyrne.com/library/guides/h4i2lcylydk/index.md): Connect the database context, inspect generated SQL and compare the query result and visualisation with the question asked.
- [Assemble an AI-clone video from script to thumbnail](https://lukejbyrne.com/library/guides/2cuoxkuoi4k/index.md): Keep the approved script linked to the voice, avatar, edit and final visual package across five tools.
- [Immersive Translate: inspect the translated content as well as the browser convenience](https://lukejbyrne.com/library/guides/xwkmo7h1ciy/index.md): Try a known source across the demonstrated formats and check meaning, layout and timing before relying on a translation.
- [n8n and MCP: give a coding agent a specific GitHub job](https://lukejbyrne.com/library/guides/zls53tcoxhy/index.md): Follow repository creation, n8n tool configuration and the local execution step while keeping GitHub changes traceable to the requested task.
- [Build an n8n SQL agent against a small known database](https://lukejbyrne.com/library/guides/hylkh3uldbo/index.md): Follow provisioning, dummy data and MCP tools, then compare natural-language answers with results you can verify.
- [n8n and MCP social agent: keep research and posting as separate tools](https://lukejbyrne.com/library/guides/42kn-vyruoa/index.md): Trace search results through the agent’s draft and Twitter API action, checking the content and destination before publication.
- [n8n and MCP: connect one agent tool and inspect its result](https://lukejbyrne.com/library/guides/df9ldfezqb0/index.md): Follow MCP community-node setup, the agent workflow and Gmail authentication as distinct parts of the recorded build.
- [Understand MCP by tracing a tool request and its result](https://lukejbyrne.com/library/guides/mgs_9deqsp4/index.md): Use the March explainer to distinguish the language model, API operation and server connection in an agent workflow.
- [n8n YouTube agent: keep the transcript behind the answer](https://lukejbyrne.com/library/guides/k4iargzlhta/index.md): Connect transcription, stored source material and agent tools, then inspect answers and generated files against the actual video content.
- [Python search agent: connect the model and tool around one question](https://lukejbyrne.com/library/guides/yhc8i2rvlua/index.md): Build the recorded Python search agent with LangChain, OpenAI and DuckDuckGo, then inspect the first run before adding a chat loop.
- [Build a web-research agent with n8n and Crawl4AI](https://lukejbyrne.com/library/guides/anvykchjkes/index.md): Separate live scraping, stored retrieval and the final answer so you can inspect the evidence at each stage.
- [n8n local installation: inspect the runtime before adding an AI agent](https://lukejbyrne.com/library/guides/8mpfbi_9wzm/index.md): Follow the npm-based local setup, open the browser interface and separate workflow hosting from model costs and availability.
- [Local RAG: inspect the path from document chunks to a grounded answer](https://lukejbyrne.com/library/guides/ys_d7kvkdmo/index.md): Follow document ingestion, embeddings and ChromaDB retrieval before checking the local chat response against the actual source passages.
- [Build a Python RAG chatbot with Drive documents and Pinecone](https://lukejbyrne.com/library/guides/h6pdzazozxo/index.md): Trace a document from ingestion through embedding and retrieval to the evidence used in a chat answer.
- [AI benchmark scores: compare the test and conditions before the model](https://lukejbyrne.com/library/guides/gztgxvaw11e/index.md): Identify what each benchmark measures and read the exact metric before interpreting the historical R1/o1 comparison.
- [Local DeepSeek with Ollama: inspect chat and API access separately](https://lukejbyrne.com/library/guides/pcpgzd6a87a/index.md): Follow the local DeepSeek-R1 chatbot and API example, then compare models on the same small task.