Closing the Effort Gap
The Contract
Education has historically operated on the simple assumption that quality output implies proportional effort. The essays, exams, and problem sets were proxies. We couldn't always observe the student's process, so we graded what they produced. This was never provably true, but it was pragmatically workable. The labor required to produce a passing essay was roughly uniform; the ladder from blank page to submission had rungs you could not skip.
Generative AI removed the rungs.
A student can now move from a blank page to a polished submission in seconds. The output is indistinguishable from work that took days. The proxy has collapsed. When we grade artifacts that require zero labor, we aren't measuring learning, we are measuring prompt proficiency.
This is not an ethics failure, It is an information failure. The signal we relied on (ie. the correlation between output quality and input effort) no longer exists.
The Effort Gap
Two students. Same assignment. Same submission.
Student A spends six hours: researching, outlining, drafting, revising, struggling, rewriting, submitting.
Student B spends thirty seconds: typing a prompt, copying the output, submitting.
The outputs are identical. The grades are identical. The learning is not.
This is the Effort Gap. It is invisible in the final output but it is undeniable in the process. And because we stopped examining process centuries ago, we now have an educational infrastructure blind to its own obsolescence.
The market has responded with detection software. This was predictable and wrong. Detection attempts to infer the gap from the output alone. It fails because the question is unanswerable from the available data. You cannot reconstruct processes from products. You can only guess.
Guesswork is not a useful protocol.
Verifiable Effort: Git for Text
Verifiable Effort is a different primitive: the journey is as real as the destination.
When an author writes inside an environment, we capture the structural history of composition. We record structured, timestamped metadata: session duration, revision distance, and the ratio of human-directed to machine-assisted composition. Software developers have worked this way for decades. Git didn't make programmers more honest and no one "detects" if a developer wrote their code, but they can inspect the commit history.
The log is the proof. TWFF brings this same primitive to text
The Process Transcript
A grade transcript tells you what happened. A Process Transcript tells you how.
For educators, this is diagnostic. It separates brainstorming assistance from wholesale generation.
It reveals which students need help with structure versus mechanics and makes intervention possible. For students, this is certification. The Process Transcript is a cryptographic artifact they control. They present it to verify that their work is their own. It transforms the burden of proof from a defensive struggle into a personal asset.
For institutions, this is evidence. When stakeholders ask "how do you know the AI isn't doing the work?", the answer is not policy. It is data. Here is the transcript. Here is the effort. Here is the proof.
The output was never the point. The point was what you did to get there.
How to Get Involved
The TWFF specification is open source and available on GitHub: firl.nl/twff
We are actively seeking collaborators to implement the TWFF standard across writing platforms, educational institutions, and research projects. If you are interested in contributing to the development of TWFF or integrating it into your tools, please reach out to us.