New here, so here is what you get. Three things. One. How to put AI to work in your business and your life. Two. How to run business software and automation, mostly ClickUp. Three. How I run my own week, which I call mindful productivity. Follow if that is your problem.

Your static files go in Markdown. Mine is 1,935 of them. A client status that changes every hour doesn't become file 1,936. It goes in a Supabase database.

I went for a run and Claude edited 10 Shorts. It opened several video editors at once, one per Short, and made the cuts. The retakes are gone. The motion graphics are in. Here's how you can set up the same thing for your own Shorts: upsys-consulting.com/en/edit…

Claude Code will break your proposal template one day. I'd make a free GitHub account, turn the client folder into a private repo, then commit and push. A commit is a photo of your files, a push sends those files to the cloud, and you can go back to the last photo in one step.

Turn on time tracking in a ClickUp space and every member can log time on every task and subtask. You can't restrict either one. I'd send each entry to a webhook and let an outside automation cancel the ones that break your rules.

Claude / pinterest is a sick b-roll finder!

I built a concept called the playground for AI. Every single complex AI task starts with an HTML file with visuals, a diagram, a task list. It then gets a status open in progress done. Makes it more visual for me and allows easy handoff between sessions. The prompt to make your own: Build me a "playground" skill for my coding agent. Here's the idea. Every complex AI task starts with one HTML file: a summary, a visual, the work done, and what's next. Each file has a status (open, in progress, done). A hub page lists every board in a sidebar, grouped by status and project, so I can see at a glance what's open and hand a subject from one session to the next. The next session has only the file, so everything it needs has to be on the page. Make these 4 things: 1. A skill file (SKILL.md, or whatever my agent uses for skills) named `playground`. It triggers before any complex task, on "handoff", and on "clean the playground". It holds the rules below. 2. A template: `playground/_template.html`, one self-contained page with its CSS inline. Dark, one column about 1070px wide, phone friendly at 390px. Every board copies it, so they all look like one product. 3. A hub: a small script (Node, no dependencies) that scans every `playground/` folder in my projects and writes `~/playground/index.html`. It shows a sidebar (status counts, then projects) and a grid of cards with the title, project, date and status. It opens on Open boards by default. 4. A check script that fails a board with no status, more than 600 words on show, or a section with no visual. The rules for every board: - One subject a file, named `YYYY-MM-DD-slug.html`, saved in the repo's `playground/` folder. - A status in the head: ``. Allowed: in-progress, blocked, ready-for-review, done, reference. - The same frame, in this order: a header with the answer in one sentence, then up to 7 sections with one claim each, then "Next" as a checklist (skip it only when done), then a collapsed "More" for evidence and sources. - I look, I don't read. Each section gets something to look at: an inline SVG diagram, a table, bars, a timeline. Keep it under 600 words. - Headings are claims ("The API is the bottleneck"), never labels ("Performance"). Reading only the headings should tell the story. - Decisions, ids, links and file paths sit on the page. Never "as discussed". - Pick the shape by what I'll do with it: understand a thing (diagram), choose between options (a table with one pick), see data (chart), try a screen (mockup), or pick work up later (timeline of what's done, then the checklist). - Real values only. No "Option A" and no placeholder text. Workflow: write the question and the one-line answer first. Copy the template, draw before you write, run the check, rebuild the hub, then give me the file:// path. When work moves forward, update that board's status and checklist instead of making a new one. Build it, make one example board about this setup itself, rebuild the hub, and open it for me.

Getting my priorities straight

Vibe catting

I notice sometimes it’s better to ask Claude to redo something that was done by another model. You get better results. I had Gemini flash do a motion design, then ask Claude Opus to completely elevate it, and it did this. Way better result than when I asked it to directly do it