Project Log

Notes and updates as the research proceeds

3 October 2026 · Building the notary

I built a small tool this week. It has been on my mind for a while, and the trigger was a conversation about precedence.

Collaborative work has a particular fragility. Texts move between people. A draft is shared, revised, folded into something else. By the time a work has travelled even a short distance from its origin, the record of who made what tends to dissolve. What makes it difficult to reconstruct is that a document carries no evidence of its own history. A PDF or a Word file can be renamed, edited, resent, saved in a dozen places, and nothing in the file itself records any of it. Whatever trace exists lives elsewhere.

The notary separates two things: being on record, and being public. Normally, establishing precedence means publishing. However, publishing may not always be a commitment you want to make. Here, the file is fingerprinted in the browser, and only the fingerprint goes on-chain, with a short description and the date. The document stays with me, and whether it ever becomes public is a decision I make later. If I disclose it, the record shows that what I disclose is what was sealed.

This is close to the Eo1 register I built last month, and uses the same infrastructure (EAS on Celo), but it points in the opposite direction. The register releases works into the commons; the notary holds a private claim, in reserve. However, both rest on the same principle: that a claim about what something is should be verifiable. That the same primitive serves both is one of the things I have found most interesting in building it.

Which brings me to the heartbeat, which has been running on this site for a few months now. It works on the same principle. Once a day, a small program looks at my website, computes the fingerprint of what it sees, and signs a statement, which it then publishes to a public log. The program has its own key and its own identity. It has one job: to notice what the site looks like, once a day, and to say so on record.

The notary extends this pattern from a website to a document. In both cases, a small agent signs a record of something that existed, so that the existence can be checked later.

22 September 2026 · Building the EO1 register

This past month I accepted the challenge to develop a prototype infrastructure and Web3 website for BLACKBOX. As documented elsewhere on this site, the intention had always been to set up a small experiment in parallel to the Amicus Curiae 8:2 Law and Cultural Production Special Section. For this, I envisaged something self-contained such as tracking, via smart contracts, a work's progression from initial creation through remix and editioning. This setup would have (helpfully!) allowed for the deferring of tricky questions regarding how the micro project would be owned, governed and situated within a non-proprietary framework.

In the end, discussions led to a different challenge: the development of an Edition of One website of which BLACKBOX would be a part. Edition of One is an organisation set up by Gavin Keeney with an existing Web2 presence. It is described as a borderless agency for works of artistic-literary scholarship, which adhere to the "No Rights" ethos of elective renunciation of IP rights and which commit to the cultural commons for the common good.

This immediately brought issues related to governance back into the frame. Eo1 in this structure has ultimate oversight of BLACKBOX. How, then, could a commitment to a non-proprietary record of cultural works (the commons) be given expression without contradiction. This was my most significant concern, and introduced the first and most fundamental decision in the design build. How would the infrastructure allow Eo1 the freedom to take up, sponsor and publish cultural works in a way that allowed the infrastructure itself to exist in common? It seemed to me that the risk of building something that became just one more proprietary publishing venture was very high. Even worse, there was a risk of doing so by the back door, thus setting sail under a false flag.

I also realised that it would be necessary to carefully document the design process so that when/if the collective took it on, there would be clarity about what precisely was at stake at each fork in the design tree. Similarly, my process needed to be transparent, subject to scrutiny and revision or correction. Or, if you will, the build itself needed to practise the ethos and be capable of remix, revival and editioning without losing its original attribution. My working formulation was this: Eo1 is not a better IP registry. It is a register that refuses to be a registry at all and records itself in the process.

As part of the (intensive) work I kept copious notes. Some of those I summarised and placed on the resulting eo1collective.org website in the structure page. For this, my personal site, I thought it would be instructive to visualise the design process in diagrammatic form which is available here: The Register in Diagrams. The entire development process was interesting to me, and personally valuable, as it threw a spotlight on the necessity that I, as developer, walked the talk in how I offered up the final work, and that the design itself needed to embody the argument it sought to iterate. The first iteration of the site is attested on-chain in the EAS schema I registered on Celo network. It means that the register I built records itself in the register. In the related graph view of the attestation records, the graph includes itself on the graph. The playfulness of this act of recursion delighted me, but it also reflects my serious intent to be mindful about the act of creating this work.

I am aware that idealism does not always survive practical implementation. If, at a future date, the collective decide to become something else, or if a future developer decides it is better to abandon the non-discretionary principle of the register, or if it becomes apparent that I was fundamentally at error in my philosophical grounds, then the origin position will remain on record. What comes after, I do not know, as in many, many ways what is after is not mine.

The design log is at Eo1 Collective Structure. The site is at Eo1Collective.org. The register and the graph are live.

A copy of the design log as it stood at handover, together with the register entries, is preserved in the Commoning cluster at commoning.amykellam.com.

25 August 2026 · On language, misunderstanding, and the problem of translation

Recently I co-authored a proposal (with Dr Gavin Keeney) for developing a dApp that implements the BLACKBOX model. The basic idea is to build a flexible decentralised application that allows creators to publish a creative work to a digital archive and register their authorship (or status as originator) on a blockchain. Unlike existing blockchain-based IP projects that register works as ownable assets for licensing and monetisation, the foundational concept of the BLACKBOX app is that it registers attribution without conferring ownership. It is rooted in a cultural commons ecosystem: use of a work is permitted, but use generates an obligation to the originators and the cultural commons.

The principle that accessing and engaging with a work generates an obligation is based on a model of stewardship — that individuals bear a responsibility to the cultural commons in a similar way to how they bear a responsibility to environmental commons such as oceans and forests. Following from this is the principle that this responsibility might be expressed in various ways and requires a design flexible enough to encompass this. This might be expressed on-chain in multiple ways depending on the individual project: for example, through a monetary contribution, or through a creative or intellectual contribution such as a re-edition, revival, or remix of the original work. The foundational blockchain infrastructure should enable each project community to manage its own protocol, setting its own rules for recognition, distribution, and participation. In other words the aim, fundamentally, is to build something in which the form adapts to, and flows from, the content.

Feedback from the collective has been useful, in part because of what it reveals about the difficulty of communicating what BLACKBOX might be. One response raised questions that circled around the idea that the application was a "rule book" or "constitution" — a framing that does not quite capture what a decentralised application is doing, but one that is entirely understandable. If you are used to thinking of governance as a set of rules imposed by an authority, then a registry and a governance protocol look like instruments of rule-making.

The terms used in the proposal (protocol, registry, treasury, governance) carry different inflections in different contexts. In computer science, a protocol is a set of rules for communication between systems. In international law, a protocol is a treaty that amends or supplements an existing agreement. In everyday language, governance refers to the exercise of authority. These are not the same thing, but they are close enough to cause confusion.

To give a sense of what a dApp might look like in practice: an artist submits a work to the BLACKBOX registry. The registration records their originator status — who made what — and attaches it to the work. The cryptographic nature of the blockchain means that this record cannot be altered or erased. Later, another artist builds on the work — riffing, remixing, or extending it. The registry records that contribution too, preserving the chain of attribution across time and across creators. Any monetary contributions flow to a treasury, and the community decides how to allocate them. In addition, the blockchain provides a bulwark against the proliferation of deepfakes, unauthorised AI versions, and the erosion of provenance that characterises the digital condition. The work lives and its life is legible.

The proposal describes BLACKBOX as a collective venture and a decentralised application that operationalises attribution without ownership. But what is a decentralised application? For a technologist, it is a piece of software that runs on a blockchain. For a legal scholar, it is a governance mechanism. For an artist, it might be a site of practice. The application, like the work itself, exists differently in each network that encounters it.

This is, in a sense, a recursion of the original problem. The proposal began by asking what a work is. The reply to this is that a work is not a substance but a relational pattern, something that exists in the network of its uses, users, and originators. Now we are asking the same question about the application itself. What is BLACKBOX? It is not a thing; it is a site of translation between different worlds: legal, technical, artistic, philosophical. The misunderstanding about the "rule book" is a symptom of this translational problem.

The question this raises is epistemological as well as ontological. How do we know what BLACKBOX is? Not in the abstract, but in practice: through the uses to which it is put, the communities that adopt it, the conflicts it generates. The pilot was designed to test the framework's viability; but it might also be seen as a test of whether BLACKBOX can be understood in practice, across different contexts and communities.

This suggests that the design challenge is not just technical or legal. It is also linguistic and communicative. The words, the terminologies chosen, are therefore important. They can also fail. They might fail because they carry baggage; because they are interpreted differently by different readers; or because the concepts they describe are new and unfamiliar. Perhaps, then, the task is not to find the perfect words but to build a practice that makes the words meaningful — a principle that has a resonance in the Madhyamaka tradition, where concepts are understood not as final truths (or the thing itself) but as means, and that their meaning emerges through use.

And yet there is an irony here. The proposal presupposes an ontology of cultural works that may not be shared by its readers. Elsewhere, I have argued that engaging with the ontology of cultural works — rather than accepting the functional conceits of law — is a justifiable and necessary endeavour because it changes our relationship to, production of, understanding of, and ability to use, cognitive outputs placed intentionally into the communal space. The legal conceit is not neutral; it actively shapes how people think about and value works. Denying the ontology of cultural works impoverishes our collective understanding and ability to use works for the greater good.

But this is precisely what is happening in the response to the project itself. The reader's instinct to read the proposed BLACKBOX dApp as a "rule book" is evidence of the very problem the ontological enquiry diagnoses. They are applying the functional conceits they know — such as governance as authority, protocol as imposition, registry as bureaucracy — to something that is trying to escape those conceits. The "rule book" reaction is, therefore, the proposal's first successful diagnostic.

18 August 2026 · On moral rights

I've used the term 'moral rights' loosely on this site, saying that BLACKBOX offers a moral rights inspired approach. However I think this phrase fundamentally fails to communicate what is at stake. I need to be clearer about what I mean. So, to explain I need to cover a bit of background context.

Moral rights (droit moral) entered positive law through the Berne Convention for the Protection of Literary and Artistic Works (1886). They protect the author's paternity (the right to be identified as the author) and integrity (the right to object to derogatory treatment). In civil law jurisdictions, these rights are often inalienable and perpetual. In common law systems, they are more limited and waivable.

The concept of droit moral can be traced back to the eighteenth-century struggle over copyright in Britain. The Statute of Anne, enacted in 1710, was the first copyright legislation to establish a limited term of protection. It granted authors a fourteen-year term, renewable once, after which works entered the public domain. This was a deliberate intervention by Parliament: it created a temporary monopoly to incentivise production, but it also guaranteed that works would eventually return to the commons.

The London booksellers, who had effectively controlled the book trade through perpetual copyright claims under common law, saw this as a threat to their business model. They had long operated as a cartel, acquiring authors' manuscripts through outright purchase and then holding the rights in perpetuity. To maintain this control, they mounted a creative and aggressive common law argument. They contended that copyright was a natural property right, existing independently of statute, and that an author's perpetual right to their work passed to the bookseller at acquisition. This was a strategic claim: the Statute of Anne, they argued, had only supplemented this common law right, not extinguished it. On this basis, they continued to assert perpetual control over works regardless of the statutory term, effectively circumventing the limitation Parliament had imposed.

The battle was fought in the courts over several decades, culminating in the landmark case of Donaldson v Becket in 1774. This decisive showdown was triggered when Thomas Becket, a prominent London bookseller, sought an injunction to protect his lucrative share in James Thomson's poem The Seasons from cheap Scottish reprints. Defending their market monopoly, Becket and his London cartel argued for a natural property right that existed independently of statute and could never expire. The Lords ultimately rejected the London booksellers' claim, holding that copyright was not a perpetual common law right but a statutory entitlement of limited duration, after which works fall into the public domain.

This history embodies a persistent idea, one that would later inform the concept of 'moral rights': an author's relationship to their work outlasts both its market value and the statutory term. But, in the contemporary legal context, the term 'moral rights' is doctrinally specific. In its positive-law sense, it is a sub-category of copyright, a right granted to a rights-holder within the proprietary order.

Whilst evoking the 'War of the Booksellers' captures a tradition of resistance against publishing monopolies, the term 'moral rights' introduces ambiguity because it collapses three distinct registers: the historical conviction that an author's connection to their work cannot be entirely severed; the statutory doctrine of paternity and integrity; and the broader philosophical claim about creativity and personhood.

So I am moving away from using the term even when qualified as 'moral rights inspired'. As I see it, what needs to be implemented is a framework for relational stewardship, or 'user obligation', where 'use' of a work generates an obligation to the originator and to the wider cultural commons. This operates, then, as a liability rule rather than a property rule. Fundamentally, however, this liability is more than a legal or monetary one because it is measured, ultimately, in socio-cultural terms.

7 May 2026 · Recent infrastructure updates

The site got a few upgrades this week.

First, I finally eliminated the Google Fonts dependency. The 3 typefaces I use (Inter, Libre Baskerville and IBM Plex Mono) are now self-hosted WOFF2 files served directly from this IPFS directory. So, no more requests to fonts.googleapis.com or fonts.gstatic.com. Google Fonts isn't exactly a privacy disaster but it's nonetheless satisfying to fully deGoogle the site ;-).

Second, I built a heartbeat agent — a tiny non-human steward that resolves the site's IPNS address every morning and signs a cryptographic attestation that says "I saw this site, at this CID, at this time." It then publishes the attestation to IPFS. The public log is at /heartbeat.html. The agent is a Python script that runs on a daily GitHub Actions workflow. It has its own Ethereum key.

For visual interest I also added a relational link graph — an interactive map of the conceptual and infrastructural connections behind this website. The graph is powered by D3.js and a plain-text links.jsonl file, both self-hosted.

1 May 2026 · Call for papers

Call for papers now open. I have launched a call for papers with guest editor Gavin Keeney for a special section of Amicus Curiae titled Law and Cultural Production. The call is open to scholars across disciplines — law, economics, anthropology, critical theory, art history, and beyond — provided the work engages substantively with law. Full details, including the timeline and submission instructions, are on the dedicated call page.

As a secondary experiment, accepted essays may also be archived on a planned BLACKBOX platform, testing smart contract registration and blockchain-based relational links. Participation is optional and does not affect the journal's CC BY licence.

3 April 2026 · First IPFS deployment test

Uploaded a static version of the page to Pinata. Works, but need to think about updating the log without breaking links. IPNS might be the solution.

2 April 2026 · Research writing in Obsidian

I use Obsidian for almost all my research and fiction writing. Its bidirectional linking and local-first philosophy align with my own distrust of cloud silos. Obsidian is an opensource note management software, that uses markdown that allows for multiple project vaults, organised by topic. The next step is to automate the project log: write entries in Obsidian and push the markdown files to IPFS via GitHub (using Quartz to convert them to html). This will hopefully keep the writing experience tied to my habitual environment. Actually, the real process starts before I sit down at the laptop: I write the old-fashioned way, away from the distraction of the internet. For this I use a ReMarkable tablet (e-ink) which also functions as my e-book reader. All my handwritten notes, annotations and chapters can be converted to markdown text and directly imported into Obsidian for final processing.

2 April 2026 · AI-assisted development: reflections

I've been tinkering with AI since the first OpenAI API release (around 2021) when I signed up to the developer's programme. My early experiments used the Python based PyGPT desktop app to query an AI-indexed database of texts. I then played with simple agents: the most memorable experiment involving an agent linked to Madhyamaka Buddhist texts, and another to a critical legal studies memoir, who I set to converse. It was entertaining but not ultimately very fruitful!

More recently, I've found AI very useful for web development: speeding up CSS (which I'm fully literate in, but not as fast as I might be) and navigating JavaScript (which I can read but write properly). The BLACKBOX cube and scroll-reveal js were written with heavy AI assistance. That said, I remain wary of relying on cloud-based AI. So, I've installed a distilled LLM on my own server, which I now use to search, summarise, and organise my document library and Obsidian notes. It's a lot more limited than ChatGPT (or DeepSeek etc) in that it doesn't have much in the way of generative 'chat' power, but it is local. And tbh I'm getting a bit fed up of the now ubiquitous faux-human machine chat experience (prompting me to finally ditch google as a default search engine. And re-read Ian McEwan's most excellent "Machines Like Me". Phew).

Ethically, I think using AI to assist with code for a site like this is fine — the output is open source, the prompts were my own, and I've understood (and often modified) every line. The danger is when AI replaces thinking rather than enabling it. I have seen it likened to a robotic exoskeleton for the mind, in which case overreliance leads to atrophy. I think that captures the issue very well. But, does that danger apply at the macro level and herald the mass atrophy of culture? Hmmm. For now, I treat it as a very patient pair-programmer who doesn't mind my stupid questions about flexbox.

1 April 2026 · Canvas cube: rotation + morphing + node regeneration

The BLACKBOX visual is now a rotating, subtly morphing wireframe cube. Clicking on it regenerates the internal node connections (diagonal lines). The holographic fill pulses inside. It is, I think, a neat metaphor for the work's life — never fixed, always re-linking.

31 March 2026 · Decision: no code repositories

To keep the site lightweight and auditable, I'm writing vanilla HTML/CSS/JS. The cube and scroll observer are the only scripts. This aligns with the project ethos. But what about serving my favoured typefaces via Google Fonts CDN? I can't quite bring myself to ditch the typography yet, so may have to look at serving the fonts locally from IPFS.

31 March 2026 · Project inception

Started sketching the BLACKBOX page as a complement to the book proposal. The goal: make the DAO's concepts tangible via interactive web design.