Addams — Your Humanitarians AI Sherpa
Addams is a weekly learning-documentation system for Humanitarians AI OPT volunteers, built to capture the friction and problem-solving that a finished artifact no longer proves happened.
Generative AI can produce a polished report, a clean summary, or a well-structured essay in seconds, without doing any of the cognitive work those documents were originally meant to prove happened. That single fact broke something quietly important in how learning gets documented, and it is the exact problem Addams was built to solve.
The problem: the artifact no longer proves anything
Before AI, a well-written report was reasonable evidence that someone had worked through a problem, made decisions, and understood what they produced. Now a system that has done none of that cognitive work can generate the same polished output. That means the artifact, the article, the tool, the write-up, no longer proves the learning happened. What actually proves it is something different: the friction you worked through, the moment you caught an AI output before it compounded into a wrong decision, the point where you had to reformulate a problem because the one you were handed was not the right one, the thing only you could supply. Most documentation systems do not ask for any of that. Addams does.
Why it is named after Jane Addams
The tool takes its name from Jane Addams, who founded Hull House and won the Nobel Peace Prize, and who built the infrastructure that turned good intentions into lasting social change. The naming reflects a specific belief: good work without documentation disappears. Addams exists to make sure that the real work Humanitarians AI's OPT volunteers do each week does not vanish simply because the final artifact looks complete on its own.
How the weekly workflow works
Addams is set up as a Claude project, with the system prompt and documentation linked from the Humanitarians AI website. At the end of each week, a volunteer types "/hai" to begin a structured intake covering their objectives, work sessions, hours, artifacts, and Substack publication. The most important part of that intake is the section asking specifically about the moments where a volunteer had to supply something the AI could not. The tool is built to account for 20 hours of degree-relevant work per week, and it enforces gates: it flags hours below 20, and if the friction section comes back thin or empty, it asks again, because real work every week involves genuine struggle, and naming that struggle is treated as evidence rather than a weakness to hide.
Who reads the report
Each weekly report is structured for three specific readers: the volunteer themselves, their program coordinator, and Humanitarians AI management. That three-way structure is deliberate. It is not a compliance checkbox; it is meant to make the report actually mean something to each of the people who rely on it, from the volunteer reflecting on their own progress to the organization tracking what its OPT program is actually producing.
Key takeaways
- Addams is a weekly documentation system built specifically for OPT volunteers at Humanitarians AI.
- It exists because generative AI decoupled the polished artifact from the learning it used to prove.
- Filing a report starts with typing "/hai," which walks you through objectives, work sessions, hours, artifacts, and Substack publication.
- The tool enforces gates: it flags hours under 20 and pushes back on a thin or empty friction section.
- Reports are structured for three readers: the volunteer, their program coordinator, and Humanitarians AI management.
- The tool is named after Jane Addams, reflecting the idea that good work without documentation disappears.
Frequently asked questions
How do I file my weekly report? Set up Addams as a Claude project using the system prompt linked on the Humanitarians AI website, then type "/hai" at the end of each week to begin the structured intake.
What happens if I don't log enough hours or skip the friction section? Addams flags hours below the expected 20 per week and will ask again if the section on struggles and judgment calls comes back empty, since that friction is treated as the actual evidence of learning.
Who this is for
Addams is built for OPT volunteers working on Humanitarians AI projects who need to file weekly learning documentation. Anyone starting a new role should set up their Claude project and file their first /hai report by the end of their first week.
Full transcript(auto-generated, with timestamps)
[0:00]Hey, Adams here. You've just joined the Humanitarian's AI project. You have a role, a team, a deadline. And at the end of the week, someone is going to ask you to document what you learned. Here's the problem. You're probably going to write about what you produced, the article, the artifact, the tool you built. And that is going to feel like enough because it looks like enough. It isn't. That's not a criticism. It's a structural problem, and it's the reason this tool exists. Humanitarian's AI is a bridge education program. The product isn't the artifact. The product is your learning. And here's what changed. Generative AI permanently decoupled those two things.
[0:43]A polished essay, a well-structured report, a clean summary. These can be produced in seconds by a system that has done none of the cognitive work those documents were designed to evidence, which means the artifact no longer proves the learning happened. What does prove it? The friction, the confusion you had to work through, the AI output you caught before it compounded into a wrong decision, the moment you reformulated the problem because the one you were handed wasn't the right one, the thing you had to supply that the machine simply could not. That's the learning record, and most documentation systems don't ask for it. This one does.
[1:16]Bring up the name calm, not promotional. This tool is called Adams, named after Jane Addams who founded Hull House, won the Nobel Peace Prize, and built the infrastructure that turned good intentions into lasting social change. She did it because she understood something most people still don't. Good work without documentation disappears. Adams is your weekly documentation system. It's built specifically for OPT volunteers at Humanitarian's AI, and its job is not to make your reports look better. Its job is to make your reports mean something to you, to your program coordinator, and to the project you're actually contributing to. Every week, Adams asks you to account for 20 hours of degree-relevant work,
[1:51]Not in a compliance box, in a record that names what you did, what you struggled with, what the AI couldn't give you, and what you're still figuring out. Because that record, that's the proof of work that matters. Keep this tight one workflow clearly stated. Here's how it works in practice. You set up Adams as a Claude project. The system prompt is in the link below. At the end of each week, you type "Hey Adams." Walks you through a structured intake. Your objective, your work sessions, your hours, your artifact, your Substack article, and most importantly, the moments where you had to supply something the AI could not.
[2:22]It asks specifically. It pushes back on finances. And it produces a weekly report structured for three readers, you, your program coordinator, and Humanitarian's AI management. It also enforces gates. No artifact documented? It flags hours below 20. It helps you surface the time you spent but didn't write down. Fiction section empty? It asks again. Because every week of real work has struggle, and naming it isn't a sign of weakness. It's the evidence. Direct one action. No hedging. Your first Hey report is due at the end of this week. Set up the Claude project now. The system prompt and full documentation are linked below. Then run. Onboard to build your profile
[3:04]And file Hey when your week closes. The record starts this week, not next week, not when you feel more settled. This week. Because a week without documentation is a week that, for the purposes of this program, did not happen. You're doing real work on real problems. Make sure there's a real record of it. Don't forget to like, subscribe, and hit that bell. Thanks for watching the Humanitarian's AI YouTube channel.
More videos
2:08Bridging the Pixel Gap in Browser Automation.
2:23How One Narrow Safety Rule Can Make an AI Less Safe Everywhere Else.
2:04Why splitting a chunk from its document makes it retrieve for the wrong question
4:20Three You Can Take Back. One You Can't.
2:21Why a 50-turn agent pays for the same screenshot 35 times unless it caches the pixels
1:53