Weekly Recap Scripting: Separating "Done" from "Next" with Claude
Agrima, a fellow at Humanitarians AI, asks Claude for a script that logs her week, then revises it to split finished deliverables from work still ahead. The video shows why a flat list reads like a highlight reel and how two lists fix it.
Most weekly updates read like a highlight reel, with the wins dressed up and the unfinished work blended in. In this video, Agrima, a fellow at Humanitarians AI, shows a more honest alternative. She uses Claude to write a small script that logs her week and then revises it so that what is actually finished stays apart from what has only just started.
Why weekly updates drift toward highlight reels
The video opens with the problem. A typical weekly update tends to present everything in the most flattering way. Agrima wanted something closer to a real log: what actually got done, kept separate from what is just getting started. That framing sets up the rest of the demonstration, which is about structure rather than wording.
Starting simple with a flat list
The first step is deliberately basic. She asks Claude for a script that simply logs the week, with one flat list and nothing fancy. The week has three items: an article published on Substack, four brutalist videos she produced, and an upcoming Suffolk University lecture with a presentation being built for it. Claude writes a flat list with three entries, no ranking and no spin. Running the script prints the week: the Substack article titled for the rise of faceless AI accounts, the four videos, and the lecture preparation.
The flaw in a flat list
The video then points out what is wrong with that output. All three entries carry the same weight. A flat list treats "I published an article" the same as "I am still building slides for a lecture that has not happened yet." The items are accurate individually, but together they blur the difference between work that shipped and work that is still ahead. A reader skimming the list could easily come away thinking all three are finished.
Revising the script to split done from next
The fix is a small revision. Agrima asks Claude to split what is actually finished from what is just starting, using the same week and the same three items but organized into two lists. Under "done" go the article and the four videos. Under "next" go the Suffolk University lecture and the presentation that is still coming together. When the script is run again, the shape of the output changes completely.
What the revised log says
The new output reports that the Substack article is published and that all four videos are finished in both formats. It then lists the Suffolk University lecture, with meetings underway, and the presentation still being built, under starting next. The narration makes the point plainly: nothing in the second list is finished yet, and saying so is the whole purpose. The result is one article shipped, four videos produced, and one lecture still ahead.
Key takeaways
- A flat weekly list gives finished work and future plans the same weight.
- Asking for a plain, unranked log first makes the problem visible before you fix it.
- Splitting the log into "done" and "starting next" changes what the reader understands about the week.
- Items that are still being built, such as slides for an upcoming lecture, belong in the "next" list.
- The difference between a recap and a highlight reel is a matter of structure, not tone.
- Revising a script with Claude is a quick way to test a different structure on the same content.
Try it yourself
Use the same prompt with your own week. Ask Claude for a recap script of what you did, then ask it to split the entries into what is done and what you are starting next. Compare the two outputs, and notice how much more honestly the second version describes what actually happened versus what you have only just started thinking about.
Chapters
- 0:00Why standard weekly updates feel like highlight reels
- 0:20Prompting Claude for a flat log of weekly tasks
- 0:45The flaw of flat lists: Treating finished work equal to future ideas
- 1:10Revising the script to separate "done" from "starting next"
- 1:35Tracking Substack posts, brutalist videos, and Suffolk lecture prep
Full transcript(auto-generated, with timestamps)
Why standard weekly updates feel like highlight reels
[0:00]Hi, I'm Agrama. Let's skip the highlight reel. Here's the actual log of what I did this week and what's just getting started. Most weekly updates are a highlight reel, the wins dressed up. I wanted something closer to a real log. What actually got done, kept separate from what's just getting started. Start simple. Ask Claude for a script that
Prompting Claude for a flat log of weekly tasks
[0:21]Just logs the week. The article, the four videos, the Suffolk lecture. One flat list, nothing fancy yet. Here's what it wrote, a flat list. Three entries, no ranking, no spin. Just published this, produced these, and what's coming up next. Run it and here's the week. No face, no problem, published on Substack, the rise of faceless AI accounts. Four brutalist videos produced
The flaw of flat lists: Treating finished work equal to future ideas
[0:46]Start to finish, all in both formats. And a Suffolk University lecture coming up with Yatra, plus the presentation being built for it. All three, same weight. That's the problem, though. A flat list treats I published an article the same as I'm still building slides for a lecture that hasn't happened yet. So, revise it. Split what's actually finished from what's just starting. Same week, same three items now in two lists.
Revising the script to separate "done" from "starting next"
[1:11]Done, the article, the four videos. Next, the Suffolk lecture and the presentation still coming together. Rerun it and the shape changes completely. Done this week, No Face, No Problem is published and all four videos are actually finished in both formats. Starting next, the Suffolk University lecture with Yatra, meetings are underway and the presentation is still
Tracking Substack posts, brutalist videos, and Suffolk lecture prep
[1:35]Being built. Nothing here is finished yet. That's the point of saying so. So, that's the real log. One article shipped, four videos actually produced, and one lecture still ahead of me, presentation and all. Separating done from next isn't a small detail. It's the difference between a recap and a highlight reel. Your turn, same prompt, your week. Ask Claude for your own recap script split the same way, done and starting. It's a small habit, but it keeps you honest about what actually happened versus what you just started thinking about. This week, what shipped, what's next? Logged for real, not hyped up.
More from Humanitarians AI Fellows
3:45Harshita Nimbal explains: Designing the Right Input for AI (Voice, Text, or Neither)
4:38Your Phone Can Do RAG — MiniRAG, Explained | Neelabh
3:55Harshita Nimbal explains: Human-in-the-Loop AI Design Pattern
4:02Simbaremuteuro explains: Prefill vs. Decode and GPU Serving Bottlenecks
2:21Harshita Nimbal explains: AI Harness vs. AI Model
4:33