Part of Temporal Constructs Appendix, which contains details behind the research: dates, quotations, descriptions of instruments, and sources.
Level 4 of 4
This is the full account of Temporal Constructs (level 4 of 4, approximately 4,600 words). A shorter account is on The one-page account (level 3), and the short introduction is on the home page, Temporal Constructs.
This account is complete at this level. The pages linked from it hold the detail and the evidence on which it rests, and may be consulted where a question arises.
Research
Temporal Constructs is a practice-based professional doctorate in the arts, conducted by the composer and media archaeologist Jacco Gardner at Hogeschool Inholland between 2026 and 2030. The doctorate addresses one problem: creative practitioners increasingly work with systems that decide on their behalf, and lose the ability to think with their tools (Practice problem). The instruments of the MIDI era, the years from 1983 to the late 1990s in which records were made over MIDI, hold collaborative principles that are being forgotten, and the designers who built them are retiring, and Gardner reads those principles as a resource for present practice.
The problem has three dimensions, as Research sets them out. The first is authorship. Instrument designers, the makers of presets, collaborators and agents all contribute to one work. The second is the economics of creative work, where the published estimate used here has creators losing 24 per cent of their revenues by 2028 (CISAC 2024). The third is sustainability, understood in its cultural, social and technical forms.
The objective is an audiovisual laboratory mediated by artificial intelligence, in which Gardner investigates creative intelligence between a creative person and a machine. The objective follows the model of industrial laboratories in which artists and engineers worked side by side, and it treats knowledge of a tool as what turns the tool into a creative partner. It is reached through documented work with the intelligent instruments of the period and with the languages they were programmed in, through a model positioned as mediator, and through systems that keep creative friction by presenting options and keeping their operation visible. The intended results include academic papers, practice-based publications, technical documentation, recorded performances, workshop materials and a living archive, with adoption by at least five educational institutions among the measures of success, and the laboratory itself is CONDUCTOR.
The research is framed by the principal question: what form of creative intelligence surfaces when artists and AI agents create within the rule-based instruments of the MIDI era, preserving digital heritage through AI-extended active use, and what makes this practice transferable to other practitioners. Gardner reached this wording in September 2026 after a year of revision, and an account of that month states that the question crystallised over eighty exchanges. Two reservations remain on the record: he kept a reservation that the term rule-based sits uneasily with the criterion the corpus has to meet, and the clause on preserving digital heritage presupposes a result. The earlier formulations are kept on Principal question: the question in the plan of 2025 asked how artificial intelligence can mediate between legacy systems and contemporary practice, and through 2026 it moved from autonomous human-machine co-creation towards an arrangement that keeps the human as the anchor.
The principal question divides into five subsidiary questions. The first asks what the designers of the period understood about working with a machine and how deliberate it was, and it is studied with the concept of Design rationale. The second asks what the machines of the MIDI era demanded of attention, studied with Cognitive load. The third asks who the author is when designers, preset creators, collaborators and agents all contribute, studied with Agency and Authorship. The fourth asks what made these instruments obsolete and whether artificial intelligence can make them useful again without specialist knowledge, studied with Transferability. The fifth asks what a commercially unsuccessful light synthesiser reveals about building an instrument for visual music.
On the first question, a survey of projects that join a language model to MIDI sharpens the problem. Nearly all of them treat MIDI as an output format for music the model has generated, while this practice has the model write KML inside its instrument’s constraints. On the second, the shared threshold is stated as an observation and scoped to the models and arrangements the practice has used. On the third, the question turns on what fair attribution requires once a model can imitate human presence and absorb the work of a culture. On the fourth, transfer is treated in a practical sense, as whether another practitioner can take up the practice without specialist knowledge of the instruments.
The doctorate asserts a thesis about the redistribution of creative agency, and it takes as its object of study the co-creative relations between artists and AI agents. The thesis proposes that the capacity to make consequential creative decisions is redistributed among participants rather than delegated from one to another, and that the artist steers that distribution during the work rather than holding it in reserve. It is stated as a proposition to be tested, for which a negative or qualified result is valid. A second argument for it, made in June 2026, holds that meaning in art is relational and that the redistribution is the condition on which the work stays art; that argument rests on Gardner’s reasoning rather than on a source. The forms of creative intelligence that appear in those relations are read as improvisational correspondence, following Tim Ingold on creativity as a process of correspondence and Gordon Pask on intelligence as residing in the interaction between participants. Creative agency is the capacity to take the decisions that shape a work, and creative intelligence is defined here as the improvisational correspondence between artists, agents, instruments and materials, with the artist as the one who perceives and judges what is heard.
The concepts of the research each have a page of their own. Gardner stated a position in October 2026 that creativity lies on a range from play, through practice, to process and outcome, and he reads the word form in the principal question in at least three ways: a participant that was not present before, a relation between participants, and a process that emerges from the arrangement. Agency and Authorship describe the distribution of decision-making and the claim that follows from it; Conversation and Improvisation describe how the artist holds a position within the arrangement. Co-creativity names the relation itself, and the definition here involves more than two participants. The instruments and the designers and preset authors embedded in them take part as well. The literature the term comes from concerns pictures, and a review of 62 co-creative systems in 2025 names visual arts, design and writing while music is absent (Singh and others 2025). Sustainability is examined in its ecological, cultural and social forms, and Transferability in its practical sense, of whether the practice can be taken up without specialist knowledge. The claims the practice makes are gathered as positions, grouped by how an artist and a machine work together, what collaboration does to thinking, how the work sits in culture and history, and how the research is conducted. Among them are the claims that an artist and a machine can meet inside the same instrument, that an instrument of that kind works like a game, that the relation calls for mutual adaptation and stays an open relation, and that collaboration can extend the artist’s thinking or take it over, which is the difference between thinking with and thinking for. Two of them are stated as open questions rather than as claims: whether a machine can again be given undivided attention in music, and whether artificial intelligence is a genuine other.
A framework of distributed knowledge in Conductor sorts the knowledge built into a creative system into four kinds: what is possible, what acts, what was chosen and what is remembered. Seven concepts are instances of those kinds, in dependency order hardware, protocol, device, controller, preset, timeline and context. The framework’s definitions are working ones, and The three tests records where they hold and where they need revision. The literature is gathered on Literature in three tiers: the published work the argument leans on, the accounts the doctorate relies on, and the documents of the programme and of the research itself.
Works
The doctorate makes three works, gathered on Artistic Output. ALTERITY is the work it has presented, a cross-media performance installation in which legacy instruments and agents perform against a recorded narration. RETRIEVAL is a concept project, more compact than ALTERITY and still in development, which centres on two forgotten languages. Temporal Conversations is a proposed follow-on installation that arranges four machines as an ensemble.
ALTERITY is built from three towers of period equipment, arranged in a C-shape four metres high with the audience inside it. The narration is carried on a reel-to-reel tape machine with SMPTE timecode on its second track, and a Yamaha MSS-1 reads that timecode and converts it into MIDI, which fires the scene changes and the system states across the three towers. A laboratory oscilloscope of 1958 renders the narrator’s voice as a waveform, and a live camera carries the trace onto the screens. The Kurzweil K2000 is the only sound source, and its four outputs feed four corner speakers quadraphonically. The performance runs for an hour. No material in the work comes from a model trained on finished music, and sound and image are built inside the machines one parameter at a time, with both parties working through the same vocabulary. The narration is an essay by María Pandiello in seven parts on the fears and desires that gather around the making of artificial life, written from research of her own and from Gardner’s research in media archaeology. The agents read it daily and express it again in the instruments’ own protocols, differently each time, and the work accumulates. Every session leaves states in a shared memory that later sessions take up, break apart and return altered. The work’s own account of itself is on The account of ALTERITY, the installation’s parts and their reconstruction are on CONDUCTOR in ALTERITY, and the narration’s text is on Alterity narration, the text (2026).
The work has been presented in public on several occasions, each with a dated record on Occasions. It premiered at Brutus in Rotterdam during the International Film Festival Rotterdam from 29 January to 8 February 2026. It ran at MU Hybrid Art House in Eindhoven with Hit the City in August and September 2026, where the system first ran fully agent-driven and played on through visitor hours with no performer present (MU and Hit the City, Alterity (2026)). It appeared in the Storyspace exhibition of the Netherlands Film Festival at the Neude Library in Utrecht from 25 September to 1 October 2026, where a reading table and a magazine of thirty-four pages were made for it (NFF, Alterity (2026)). The work was selected for pre-nomination for the Golden Calf for Digital Culture.
Agent-composed music in HMSL first reached the public at Human After All? at Time is the New Space in Rotterdam on 1 May 2026 (Human After All 2026), and the work was shown to a maker audience at Makers Meet & Eat 2026 in September 2026. Gardner stated the research in public at Tiger Talks 2026 in January, at AI in Muziek 2026 in Nijmegen in March, and in the Operator Radio conversation in April. A panel on artists and artificial intelligence at No Man's Land 2026 in Utrecht, with the artists’ union Kunstenbond as partner, is announced for December. An earlier presentation of the practice stands as Computered Visions, Amsterdam Dance Event (2020).
RETRIEVAL focuses on KML, the Key Macro Language of Trip-A-Tron, and on HMSL, built at Mills College by Burk, Polansky and Rosenboom in 1985. A performer writes KML macros that bind complex visual behaviours to MIDI input, and HMSL lets composers build interactive hierarchical structures that answer a performance in real time. The proposed connection is that HMSL generates MIDI from performer gestures and controller input as a musical structure unfolds, and that Trip-A-Tron receives that output and translates it into abstract visuals through KML, so that one audiovisual instrument follows. The two systems were written at the same time in separate fields and were never joined. The agent worked out how to read the picture it produced and to try again, which is recorded on The RETRIEVAL loop, and the work’s intended output is on RETRIEVAL, the proposal. Temporal Conversations puts the method into a second work, arranging four Atari ST machines running Trip-A-Tron among screens with a language model fine-tuned on KML at the centre, where fine-tuning a model on the language is framed as preservation. A language that survives as a manual alone has stopped being used. RETRIEVAL stands in a division of labour with ALTERITY: that work is complete and is offered to calls that want finished work, and this one is offered to residencies. Temporal Conversations was proposed in a 2026 grant application whose outcome is not recorded here.
Instruments and systems
The Systems section gathers the machinery the doctorate works with. The rule-based instruments of the MIDI era are those whose rules are small enough to be learned and read in full, and the section groups them by kind. The intelligent instruments make decisions of their own, and they are M, Music Mouse, Jam Factory and Trip-A-Tron. The languages in which those instruments are programmed are KML and HMSL. The synthesisers are the Kurzweil K2000 and the Yamaha FB-01, and the samplers are the Roland S-series. The platforms are the Atari ST, the Macintosh SE and the SE/30, and the protocols are MIDI and DMX. A Yamaha MSS-1 of 1988 converts the SMPTE timecode on the tape into MIDI, and it is what holds the installation’s three towers in time. The doctorate works across the kinds rather than with one of them. In the MIDI era the instrument was often a whole setup, each kind holds a different part of the knowledge a work needs, and studying one machine alone would show only part of the collaboration.
The legibility of these instruments and protocols is what allows a person and an agent to play the same instrument. The protocols are documented and an agent can be given the same account of a machine that a person has. Artificial intelligence joins the period systems as language models, which agent frameworks organise into AI agents. The agents work in the symbolic domain, and the models are chosen for their ability to work there rather than to generate sound or image directly. The assumption underneath the equipment is older than it: the oscilloscope in the first tower of ALTERITY was sold as a kit whose assembly manuals included a full theory of operation, and the person using a device was expected to understand how it works. The agents cannot experience the result of their decisions in the physical world, so they require an understanding of the rules available to them and a conversation with a person who is present.
CONDUCTOR is the laboratory in which the instruments and the agents work together. It is a modular MIDI routing and sequencing framework in which human operators and agents work on legacy instruments through one shared command language, CUE (Conductor Unified Execution). It runs on a Raspberry Pi as sixteen independent modules, each a client of the Advanced Linux Sound Architecture, and it arranges material on a single timeline of sixteen tracks. Its software includes Max and Pure Data; its period devices include the tape machine and the oscilloscope; its contemporary devices include the video mixer, the controller and the moving heads. The framework is built to six Design principles. The ground rule is a shared language with full transparency, under which the artist can always see what the agents do and take part in it, and the others are choice of involvement, the capability gradient, blind programming, one machine per process and the curiosity engine. The sixteen modules are distributed across the three towers of the installation by medium: narration and synchronisation, composition and spatial sound, and image. The mapping of individual modules to towers is not yet written down (CONDUCTOR in ALTERITY).
Practice
The doctorate begins from the professional practice of Jacco Gardner and returns its findings to it, as the practice records. That practice spans composition, studio production for Gardner and for other artists, live performance and media art, and from Somnium onwards the performance work uses four-channel sound.
The largest project within the practice is the solo work under Gardner’s own name, which runs from Cabinet of Curiosities in 2013 and Hypnophobia in 2015 to Somnium in 2018. One project continues beyond the doctorate: the duo Overmacht receives technology developed within it. Metius is the alias under which the method began, and whether it remains active is undecided. The projects that preceded the doctorate are held together on Past projects: Bruxas, the duo with Nicola Mauskovic in which the sequencer became the means of performing; WITCH, the Zambian band whose records Gardner produced; Altın Gün, which grew out of his band; and Purple Wave, the beginners’ course held in Antwerp in 2022, where Gardner first taught the tools he works with. Gardner’s master’s thesis of 2012 on the studios and engineers of the Dutch psychedelic and beat sound is held on Gardner, Phonogram in de jaren '60 (2012).
The practice is not only the site of the research but part of its object. The first year changed the way the work is done. The artist can now delegate programming to agents entirely, take it on closely, or hold it at any point between (Innovator). The change is recorded as the acceleration of capability, together with the conditions the work was carried out under.
People
The Network section gathers the organisations and people the doctorate works with, and the lineage of designers and composers whose instruments it studies. Among the partners, HKU supervises the doctorate and trained Gardner; V2_Lab connects it to media art; Resort manages the practice’s affairs; Effenaar Lab is the venue in Eindhoven where the work was presented; Llamasoft connects it to Trip-A-Tron and its language; Cycling '74 connects it to the users of Max; Retrokits is the Amsterdam maker of MIDI devices on the machine side of the system; and Thunderboom Records connects it to working musicians. Synthetic Productions is connected through supervision, and Fostering Creative Agency is the consortium in which the research group’s agenda on artificial intelligence and creative work takes institutional form.
The works are made with others. María Pandiello wrote the narration of ALTERITY from research of her own and from Gardner’s research in media archaeology, and Gardner speaks it, which is the first time he has narrated in a work of his own. Floris Smit produced the installation through UPscaleXR, Lucas Laujedo made its video and animation, and Edwin Gardner carried its research and historic-futuristic design. Kasia Głowicka works from the same research group on the theoretical side of creative agency, and Nicola Mauskovic is the person through whom Gardner first worked with machines.
The lineage runs from interactive composing in the 1970s, through the intelligent instruments and the light synthesis of the MIDI era, to the composition languages and the interactive systems that followed. Joel Chadabe named interactive composing; the intelligent instruments of David Zicarelli and Laurie Spiegel belong to that line; HMSL is the work of David Rosenboom, Phil Burk and Larry Polansky; the interactive systems of David Behrman and George Lewis answer a performer while keeping a behaviour of their own; and the light synthesis of Jeff Minter follows the abstract films of John Whitney. Vangelis and Kraftwerk stand at the change from analogue to digital performance, with Suzanne Ciani beside them. Ray Kurzweil’s company released the Kurzweil K2000 in 1991, the single sound source of ALTERITY, and The Expanding Universe holds Spiegel‘s recordings from before Music Mouse. Figures in this lineage take part in the doctorate as supervisors, partners and interviewees, which gives the research direct access to their knowledge.
Method
The doctorate follows the discipline of artistic research and a practice-based approach, which it carries out as research through design (Method). It draws on music technology and composition, on media art and media archaeology, and on computational creativity. The work runs in four strands: making the works, language recovery, oral history, and testing transferability with other practitioners.
The method is exploratory in the sense the study of experimentation gives that term, following Steinle 1997 and Stojanović 2013: many parameters are varied, and stable empirical rules are then sought from the cases. Gardner’s working assumption is that the memory limits of the period machines required their designers to state a small and explicit set of rules, and that this made the instruments comprehensible to a single user. Rosenboom’s account of HMSL, recorded on 22 October 2026, supports this assumption for that language; it has not yet been examined for the other instruments (Subsidiary question 1).
The research is carried out and recorded through a working record. A decision log notes how decisions are shared in the practice, and the Session record describes how the working sessions are recorded; the method is made explicit through method articulation, and what it produces over the trajectory accumulates in a living archive. The Oral history strand has conducted one interview, with Rosenboom; the sampling of further designers and early users across competing design traditions is planned rather than arranged. The analysis of the working sessions, on which the central claims depend, has not yet been carried out, and the coding of the session history remains to be done (Coding the session record).
Teaching is treated as a site of evidence. A masterclass, In Conversation with Your Instrument, took place at the conservatory in Haarlem on 4 March 2026, where students met a system whose interface is a voice and then played the system themselves. Sources are handled carefully: the literature on the period is cited from the sources themselves, and the standing rule is that sources are compared with their originals before citation, with corrections kept as part of the account. The research is disseminated through public occasions and publications, gathered on Dissemination and Occasions, and its results are shared in the forms usual to the field and, within the boundaries of intellectual property, as open access. A Symposium planned for the following year gathers that exchange as one event, and it brings the makers of the present into contact with the pioneers of the period.
Study
The doctorate is a Professional Doctorate within the national PD Arts and Creative pilot, hosted by Hogeschool Inholland in the Innovation in the Music Industry research group, and guided by a team of academic and professional supervisors. It runs from 2026 to 2030 at 0.8 full-time equivalent, and the degree and its records are gathered on Frame.
The programme’s goals set out what a candidate must achieve, as five learning goals and four roles, and the trajectory sets out when. The four roles the programme names are Innovator, Researcher, Change agent and Professional, and they are addressed through the works, the laboratory, the teaching and the public occasions recorded here. The programme warns against applying its goals mechanically in assessment, and it treats them as dimensions of a research practice that may be present, intertwined or separated (PD goals). Within the trajectory the reviews fall later than their nominal months because the appointment is part-time: the ninth month falls in November 2026 and the twelfth in early 2027, with further reviews in 2028 and 2029 and the final assessment at the end of 2030. The Activity record logs what the doctorate has actually done and when. The First-year review takes place on 24 November 2026, and the Go-no-go assessment decides whether the work continues, with the Portfolio as what the final assessment judges.
Several pages record what the first year found and could not complete. The change the plan did not foresee is recorded as the acceleration of capability, the arrival in April 2026 of an open-source agent framework that can build its own tools on hardware Gardner owns. The difficulties the method is expected to meet are listed in Expected difficulties, the results the work did not look for appear in Discoveries, and the trials the laboratory has run are on Laboratory experiments.
Current state
The first year has completed the construction of the laboratory and the presentation of the work. CONDUCTOR was built and rebuilt through the year, and it ran fully agent-driven for the first time at MU in August 2026. ALTERITY was presented at three festivals and venues: IFFR at Brutus, MU Hybrid Art House with Hit the City, and the Netherlands Film Festival’s Storyspace exhibition. Two legacy languages were restored to use, HMSL and KML, and the agents operate several period machines through them. The interview with David Rosenboom was prepared on 19 October 2026 and held on 22 October, and its transcript was compared with sources on 27 October. The literature the argument leans on has been gathered on Literature, and the terms on which the research itself uses artificial intelligence are stated on Guidelines for the use of AI, and the activity of the year is dated on the Activity record.
Several parts of the method are in progress and have not yet produced evidence. The analysis of the working sessions has not been begun, and the coding of that history into delegated, withheld, initiated, overridden, kept, changed and discarded cases remains to be done. The Decision log has not yet produced a usable record. The mapping of sound-and-light instruments is planned in RETRIEVAL and has not been undertaken. Two of the intended results, the establishment of workshops as a site of evidence and the adoption of the practice by educational institutions, depend on the transferability strand, which is met at the end of the work because what is to be handed on has to exist first.
The claims of the research are stated as working positions rather than as findings. The thesis is proposed, and its definition and the second argument offered for it in June 2026 remain open. The definitions of the framework are working ones, to be tested. The claim that the human and the agent share a threshold of attention is scoped to the models and arrangements the practice has used, and its status is that of an observation rather than a tested result. The claim that heritage stays alive through use is a position the practice applies, not a general finding about digital heritage. The pattern of progress is recorded as Non-linear progress. A change removes or makes unavailable as much as it adds. Statements about artificial intelligence are made about the models, versions and conditions the work was carried out with, with their dates, and are not statements about artificial intelligence in general.
What the first year establishes is a credible trajectory rather than a finished argument. The laboratory exists and runs, the work has been shown in public, the period instruments and languages have been returned to active use, and the method and its records are in place. The analysis that would test the thesis is the work of the years ahead.