https://itsaimathias.com/shorts/short-claude-code-cloud-vs-remote-control-wsyz8mvxxeshttps://i.ytimg.com/vi/wsyz8mvxxes/hqdefault.jpgClaude Code: Cloud vs Remote ControlA short practical Claude Code explanation, using official documentation and Mathias’s digital avatar. Cloud execution and Remote Control run in different places. Full walkthrough: https://www.youtube.com/watch?v=HboQzPhlwYU Official guide: https://code.claude.com/docs/en/claude-code-on-the-web #ClaudeCode #AI #Codinghttps://www.youtube-nocookie.com/embed/wsyz8mvxxes292026-10-21T08:21:24-07:00https://itsaimathias.com/shorts/short-why-your-private-repo-is-missing-in-claude-code-gL7MWEE9o84https://i.ytimg.com/vi/gL7MWEE9o84/hqdefault.jpgWhy Your Private Repo Is Missing in Claude CodeA short practical Claude Code explanation, using official documentation and Mathias’s digital avatar. Connecting GitHub and granting repository access are separate steps. Full walkthrough: https://www.youtube.com/watch?v=HboQzPhlwYU Official guide: https://code.claude.com/docs/en/claude-code-on-the-web #ClaudeCode #AI #Codinghttps://www.youtube-nocookie.com/embed/gL7MWEE9o84272026-10-21T08:20:19-07:00https://itsaimathias.com/watch/coding-on-the-go-with-claude-code-HboQzPhlwYUhttps://i.ytimg.com/vi/HboQzPhlwYU/hqdefault.jpgCoding on the go with Claude CodeCode on the go with Claude Code: how cloud sessions work, how GitHub access and environments fit together, and how to review a change from your phone. This is a documentation-led walkthrough using official Claude screens and an original fictional local newsletter example. The screens from Claude’s field guide are attributed vendor examples. We did not run Claude cloud, create its pull request, or test a cloud session in this production. The video explains the setup and review workflow and the difference between Cloud and Remote Control.https://www.youtube-nocookie.com/embed/HboQzPhlwYU2322026-10-21T08:19:04-07:00https://itsaimathias.com/shorts/short-this-ai-world-is-not-running-in-real-time-P3yTAPLvjdUhttps://i.ytimg.com/vi/P3yTAPLvjdU/hqdefault.jpgThis AI World Is Not Running in Real TimeStreaming does not automatically mean real-time interaction. Streaming Code World Model reports about 5.1 seconds to generate a chunk that plays for 0.71 seconds on one H200 GPU. Smooth playback and generation latency are different measurements. Would that delay work for your use case?https://www.youtube-nocookie.com/embed/P3yTAPLvjdU392026-10-19T06:53:23-07:00https://itsaimathias.com/shorts/short-moss-image-give-ai-a-layout-not-just-a-prompt-vM2K3rDJYt8https://i.ytimg.com/vi/vM2K3rDJYt8/hqdefault.jpgMOSS Image: Give AI a Layout, Not Just a PromptMOSS Image accepts structured placement alongside a text prompt. Reserve a headline and product area, then check the result. The visual here is our original layout illustration, not output from the model. What would you want the image model to keep exactly in place?https://www.youtube-nocookie.com/embed/vM2K3rDJYt8382026-10-19T06:51:59-07:00https://itsaimathias.com/shorts/short-gemini-4-argon-s-1m-tokens-have-a-catch-iopKU4MI5SYhttps://i.ytimg.com/vi/iopKU4MI5SY/hqdefault.jpgGemini 4 Argon's 1M Tokens Have a CatchGemini 4 Argon's 1,048,576-token window is shared by input, thinking and output. The default output ceiling is 65,536 tokens; longer background generation stays within the remaining context. Our animation is an original explanation. What task would you put in that context window?https://www.youtube-nocookie.com/embed/iopKU4MI5SY372026-10-19T06:49:07-07:00https://itsaimathias.com/watch/gemini-4-argon-is-here-ai-news-you-can-actually-use-wScATWl_u6whttps://i.ytimg.com/vi/wScATWl_u6w/hqdefault.jpgGemini 4 Argon Is Here — AI News You Can Actually UseGemini 4 Argon leads an announcement-based AI news episode about control: managing context, specifying image placement, keeping scenes consistent and choosing hardware. The episode uses original labeled illustrations and credited project excerpts; none of the featured models was independently benchmarked by ItsAIMathias.https://www.youtube-nocookie.com/embed/wScATWl_u6w10692026-10-19T06:48:22-07:00https://itsaimathias.com/shorts/short-claude-project-stuck-check-this-before-you-retry-jNU6j6gOBYUhttps://i.ytimg.com/vi/jNU6j6gOBYU/hqdefault.jpgClaude Project Stuck? Check This Before You RetryA stuck thread needs a specific diagnosis. Find the failed request, identify missing access, and answer approval prompts inside the thread that requested them. Watch the full Claude Projects setup guide: https://www.youtube.com/watch?v=iH0bSLJbLf0 Official documentation: https://code.claude.com/docs/en/claude-projects Checked 19 October 2026. AI-generated presenter and voice.https://www.youtube-nocookie.com/embed/jNU6j6gOBYU372026-10-19T00:10:04-07:00https://itsaimathias.com/shorts/short-claude-projects-vs-subagents-the-difference-that-matters-qzVMuF-f7Mohttps://i.ytimg.com/vi/qzVMuF-f7Mo/hqdefault.jpgClaude Projects vs Subagents: The Difference That MattersProject threads are independent sessions you can revisit and steer. A subagent helps a parent session with a bounded assignment. Here is the practical distinction. Watch the full Claude Projects setup guide: https://www.youtube.com/watch?v=iH0bSLJbLf0 Official documentation: https://code.claude.com/docs/en/claude-projects Checked 19 October 2026. AI-generated presenter and voice.https://www.youtube-nocookie.com/embed/qzVMuF-f7Mo402026-10-19T00:06:42-07:00https://itsaimathias.com/watch/claude-projectshttps://itsaimathias.com/assets/claude-projects.webpClaude Projects: Stop Managing Every AI SessionClaude Projects adds a coordinating conversation around separate work threads. Mathias uses an AI Release Desk to explain how to define the first assignment, supply useful context, and inspect blocked work; the board is a local teaching example, while the product explanation follows Anthropic’s documentation.https://www.youtube-nocookie.com/embed/iH0bSLJbLf010602026-10-19T00:03:43-07:00https://itsaimathias.com/shorts/short-your-ai-slide-changed-did-the-old-number-survive-YW2OcJJcffkhttps://i.ytimg.com/vi/YW2OcJJcffk/hqdefault.jpgYour AI Slide Changed. Did the Old Number Survive?Changing 107 to 110 should also update the growth claim from 7% to 10%. Check the headline, chart and source note, then open the file you’ll send. This fictional local example shows a review method. It does not demonstrate ChatGPT generating, editing or exporting a deck. Incoming tickets remain 140; the revised aggregate activity ratio is about 78.6%, not a same-cohort resolution rate.https://www.youtube-nocookie.com/embed/YW2OcJJcffk282026-10-18T03:11:13-07:00https://itsaimathias.com/shorts/short-7-more-tickets-resolved-is-that-actually-better-Fbn94jsCTVIhttps://i.ytimg.com/vi/Fbn94jsCTVI/hqdefault.jpg7% More Tickets Resolved. Is That Actually Better?A 7% rise in tickets resolved is one count, not proof of better service or a shrinking backlog. This short checks the claim against a fictional support example. All figures are synthetic teaching data. This is a local deck illustration, not a recorded ChatGPT Slides product test.https://www.youtube-nocookie.com/embed/Fbn94jsCTVI272026-10-18T03:08:54-07:00https://itsaimathias.com/watch/chatgpt-slideshttps://itsaimathias.com/assets/chatgpt-slides.webpChatGPT Slides Is Rolling Out. Here’s What to Check FirstA five-slide support update shows why an AI presentation needs more than a convincing headline. Starting with 100 versus 107 resolved tickets, Mathias adds incoming workload, revises the result to 110, and follows the correction through the deck; the figures and editable deck are original teaching material, not a native ChatGPT Slides test.https://www.youtube-nocookie.com/embed/wMwduMhCQyg4282026-10-18T03:06:42-07:00https://itsaimathias.com/shorts/short-an-ai-budget-alert-is-not-a-spending-limit-tgqpPHitiX4https://i.ytimg.com/vi/tgqpPHitiX4/hqdefault.jpgAn AI Budget Alert Is Not a Spending LimitWarning at 6. Settled total 12. Nominal budget 10. A notification is different from deciding whether the next task may start. Our original local ledger uses fictional teaching units. This is not an n8n execution or a billing benchmark.https://www.youtube-nocookie.com/embed/tgqpPHitiX4412026-10-15T08:28:04-07:00https://itsaimathias.com/shorts/short-shorter-video-same-natural-voice-here-s-the-ai-edit-RpVn8XM544Ehttps://i.ytimg.com/vi/RpVn8XM544E/hqdefault.jpgShorter Video. Same Natural Voice. Here’s the AI Edit.A shorter edit. The same natural voice. Codex proposed one cut in our owned Short, and a local renderer preserved the original speech speed. This is one prepared example with supplied speech blocks. It does not grade raw-footage storytelling or prove that shorter is always better.https://www.youtube-nocookie.com/embed/RpVn8XM544E382026-10-15T08:26:25-07:00https://itsaimathias.com/shorts/short-gemini-skills-save-the-method-keep-the-facts-current-j4jn39tgtm8https://i.ytimg.com/vi/j4jn39tgtm8/hqdefault.jpgGemini Skills: Save the Method, Keep the Facts CurrentSave the method. Keep the facts current. A reusable skill should preserve the source status instead of turning a pending decision into a confirmed launch date. This is our authored local teaching example with fictional project notes, not an observed Gemini response.https://www.youtube-nocookie.com/embed/j4jn39tgtm8382026-10-15T08:25:51-07:00https://itsaimathias.com/shorts/short-your-ai-document-test-needs-one-missing-question-FfWMPJ11rzUhttps://i.ytimg.com/vi/FfWMPJ11rzU/hqdefault.jpgYour AI Document Test Needs One Missing QuestionA real citation can still support the wrong answer. Add a question your source cannot answer before trusting a document assistant. This is our authored fictional 20-question document QA pack. No Haiku inference or model accuracy test was run.https://www.youtube-nocookie.com/embed/FfWMPJ11rzU352026-10-15T08:25:07-07:00https://itsaimathias.com/watch/ai-video-editinghttps://itsaimathias.com/assets/ai-video-editing.webpI Let Codex Recut My Video—Without Speeding Up the VoiceCodex receives one tightly defined recut: shorten a 36.92-second owned video while preserving the opening, identity, evidence, caveat, and ending. The episode inspects the removed example, renders the selected intervals with FFmpeg, and checks the 26.84-second export with the retained speech at its original speed and pitch.https://www.youtube-nocookie.com/embed/UYNAqwco9dE6432026-10-15T08:24:49-07:00https://itsaimathias.com/watch/gemini-skillshttps://itsaimathias.com/assets/gemini-skills.webpGemini Skills Are Rolling Out: Build a Reusable Weekly UpdateTwo small reusable skills, Evidence Brief and Weekly Update, organize fictional Harbor project notes without turning a pending release into a confirmed date. The episode reuses the method with a second week and improves the instructions to name the decision owner, using authored examples alongside Google’s product guidance.https://www.youtube-nocookie.com/embed/2wTdUs-NIpo6762026-10-15T08:23:24-07:00https://itsaimathias.com/shorts/short-20-booklets-or-80-the-cheapest-quote-changes-bag2Je_sznohttps://i.ytimg.com/vi/bag2Je_szno/hqdefault.jpg20 Booklets or 80? The Cheapest Quote Changes.At 20 booklets, Maple is cheapest at $150. At 80, Harbor is cheapest at $475. The assumptions changed, so the recommendation should change too. These are fictional printing quotes in our original local teaching tool. This is not recorded GPT-6 output or a model benchmark.https://www.youtube-nocookie.com/embed/bag2Je_szno362026-10-15T08:23:04-07:00https://itsaimathias.com/watch/n8n-agent-budgetshttps://itsaimathias.com/assets/n8n-agent-budgets.webpn8n Agent Budgets: Why an Alert Isn’t a Spending LimitA local budget replay reaches 12 teaching units against a nominal limit of 10 even though a warning arrived at 6. Mathias uses that timing problem to explain alerts, admission checks, reserved capacity, and final charges, then examines timeouts, paid failures, duplicate events, and an underestimated cost bound.https://www.youtube-nocookie.com/embed/xt3v-WjT1ps6942026-10-15T08:23:04-07:00https://itsaimathias.com/watch/haiku-switch-checklisthttps://itsaimathias.com/assets/haiku-switch-checklist.webpHaiku 5.5: Ask These 20 Questions Before You SwitchA 20-question document evaluation pack helps you decide what to test before switching to Haiku 5.5. Using fictional Northlight event notes, the episode covers missing answers, conflicting versions, misleading citations, and the cost of an answer you can actually accept; no Haiku inference or model comparison was run.https://www.youtube-nocookie.com/embed/jzCMtdYISOY8032026-10-15T08:18:36-07:00https://itsaimathias.com/watch/chatgpt-intelligent-uihttps://itsaimathias.com/assets/chatgpt-intelligent-ui.webpChatGPT Intelligent UI: Turn Long Answers Into Useful ToolsThree fictional printing quotes become a small decision tool whose recommendation changes with quantity, deadline, and priorities. The episode explains ChatGPT Intelligent UI, then uses original local HTML and 16 reported logic checks to show how an interactive answer can make its arithmetic and decision rules inspectable.https://www.youtube-nocookie.com/embed/IMIdMN_VXMs7572026-10-15T08:18:21-07:00https://itsaimathias.com/shorts/short-3-codex-agents-do-not-mean-1-finished-build-iYmSjl9ewXUhttps://i.ytimg.com/vi/iYmSjl9ewXU/hqdefault.jpg3 Codex Agents Do Not Mean 1 Finished BuildThree Codex CLI sessions worked in parallel in separate Git worktrees. Our test went from 3/21 to 8/21, 11/21 and 21/21 as we reviewed and integrated each change. Those are checks on our own small example, not a general Codex benchmark. The full walkthrough shows the prompts, changes and combined result.https://www.youtube-nocookie.com/embed/iYmSjl9ewXU362026-10-15T05:30:13-07:00https://itsaimathias.com/shorts/short-your-ai-passed-57-checks-and-still-failed-the-job-0OJ4-6pUCbAhttps://i.ytimg.com/vi/0OJ4-6pUCbA/hqdefault.jpgYour AI Passed 57 Checks—and Still Failed the JobA support follow-up loses its duplicate link. Our authored packet passes 57 of 58 checks, but still misses a requirement that matters. This is a deliberate local fault in our fictional teaching example, not output from a live Grok Bot run. The full walkthrough uses xAI's official guides to design a draft-only support workflow.https://www.youtube-nocookie.com/embed/0OJ4-6pUCbA392026-10-15T05:28:56-07:00https://itsaimathias.com/shorts/short-your-ai-app-looks-right-test-the-boundary-GOADUGGlM5ghttps://i.ytimg.com/vi/GOADUGGlM5g/hqdefault.jpgYour AI App Looks Right. Test the Boundary.Test nine, ten, and eleven units—not simply the happy path. A quick lesson from our Google Antigravity documentation-led tutorial using an owned synthetic demo. The known starter bug is intentional, not an Antigravity failure claim. Watch the full tutorial using the related video link. #Antigravity #AICoding #VibeCodinghttps://www.youtube-nocookie.com/embed/GOADUGGlM5g372026-10-15T05:27:37-07:00https://itsaimathias.com/shorts/short-your-ai-dashboard-shows-3-tickets-why-does-it-export-12-Wk-dpkXVcKchttps://i.ytimg.com/vi/Wk-dpkXVcKc/hqdefault.jpgYour AI Dashboard Shows 3 Tickets. Why Does It Export 12?The table shows three tickets, but the deliberately faulty starter exports twelve. This Claude Code workflow connects the visible filters to the actual CSV file. Original synthetic demonstration. AI presenter and synthetic narration. #ClaudeCode #AICodinghttps://www.youtube-nocookie.com/embed/Wk-dpkXVcKc392026-10-15T05:26:27-07:00https://itsaimathias.com/shorts/short-your-project-renders-that-does-not-mean-it-is-ready-ftHwF5vcwjIhttps://i.ytimg.com/vi/ftHwF5vcwjI/hqdefault.jpgYour Project Renders. That Does Not Mean It Is Ready.A page that renders can still have untested behavior. This genuine Codex handover example preserves the difference—and keeps the launch decision pending. Fictional teaching project. AI presenter and synthetic narration. #Codex #AIWorkflowhttps://www.youtube-nocookie.com/embed/ftHwF5vcwjI372026-10-15T05:25:03-07:00https://itsaimathias.com/watch/codex-parallel-worktreeshttps://itsaimathias.com/assets/codex-parallel-worktrees.webpI Gave Codex 3 Jobs at Once—Then Checked Every MergeThree Codex workers receive separate search, summary, and snapshot-restore tasks in a fictional task board. This actual CLI experiment uses manually created Git worktrees, then reviews each change as the shared suite progresses from 3 of 21 checks to 8, 11, and finally 21 of 21 after integration.https://www.youtube-nocookie.com/embed/PLpfavuYTnc10402026-10-15T05:24:19-07:00https://itsaimathias.com/watch/grok-ai-teamhttps://itsaimathias.com/assets/grok-ai-team.webpGrok Bot: A Practical Workflow for Your First AI TeamEight fictional support cases show how to design a useful first Grok Bot workflow around a draft packet someone can review. The episode separates incoming records, policy, specialist handoffs, and escalation, then uses three deliberately faulty packets to reveal why a high aggregate score can hide a critical failure.https://www.youtube-nocookie.com/embed/9n3bfsbk1VY10662026-10-15T05:23:54-07:00https://itsaimathias.com/watch/google-antigravityhttps://itsaimathias.com/assets/google-antigravity.webpGoogle Antigravity Tutorial: From Pretty UI to a Verified AppA fictional quote builder turns Antigravity’s planning and review guidance into an inspectable feature exercise. The episode follows a discount boundary at 9, 10, and 11 units, immediate input changes, saved drafts, and JSON exports, using an authored broken starter and working reference rather than a recorded Antigravity build.https://www.youtube-nocookie.com/embed/T4RamaMOfHc9552026-10-15T05:23:20-07:00https://itsaimathias.com/watch/claude-csv-exportshttps://itsaimathias.com/assets/claude-csv-exports.webpClaude Code: Build CSV Exports That Match Your DashboardA support dashboard shows three matching tickets while its seeded CSV export contains twelve. This authored teaching example turns that mismatch into a precise Claude Code feature brief, then checks composed filters, dates, quoting, and empty results with an independent Python CSV reader.https://www.youtube-nocookie.com/embed/DSjdgLrwu4k9962026-10-15T05:22:50-07:00https://itsaimathias.com/watch/codex-for-real-workhttps://itsaimathias.com/assets/codex-for-real-work.webpCodex for Real Work: From a Messy Folder to a Clear HandoverA real Codex CLI run turns a fictional website brief, decision log, and asset inventory into Markdown and JSON handovers. Mathias follows the output back to its sources and the 22 reported artifact checks, showing how to preserve owners, dependencies, deferred work, and decisions that still need approval.https://www.youtube-nocookie.com/embed/eG9VgKCOWV410052026-10-15T05:21:32-07:00https://itsaimathias.com/shorts/short-gemini-4-argon-worked-overnight-only-2-changes-shipped-Ai4vjKMfFxYhttps://i.ytimg.com/vi/Ai4vjKMfFxY/hqdefault.jpgGemini 4 Argon Worked Overnight. Only 2 Changes Shipped.Google’s reported Gemini 4 Argon ADK run proposed six optimizations. Two shipped after human review. What would you check before accepting an AI change? This is a source-based analysis, not our model benchmark. Watch the related full video for the context, three suggested evaluation tasks and an API cost example.https://www.youtube-nocookie.com/embed/Ai4vjKMfFxY342026-10-14T13:54:05-07:00https://itsaimathias.com/watch/gemini-argonhttps://itsaimathias.com/assets/gemini-argon.webpGemini 4 Argon Is Here—Here’s Where I’d Actually Use ItMathias examines Gemini 4 Argon through three jobs: improving existing code, organizing project documents, and interpreting a chart with a misleading detail. The episode connects Google’s engineering report and published pricing to suggested evaluation tasks, with clear separation between Google’s reported results and the channel’s synthetic examples.https://www.youtube-nocookie.com/embed/E9TUdATpG6Q4702026-10-14T13:53:06-07:00https://itsaimathias.com/shorts/short-24-green-checks-are-not-a-whole-app-guarantee-gZDxAnVwAishttps://i.ytimg.com/vi/gZDxAnVwAis/hqdefault.jpg24 Green Checks Are Not a Whole-App GuaranteeThe same local contract: original 24/24, deliberate control 16/24, Claude candidate 24/24. Check individual cases and keep the conclusion within the tested scope. Watch the full REA + Claude Code test: https://www.youtube.com/watch?v=RriVB3ln6ishttps://www.youtube-nocookie.com/embed/gZDxAnVwAis342026-10-14T08:32:19-07:00https://itsaimathias.com/shorts/short-rea-says-no-match-that-does-not-mean-no-feature-EPvhCYIacEIhttps://i.ytimg.com/vi/EPvhCYIacEI/hqdefault.jpgREA Says No Match. That Does Not Mean No Feature.A real REA graph query returned no-match. Another evidence path located the search function but retained five unknowns. Read the scope before you trust the result. Watch the full REA + Claude Code test: https://www.youtube.com/watch?v=RriVB3ln6ishttps://www.youtube-nocookie.com/embed/EPvhCYIacEI342026-10-14T08:25:41-07:00https://itsaimathias.com/shorts/short-your-ai-search-looks-right-until-this-query-4JECs7dk2Yohttps://i.ytimg.com/vi/4JECs7dk2Yo/hqdefault.jpgYour AI Search Looks Right—Until This QueryA basic search can pass the happy path and still lose Café when you type cafe. Here is the actual controlled demo and the matching rule that changes the result. Watch the full REA + Claude Code test: https://www.youtube.com/watch?v=RriVB3ln6ishttps://www.youtube-nocookie.com/embed/4JECs7dk2Yo422026-10-14T08:18:40-07:00https://itsaimathias.com/watch/rea-one-featurehttps://itsaimathias.com/assets/rea-one-feature.webpREA + Claude Code: Rebuild One Feature—and Prove It WorksAn owned JavaScript app becomes a focused reconstruction exercise: rebuild its search behavior and compare the candidate against 24 fixed cases. The episode follows REA findings, a bounded Claude Code request, accent handling, multi-word search, filters, and ordering, while explaining exactly what the reported result covers.https://www.youtube-nocookie.com/embed/RriVB3ln6is7752026-10-14T08:09:49-07:00https://itsaimathias.com/shorts/short-claude-code-auto-model-small-task-bigger-build-qsET-wo4Ga8https://i.ytimg.com/vi/qsET-wo4Ga8/hqdefault.jpgClaude Code Auto Model: Small Task, Bigger BuildA small task and a larger build can use different Claude models. This Auto Model walkthrough shows a working local routing rule, a manual-choice idea, and why total usage and caching still matter. Full five-mod guide: https://www.youtube.com/watch?v=qSeSasK7xB8https://www.youtube-nocookie.com/embed/qsET-wo4Ga8522026-10-13T09:17:33-07:00https://itsaimathias.com/shorts/short-muse-wonder-is-live-your-muse-app-may-still-be-waiting-msZKAuCm1Rkhttps://i.ytimg.com/vi/msZKAuCm1Rk/hqdefault.jpgMuse Wonder Is Live. Your Muse App May Still Be Waiting.Muse Wonder is available, but the Muse app update is a separate rollout. Wonder, Wonder Full and max reasoning effort are different choices. Full news breakdown: https://www.youtube.com/watch?v=_3LtxYRUO-Ahttps://www.youtube-nocookie.com/embed/msZKAuCm1Rk462026-10-13T04:35:37-07:00https://itsaimathias.com/shorts/short-your-claude-md-split-might-still-load-everything-mvURI1GMo-Uhttps://i.ytimg.com/vi/mvURI1GMo-U/hqdefault.jpgYour CLAUDE.md Split Might Still Load EverythingSplitting Claude Code instructions into more files does not automatically make them load on demand. Watch the @import trap and a clearer task-specific reference pattern. Full context engineering guide: https://www.youtube.com/watch?v=RsSiyoi4Azshttps://www.youtube-nocookie.com/embed/mvURI1GMo-U492026-10-11T22:09:34-07:00https://itsaimathias.com/watch/stop-overloading-claude-code-fix-your-claude-md-skills-RsSiyoi4Azshttps://i.ytimg.com/vi/RsSiyoi4Azs/hqdefault.jpgStop Overloading Claude Code: Fix Your CLAUDE.md & SkillsA local example shows how to remove duplicated Claude Code instructions while retaining the rules that matter. Mathias examines CLAUDE.md imports, task-specific skills, other sources of context, and the evidence passed from research to writing, with a clear distinction between file cleanup and measured model performance.https://www.youtube-nocookie.com/embed/RsSiyoi4Azs3532026-10-11T22:08:15-07:00https://itsaimathias.com/shorts/short-make-claude-code-show-the-evidence-before-you-launch-5DOJYKQLySwhttps://i.ytimg.com/vi/5DOJYKQLySw/hqdefault.jpgMake Claude Code Show the Evidence Before You LaunchA homepage screenshot tells you very little about whether the booking flow works. Ask your AI agent to test the customer's actual route: choose a service, submit the form, and check the result. Use an unavailable slot and an invalid email as example failure cases. Ask the agent to report what it changed, what it tested, and what remains unresolved.https://www.youtube-nocookie.com/embed/5DOJYKQLySw472026-10-11T13:06:17-07:00https://itsaimathias.com/watch/6-claude-code-workflows-you-should-actually-use-lzw4DksJiWIhttps://i.ytimg.com/vi/lzw4DksJiWI/hqdefault.jpg6 Claude Code Workflows You Should Actually UseSix workflow categories connect Claude Code to a concrete output: editable motion, connected data, better web experiences, project planning, app development, and browser testing. The episode explains where specialist tools fit and keeps the final question practical: what evidence would show that your chosen task works?https://www.youtube-nocookie.com/embed/lzw4DksJiWI5102026-10-11T13:03:57-07:00