Horus is a free web tool that shows a student, course by course, how artificial intelligence is reshaping the value of what they're learning. Most students put four years and a lot of money into a degree with one quiet worry running underneath it: will AI make parts of what I'm studying obsolete before I even graduate? The public conversation only swings between hype and doom. We're building the calm, evidence-based answer.
What does it do?
A student picks their degree program. Horus breaks it into the specific skills their courses are built to teach, then gives each one an honest verdict:
| Verdict | ||
|---|---|---|
| Automating | Transforming | Untouched |
| Machines increasingly do this. Your role shifts to directing and checking the work. | You will do this alongside AI. The skill worth building is the augmented version. | A durable human edge AI is not reaching. Worth protecting. |
What Horus never does is tell a student their major is doomed — because the data doesn't say that. Most of what AI does today augments human work rather than replacing it, and the product is built to reflect that, not to scare people.
Where do the answers come from?
The verdicts aren't opinion. They're grounded in real data on how people use AI across thousands of actual work tasks — the open Anthropic Economic Index — cross-referenced with the U.S. Department of Labor's O*NET skill framework and mapped onto a program's real curriculum. The result is a now-cast: an honest read of where things stand today, shown with its limits visible — not a crystal-ball promise about the next decade.
What shipped?
Horus reads 11,320 course entries across 285 degree plans. 7,604 of them publish learning outcomes, and those are the only ones it draws conclusions from — a course without published outcomes is shown in the plan and marked as unread, never guessed at.
From those it identifies the 34 O*NET work skills the degrees actually teach, and writes 5,989 distinct lines of course-specific advice — not a template repeated, but a sentence grounded in what each course has students do.
Each degree gets a plain answer to the question the site opens with. On the shipped data, 103 degrees read "Yes" and 182 read "No" — and the "No" is never left to sound like "untouched", because every one of those degrees still teaches skills that change shape.
Built to be checked
It publishes its own error rate. An independent check re-reads a course's outcomes and one skill matched to it — without seeing the original reasoning — and judges whether the match holds. On a random sample of 1,000 pairings: 55.5% upheld, with a 95% confidence interval of 52.4–58.6%. That number is on the site, not buried in a report.
It says which verdicts are shaky. Every skill's score was recomputed 300 times on different samples of the jobs behind it. Nine of the 34 land on a different verdict more than once in ten runs, and each of those says so on the page rather than presenting a coin flip as a finding.
It says what a person has and has not checked. All 34 skill verdicts were reviewed by a human, and the course ranking was checked across 27 skills. The tagging beneath that was measured by sampling, not reviewed — and the site states exactly that distinction.
Five degrees get no verdict at all. Where fewer than three courses publish outcomes, Horus says "We can't say yet" and shows what it could read. A guess dressed as a reading would be worse than no answer.
How was it built?
- Scope: every undergraduate program in the Texas A&M catalog — 285 degrees across 13 colleges, each read course by course.
- Pipeline: scrape the public catalog and course syllabi → match each course's published learning outcomes to O*NET work skills with the Claude API → score those skills against measured AI usage → generate a page per degree. Nothing runs while a student reads; every page is a file built in advance.
- Runtime: static export, no database, no server, no account, no uploads.
- Ship it: DigitalOcean, with an automated accessibility and end-to-end test gate on every change.
Stack: Next.js · React · TypeScript · Tailwind · Python · Pydantic · Claude API
The payoff
Agency. Instead of guessing, a student comes away with a concrete sense of what to focus on, which AI skills to pick up alongside their major, and where their lasting advantages lie — turning a vague fear about the future into informed choices they can make right now.
How it’s built
One cohort. One ship.
Help us maintain Horus.
Horus is free to use — no ads, no accounts, nothing to upload. Donations cover the hosting it runs on and the upkeep that keeps its readings current.