People serving time in prison often tell Prison Professors that they feel forgotten. Many of them write every day anyway. They publish journal entries, book reports, course responses and release plans on their Prison Professors Profiles. Each entry is dated, and over months the entries add up to a record of how a person used their time. Participants earn points for that work, and a monthly leaderboard shows who has done the most.

I'm an engineer on the Prison Professors team. This month I worked on a project we call Profile Leaders. The goal is simple: notice people who are building that record and recognize them in public. The path to a working version was less simple. It ended up teaching me the same principles Prison Professors teaches the people we recognize.

Where the Project Started

When I took over the project on October 13, the plan was ambitious. Each day we would feature one top participant from the leaderboard. That meant a blog post, social media posts and branded images. It also meant a recognition letter mailed to the person, with a copy to their reentry coordinator.

I built the first version as a pipeline run by AI agents on one of our servers. I didn't trust the agents to get every step right, so I kept adding gates, verification checks and recovery routines. By the end, the project had about 14,600 lines of scripts and about 20,500 lines of tests.

Technically, it worked. The posts were another story. The writing was just not even close to what Michael Santos had in mind. The images looked plain and stock-like, and they came out in the wrong sizes. I had built a system that could prove each step had run. It couldn't produce work that sounded like Prison Professors.

Stepping Back to Define Success

The first principle in Michael's Straight-A Guide is to define success. On the Prison Professors blog, he warns that if you don't define success, someone else will define it for you. He also borrows a line from Stephen Covey: begin with the end in mind. I had defined success as a pipeline that never failed. Michael had defined it as recognition that sounds like our organization and honors the person's actual work.

On October 23, I stopped adding machinery. I reread what Michael had asked for and went through his spec with him one piece at a time. He didn't want a bot running the project. He wanted repeatable code on a server we own, running on a schedule, that we could keep training toward our brand voice.

That conversation shrank the project. Each choice became a plain rule that anyone could check:

- Choose one person at random from the top 100 of the monthly leaderboard.

- Include only people currently in a facility, and skip anyone we featured in the last 90 days.

- Write a post of about 300 words, in a narrative organizational voice, and say plainly that AI helped write it.

- Draw on the person's published work, link to their full profile, and end with a Lesson section and one challenge question.

- Post short teasers on X and Facebook with brand images sized for each platform.

Michael has spent months training his writing guide, which he calls the Master Writing Skill. We built it into the code as reference material. It tells the writer to show the reader what someone did, not to announce why it matters. It also lists phrases to avoid and forbids inventing scenes, quotes or achievements. The first version had worked around Michael's standard. The new version starts from it.

Learning From the Editor

The Straight-A Guide treats accountability as feedback, not punishment. In Lesson Seven, Michael describes keeping daily logs and tracking what he did rather than what he meant to do. We tried to build that same idea into how the system learns.

Every post gets its own Google Doc, shared with Michael. The post goes live first. If Michael adds a heading called "Revision" and edits the text, the system updates the blog post in place. It also saves his changes as examples for future writing. Over time, his edits become the training record.

The recognition letters follow a stricter rule. Each one comes from Michael. For the first 30 letters, the system holds each one until Michael marks it Approved or sends a revision. After that, letters go out on their own. Trust comes from a track record, for software as much as for people.

A Record Anyone Can Check

Michael has written about another community that has noticed this idea of a visible record. After CZ, the founder of Binance, publicly supported Prison Professors, members of the BNB Chain community formed their own independent group. They launched the Prison Professors community token and PrisonProfessorsToken.com, and part of each transaction goes to a donation wallet. Michael has said publicly that the group is separate from our nonprofit. He has also pledged to keep the more than 942 BNB it has sent our treasury in cold storage until at least the summer of 2027. In his words, the nonprofit will not spend resources simply because they are available.

The crypto community understands things it can verify. A blockchain keeps a public, dated history of transactions that anyone can inspect, and it doesn't ask anyone to take a promise on faith. In a recent post about Web3 education, Michael used a heading that bridges the two worlds: "A Profile Can Become Proof of Work." A profile doesn't guarantee release, a job or any other outcome. It does give employers, mentors, families and advocates something to evaluate besides a person's worst decision. They can see the work.

Profile Leaders tries to read that record carefully and recognize it in public. The posts quote no one who didn't write the words, and they credit no one with work they didn't do. Each one links back to the source, so readers can check it for themselves.

Where We Are Now

On October 23 we published the first five posts:

• Art Wood

• Tong Hyon Suh

• Quinton Handlon

• Evan Matthew Lawbaugh

• Daniel Lloyd Avery King

Art Wood's post, for example, follows his September journal. It covers welding and plasma-cutting assignments, two hours spent setting up a CDL driving simulator, and his preparation to teach a class. Every detail comes from what he published.

Michael reviewed all five posts, and he loved them. The schedule now runs every day at 9 AM Central. We're finishing the recognition letters, and mailing will begin soon.

The system still has a lot to learn, and Michael's revisions will show us where. In a post this summer, Michael summed up his own plan for the coming months as "build 'n build." Our plan is similar: one post at a time, one edit at a time.

Lesson

In the first version, I mistook effort for progress. Thousands of lines of code showed that I had worked hard, but they didn't show that I had built what we needed. Progress started when I went back to the definition of success. I simplified the plan, made each piece checkable, and let the editor's feedback guide the next round of work. People building a record in prison work the same way: one dated entry at a time, measured against the life they're preparing to lead.

Challenge question: Think of something you're working hard on right now. Have you defined what success looks like at the end, and what is one piece you could simplify this week so your progress is easier to see and check?