It's no secret that I use Generative AI, specifically Claude Sonnet, to assist with the Ooloi project. I use it for writing Clojure tests TDD fashion, for generating Clojure code, for generating documentation, READMEs, architectural design documents and much more.

Above all, I use Claude for exploring architectural strategies before coding even begins. It's somewhat reminiscent of pair programming in that sense: I'd never just task GenAI with generating anything I wouldn't scrutinise very carefully. This approach works very well and allows me to quickly pick up on good design patterns and best practices for Clojure.

It then continued, quoting the code base:

I guess this answers my question about Clojure vs Python. It's not just the functional and immutable aspects of Clojure, it's also specifying the domain stringently through design, architectural discussons, specs, and tests.
​

I continued: "I also use Claude to create ADRs, READMEs, introductions for newcomers, Dev plans, etc. I'm quite open about this; transparency is paramount in an open-source project, and the idea is to facilitate collaborators' understanding of the Ooloi project as much as possible. This means that adherence to best practices, the use of sane architecture, abstracting away the complex stuff without compromising on performance or flexibility, etc, all are central."

Claude responded:

We then went on discussing open-source strategies, tutorials and onboarding materials for contributors and much more which I'll spare you from for now.

​Finally, I asked it to summarise its views on Ooloi as a project:

Now, this level of enthusiasm and praise from an anthropomorphic-presenting LLM (essentially a glorified Markov chain) always makes me deeply suspicious – I'm absolutely allergic to toxic positivity, especially the kind from the USA – so I asked it "Are you sure you're not overly enthusiastic just to please me?". It replied:

I then asked Claude 3.7 to comment on my journey from Common Lisp to Clojure and on the influence from Common Lisp and CLOS on Ooloi:

And finally, I just couldn't resist asking: "What would Rich Hickey say?"

I guess that hypothetical assessment is good enough for me at this point. Still not quite sure that Claude isn't giving me a friendly reach-around, though... ;)

It's been five months since my last update on Ooloi – or FrankenScore, as it's still known in its pre-release incarnation. This silence wasn't planned; rather, it happened because life got in the way. A demanding day job, a significant career change – we had to liquidate Delegat AB and I had to find a new job as a principal-level AWS Cloud Architect – and other responsibilities all conspired to slow Ooloi's momentum.

​I won't bore you with excuses – sometimes one simply must pause to change course, and I really needed to devote all time and mind space to finding what I hope is my final employment. Now that I've secured a great position with HiQ in Stockholm, I can return to Ooloi with full force.

Five months of relative silence offers time to think. Perhaps there's value in stepping back from the constant pressure to show visible output. In such moments, the architecture is refined not through frantic coding but through careful consideration.
​
The journey from Igor Engraver to Ooloi spans decades, and a few months of slower progress hardly register on such a timescale. What matters is that the vision remains clear and the foundation solid.

After all, the whole purpose of the Ooloi project is not to "disrupt the market". Like Octavia Butler's ooloi aliens, we're neither aggressive nor competitive. What is important, however, is doing this right using modern tools. The idea is to create an architecture and a platform that'll last and that musicians and publishers will want to use.

It's also to provide a powerful environment that can be easily extended through any JVM language. Ooloi has a tight, lean and efficient core, organically and seamlessly  augmented by a flora of plugins for any vertical. This would include jazz, early music, tablature, etc - but also commercial plugins to support things like virtual instruments, extremely intelligent playback, or perhaps GenAI used for musical purposes. The idea is to shift the initiative to the users, not to a central committee trying to anticipate user needs.

Ooloi is designed for flexibility and efficiency. Uniting these two aspects sucessfully requires careful architectural design. (And a language like Clojure for the core and the JVM for the plugins.)

With the core architecture stabilising, I'm thinking more about community. Ooloi is intended as an open-source project, a collaborative effort that will benefit from diverse perspectives and expertise.

The extensive documentation work completed earlier – including the architecture decision records, READMEs, and technical specifications – was not merely for my benefit. It prepares the ground for future collaborators, creating a clear map of the territory for those who will join us.
​
The website, this blog, and the growing collection of documentation all serve as beacons for those who might be interested in contributing. They signal our commitment to transparency and proper communication – essential ingredients for any successful open-source project.

Five months of comparative quiet doesn't mean I've abandoned ship; it simply reflects the natural ebb and flow of a project undertaken alongside life's other commitments. Ooloi continues to grow, perhaps not as swiftly as in those heady initial weeks, but with steady purpose nonetheless.

I'm reminded of how musical compositions themselves develop – sometimes in great creative bursts, other times through careful refinement of existing material. Both approaches have their place.

To those following Ooloi's progress, thank you for your patience. The work continues, and updates will come more regularly as we approach the milestone of public release. The vision of a modern, efficient, and elegant music notation system – one built on sound architectural principles and open to community collaboration – remains as compelling as ever.
​
Until next time (which will be considerably less than five months hence),
/ Peter

An Architecture Decision Record (ADR) is a point-in-time document that records architectural decisions and the reasoning behind them. And since communicating architectural ideas clearly to collaborators is extra important in open-source, I've just added a bunch of ADRs to the Documentation page.​

​They cover various aspects of the FrankenScore architecture, from high-level design decisions to specific technical choices. They are:
​

• ADR-0000: Choice of Clojure as the Primary Programming Language
• ADR-0001: Separation of Frontend and Backend into Distinct Clojure Applications
• ADR-0002: Adoption of gRPC for Communication Between Frontend and Backend
• ADR-0003: Integration of Plugins as a Core Architectural Component
• ADR-0004: Implementation of Software Transactional Memory (STM) for Concurrency
• ADR-0005: Selection of JavaFX and Skija for the Frontend GUI
• ADR-0006: Adoption of SMuFL (Standard Music Font Layout) for Music Notation
• ADR-0007: Implementation of Nippy for File Persistence
• ADR-0008: Implementation of Vector Path Descriptors (VPDs)
• ADR-0009: Implementation of Real-Time Collaborative Features

In the past weeks, I've been focused on FrankenScore's core architecture. I'm not rushing to open-source this; instead, I'm taking my time to craft a solid platform that will do the heavy lifting for future users and collaborators. All the complexities involving data representation and manipulation in a multi-threaded environment must be solved so collaborators can concentrate on the essentials. Clojure is ideal here, just as Common Lisp was the clear choice for Igor Engraver back in 1996.

Key developments:

1. The API is now fully polymorphic and can be used in the same way internally in the backend as in the frontend. There is a system of pointerless vector path descriptors (VPDs) implemented for this purpose that all API operations can accept as part of their polymorphic setup. I wouldn't be surprised if core collaborators will use the API for internal purposes as well, as it is highly efficient and exposes the underlying functionality in an abstract, domain-specific way. There should be little need to go directly to the underlying data structures, at least not for speed - and certainly not for expressivity. This also bodes well for plugin development in other languages than Clojure, which is an important feature.

2. This beast is fast. Clojure's STM facilities ensure high-speed ACID-compliant transactions with automatic retries. They are also composable. This means that plugins can bombard the backend with hundreds of thousands of mutation requests, for instance to implement MusicXML, with the same efficiency as the pure Clojure backend.

3. Piece Manager Implementation: There's now a Piece Manager, providing functions for storing, retrieving, and resolving pieces from IDs. This allows for multiple clients to work simultaneously on the same piece in a distributed arrangement. The FrankenScore backend can run in the cloud with multiple people collaborating on the same piece. Multiple pieces can be open simultaneously to allow copy-and-paste operations between them.

My next steps involve implementing file persistence (saving and opening music files), as well as tackling printing. These are foundational features, not mere add-ons. Persistence forces a clear definition of the data model and enables easier testing. Printing isn't just about output; it's about representation and serves as a sanity check on the entire system design. Both will likely inform further refinements of the core architecture, potentially revealing oversights or opportunities for optimisation.

Additionally, sequencing is a crucial part of the core platform. And by sequencing I mean support for converting musical representations to timed sound events - though not necessarily via MIDI; a software synth may use direct means of control, for instance. The core sequencer can be used by plugins to generate MIDI, or to input MIDI, but the actual MIDI implementation will be done in the plugin layer. But that's a whole blog post of its own.

FrankenScore is still private. There is a number of things that need to be in place before the project can go public and I can start inviting collaborators, so let's touch a little on names, releases and versions.

At this stage I'm finalising the robust, high-performance platform for ACID-compliant transactions which forms the basis of everything in FrankenScore and is manifested through the backend API.

FrankenScore becomes Ooloi when released as open source. But the first release doesn't need to be Ooloi 1.0, which by definition would be feature-complete. In fact, it should be Ooloi 0.n with an n as low as possible, meaning it's best to go public as early as possible, yet feature-complete enough so Ooloi's promise is immediately apparent.

Here's a very rough project plan:

1. Finalise the API (nearly done)
2. Implement File Persistence using Fressian (saving and loading pieces)
3. Implement gRPC Layer and Event Handling (for communication between the backend and the frontend)
4. Create Initial Frontend with Hello World Window (proof of concept)
5. Implement printing in the frontend client (of the Hello World Window)
6. Implement enough of the frontend/backend interaction so that a score can be set up and simple elements can be entered and manipulated. This is an complex. multi-step process that also involves setting up event handling in the frontend. It's the basis for everything that is to come and must be rock solid.
7. Implement measure reformatting locally for measures according to a simple linear method that will be upgraded later
8. Implement backend reformatting and reflow of measures over systems and pages
9. Iteratively add more notational elements and features until Ooloi 0.n can be released, meaning that the project is released as open source.
10. After the release, adding more notational elements continues in parallel with the remaining points:
11. Implementation of a more advanced measure formatting algorithm
12. Implementation of the plugin system.
13. Implementation of an open source plugin for simple MIDI in- and output.
14. Implementation of an open source plugin for reading and writing MusicXML files.
15. When all notational elements are supported, Ooloi 1.0 can be released.

​So, point 9 represents the point where the project goes public and Ooloi 0.n appears. It remains to be seen how feature-complete the notation must be to confidently take that step.

However. There might of course be room for collaborators in the project before the public release as open source, as there are points in the above list that cover isolated features that could be delegated to an experienced Clojure programmer. Hmm. Let's think about that.

I should perhaps say something about how Generative AI is used in the FrankenScore project. First of all, I have a prompt of about 4100 lines which prefaces every conversation with the AI chat client. The prompt consists of project documentation, background, design principles and goals, coding principles and conventions, explanations of central code and code examples. It also includes a major part of the source.

This allows the AI to:

1. Analyse and describe the project,
2. Converse about architectural choices,
3. Assist with project planning and open-source strategies,​
4. Judge how well the project embodies best Clojure practices,
5. Generate, with guidance, Clojure code and tests, and
6. Produce meaningful documentation and READMEs.

The copy on this website was almost entirely created by AI means, often using multiple iterations until I arrived at something suitable for publication. There remain a few passages that slipped me by as the AI produced text that reads a little too self-congratulatory on my part, but it was simply the opinion of the AI (though it is of course nice that it likes the code). I'll fix that during the days to come. Also, the technical comparison with other software is a bit too speculative and monotone. I'll change that, too.

In terms of code, I've found that Claude 3.5 Sonnet reasons better at depth about Clojure code than GPT-4o and consequently is the superior choice for complex coding. GPT-4o is still useful for producing text, though. It isn't exactly bad at coding, but it has a tendency to vomit code at you at every opportunity, which is both tiresome and expensive. Also, it kind of loses track when conversations get very long.

And they do; the chains of thought are sometimes complex, and a meandering AI can get costly. Therefore using Claude saves money in the long run.

By the way, it's easy to tell when I am writing. Just look for signs of British English. You know, -ise and colour and whilst and so forth. The AI invariably produces American English.