0:46Gemini 4 Carbon Is Official — But There’s a Catch
A source-attributed explanation of Google’s announcement, with attention to access and verifiable work.
24 ways into the workflow. Pick a video to see what was tried, what to check and where to begin.
0:46A source-attributed explanation of Google’s announcement, with attention to access and verifiable work.
3:28A source-attributed explanation of Google’s announcement, with attention to access and verifiable work.
12:25Four practical Google AI examples: email drafting, project spreadsheets, an editor handoff app and a test of its published behaviour.
18:39Compare seven Google programmes by goal, workload and practical AI content.
13:48Compare model answers, inspect disagreements and keep imported memory and pricing decisions separate.
8:19Prepare a provider fallback, preserve project context and inspect usage before resuming a large task.
23:09Compare three levels of direction on the same SaaS page and separate visual judgement from integration work.
8:30Use Luke’s course takeaways to test spreadsheet analysis, source-based research and a small reusable AI tool.
16:50Review the flight simulator and SaaS feature separately, recording failures and usage beside completed work.
18:49Connect research, carousel writing, image generation and PDF rendering, then review the emailed package before posting.
7:38Inspect an existing-codebase change through diffs, tests, source control and review before expanding to parallel agents.
7:57Choose an app task and model, check its AI inputs and inspect the deployed version separately from the builder preview.
19:20Review instruction following, realism and consistency without turning one attractive output into an overall model ranking.
19:26Connect video analysis, dynamic thumbnail rendering and a deliberate upload check before adding an RSS trigger.
17:14Compare five task types without turning one impressive output into a universal model ranking.
14:55Compare the Browserbase demo, local preview repository and API route, keeping the place where the tools run distinct from model access and cost.
12:50Build the recorded Gemini voice assistant, then inspect function calls, language changes and the attempted email action separately.
14:59Refine the app rules, connect Gemini Flash and test the roast behaviour before deploying the reviewed build.
16:20Follow the Rork plant-identifier build through phone testing, Gemini image analysis and the final identification example.
20:10Extract the transcript, generate a draft and carry explicit feedback through to a checked metadata update.
15:51Connect Perplexity and Gemini 2.5 in n8n, map the actual chat input and inspect the research output’s references.
25:25Connect a model and Gmail tools, then test reading, drafting and an explicitly approved send as separate behaviours.
10:51Connect Dart’s project context to the coding agent, inspect generated tasks and compare their implementation with the plan.
51:05Trace a document from ingestion through embedding and retrieval to the evidence used in a chat answer.