Workings: show what’s yours in the age of AI
Capturing your processes for peace of mind
Workings is a private record of human authorship. It is a local-first provenance and proof system that cryptographically-verifies the creation process of how your work was created digitally, without exporting the underlying content.
Will Griffiths, Founder of Workings wanted to build a,
"totally private, high-fidelity record of your process to protect your work and give you peace of mind in the age of AI".
To bring it to market, he reached out to Mechanical Rock, combining his background in product, cybersecurity and engineering, with Mechanical Rock’s software engineering capabilities and agile way of working.
The rise of generative-AI and the loss of trust
AI can now write an essay, book or report in seconds, and suspicion has hardened into blanket mistrust. Teachers don’t know if students are learning, clients don’t know if real effort went into the work they have commissioned, whether it represents original work or is infringing on IP rights, or whether AI-assisted work has been appropriately supervised.
AI-detectors came onto the scene, claiming to be the fix, but people very quickly learned that, alone, these tools can only guess where and how AI has been used and otherwise are unable to show the effort that went into producing a piece of work.
“Creators” are encouraged to maintain a record of how their work is developed, which is a laborious and flawed process in many cases; it’s problematic to create a full body of evidence, and evidence is easily fabricated.
Will shared, “It became clear that the key to addressing this was being able to show your workings. For that we needed a private, yet verifiable, record.”
Workings wasn't designed as a stance against AI. “It’s built to objectively describe the process behind the work: was it human, AI or both?”
From validated concept to a product ready for scale
The Workings Team had validated the approach with a proof-of-concept, user testing, a filed patent and a technical whitepaper. Taking it to market at scale required engineering. The solution had to deliver deterministic, independently-verifiable proof of authorship, generated entirely on-device so raw data never left the user's machine, at a unit cost that worked for a consumer subscription.
Mechanical Rock was engaged to deliver Workings across three surfaces: a cross-platform desktop application that captures working processes locally with on-device inference, a cloud backend for authentication and reporting, and a web portal for users and administrators.
There was no template for a product like this so Will explained how the two teams approached it:
“We documented everything from day one. With a product changing this fast, that’s the only way to stay ahead of tech debt.”
Designing for evidence without exposure
Mechanical Rock built Workings around five layers of architecture, each one designed to protect the user while strengthening the evidence.
Starting with capture, Workings takes periodic screenshots and timing data to show how someone works, recorded as evidence of authorship in sessions the user starts explicitly. When AI is used, it also records the tools and time taken for each session, so the record doesn't just say a human made this. It can prove, "this was made under these declared conditions, including the assistance of an AI model", so the workings and the tools used are attested rather than hidden.
The first layer handles capture. We built Workings to take periodic screenshots and timing data showing how someone works, recorded as evidence of authorship in sessions the user starts explicitly. When AI is used, it also records the tools and time taken for each session, so the record doesn't just say a human made this. It can prove the work was made under declared conditions, including the assistance of an AI model, so the workings and the tools used are attested, not hidden.
Each element captured is then sealed and becomes a cryptographic, verifiable commitment with time anchors using SHA-512 hashing, chained records and externally-anchored timestamps to ensure the record can’t be modified.
On top of that, we built a human signal analysis layer that checks the evidence against a set of human-plausibility constraints, looking for patterns consistent with human ways of working. These signals are captured within the record too, but the architecture is built so they can never be exposed or exported.
The fourth layer, was one of the harder engineering problems: making it possible to verify someone's work without disclosing raw data. We built Workings to generate commitments spanning a specific time frame, so an external verifier can confirm the work without ever seeing the underlying content or behavioural data.
Lastly, we anchored each sealed record to the blockchain, using cryptography to make the process time-sealed and tamper-evident, so it functions as an immutable public ledger.
Privacy first and foremost
Privacy shaped every decision in the build. Will explains why capturing the process itself, rather than analysing the output, was the only reliable path:
“If you want real evidence of how your work was made, you have to look at the screen. That only works if the record is completely private, so that’s where we started.”
The resulting privacy architecture keeps evidence on-device, with only cryptographic fingerprints stored in AWS. As Will puts it, “Workings stands and falls by its privacy. We don’t have to reveal anything but we can use this method to make true attestations about it.”
Using AI in delivery to keep up with AI
The project ran on practices that let a small team move quickly without accumulating risk. AI was a big part of how quickly Workings came together, but speed was never the point on its own. Mechanical Rock used AI inside the same disciplined engineering practices it brings to every build.
Will was glowing in his praise of the team:
"Dik and Brandon were legendary to work with, their can-do attitude and deep knowledge meant we could iterate at pace. Fast cycles.”
Within six months, Mechanical Rock and the Workings team had delivered a novel local-first concept, cryptographically-verifiable proof of how work was done, from a blank repository to a functioning, installed Alpha product shipped to market. Three months later, Workings has 100 closed alpha signups with users spanning big tech, academia, writers and more.
Sign up and learn more at workings.io
How can Mechanical Rock help with building your product?
At Mechanical Rock, we help companies deliver cloud-native solutions, leverage data platforms and unlock AI so systems can run securely, reliably and at scale.
Whether you're taking a new idea to market or scaling an existing product, we can help.
More from our work

Product Development
CareCorner: Bringing peace of mind to families through technology
At Mechanical Rock, we believe technology should serve people - not the other way around. One project that truly embodies this philosophy is CareCorner, a digital care coordination platform designed to help families, carers, and medical professionals work together seamlessly.


Product Development
Snowvault: Turning fragmented enterprise documents into a governed AI knowledge platform
Snowvault connects to a customer's existing document repositories, email systems and data stores, and transforms that content into a live, searchable, AI-powered knowledge layer that supports structured workflows, specialist agents and auditable work outputs.

Product Development
Rio Tinto: Accelerating rail maintenance data processing with the RSM Scanner
Rio Tinto engaged Mechanical Rock to automate and accelerate the processing of paper-based maintenance work pack data using AI/ML learning, unlock data from historical archives and design a solution to meet their future data processing needs.