Interests by period, 1989–2026

The themes behind the archive, one period at a time, from dated evidence rather than memory

The chronology lists what was posted and when. This page asks a different question of the same years: what was I actually interested in, and how do we know? The answer comes from things that carry their own dates. Mail folders and the lists and daily digests that filled them, 1993–2004. Browser bookmarks, 1995–1997 and 2001–2026. Posts to public lists and forums. The texts left on an old laptop. Papers, talks, blog posts and repositories. Each period below names its themes and then lists the evidence, linked to the pages of this site that hold it. People are left out on purpose: the subject here is the interests, not the correspondents.

Browse the archive by subject

At a glance

The periods are those of the chronology. Five strands run through all of them; the table gives the form each strand took, and the sections below give the evidence.

Main themes by period and strand
Period	Physics and mathematics	Computing and networks	Data and society	History, culture, fiction	Making and machines
1989–1993	quantisation, geometry, the thesis	Macs, Unix, the first Linux, the department server	—	—	hardware for the lab
1994–1998	noncommutative geometry, quantum symmetries; five arXiv dailies	the early web, backlinks, HTTP; a dial-up IP network; Mozilla	quantitative finance, world systems, the physics of ownership	anarchist archives, folklore and myth, Alexandria, SF, cinema	PC internals, VGA to TV
1999–2003	renormalisation seen geometrically, spectral triples, coincidences in constants	system testing; teaching programming; pine and procmail	finance and physics; the euro	the Grail, Democritus, alchemy; essays, stories, letters	the webcam mouse
2004–2008	lepton masses and the Koide formula; the supersymmetric bootstrap	e-paper readers, GPL firmware; a first blog; volunteer computing	Zivis and Ibercivis	the hovercraft blog: bicycles, politics, Teruel	ciclorail
2009–2013	the bootstrap on forums and viXra; lattice, Kaluza–Klein	Linux, Android, the data stack	15M on Twitter, social-network papers, search maps	local archives, participation tools	—
2014–2018	families from SO(32); nuclear data; strings and compactification	Spark and Scala, scraping	Kampal: elections, scientometrics, deep learning	Spanish politics, manga, games	Lego mechanisms, Arduino, teaching children to compute
2019–2023	octonions and E6; NCG and supersymmetry	back-end engineering, Python tooling	language models; a pandemic modelled at home; collective intelligence	the fediverse; chess by vote	sensors, drones, home networks
2024–2026	papers again: SO(32) scalars, Koide sum rules, quantum gravity	agents, distillation, this archive	collective intelligence in schools; HPC	genealogy, languages, the archive itself	—

1989–1993: a physics student with the department's machines

The earliest traces are a physics student's, then a doctoral student's who was also running the department's computers. Bitnet digests and Usenet carry questions about a Mac II, a scanner and Mathematica; the first programming job, in 1989, was a lexicographic search engine for the university; a bug report reached the Linux kernel list in 1993, two years after the first release. The physics is already the physics of the whole archive in miniature: quantisation and geometry, with the first four papers in 1992 and 1993 and a winter school in 1992. In December 1993 the department's web server went up, with the choice recorded as “Mac or PC? BSD or Linux?”, and the first test message of the surviving mail was sent on the 13th. A question of that year about toposes and geometry is the oldest thread on the early-web page.

Evidence: Usenet and digest posts, the 1993 mail, the CV, the 2008 CV extracts, the first papers.

1994–1998: the web as a subject, and everything else at once

These are the years with the most evidence and the widest spread. Four strands run in parallel.

Physics and mathematics. The thesis, on scattering potentials, was defended in 1995 after a stay at a theory centre in Marseille in 1994; a summer school on quantum symmetries in 1995 left a list alias on the department server that was still receiving announcements in 1998; a school on noncommutative geometry followed in 1997. The first mail subscription that survives is a publisher's contents alert for mathematical physics, 1994–1995. Then, through the whole of 1998, five arXiv daily listings arrived every weekday: math, hep-th, physics, quant-ph and cond-mat, about nine hundred messages in ten months, kept in a folder called dftuz y física. Physics reading continued at the same pace while the day job was elsewhere.

The web, then networks. The department server carried the wwwlab experiments of 1994–1997: map tiles of Spain and Zaragoza, image maps, a referrer log and the backlinks mechanism that became a paper submitted to WWW6 in December 1996, and a message to the HTTP working group in January 1997. The 1995–1997 bookmarks read like a syllabus of the young web: hosting prices, the W3C design notes, Dublin Core, distributed indexing, the IETF and the Internet Society. From November 1997 to November 1998 the job was systems engineering on a telephone company's new dial-up IP network: keeping an integration testbed alive, naming and routing its machines, Solaris, DNS, CORBA, tunnelling protocols, interconnection with other providers, server certificates. The same year the browser's source was opened and its mail-news group produced nearly seven hundred messages in the mailbox; a campaign for refunds on bundled Windows licences was run under a pseudonym; a folder of security advisories was kept from December 1997.

Economy and society. The Flexion archive of “archivos ideológicos” ran on the server from 1995 to 1998, with the anarchist reading list of the politica bookmark folder behind it and the 1995 Zapatista consultation ballot beside it. A research list on anarchism was read in 1997–1998, and a folder on self-organisation in the summer of 1998 fed the electronic edition of La conquista del pan. An education forum from Zaragoza took twenty posts in 1998. Six posts went to a world-systems list in July 1998 on unequal exchange. Finance appears in the autumn of 1997 as a job search among quantitative-finance recruiters and stays as a mail folder, finanza, from September 1997 to the end of 1998, together with a working-paper service; the Physics of Ownership page on the homepage grows from it.

History, myth and fiction. In May and June 1997 the bookmarks turn to Alexandria, Euclid, Ptolemy and the desert fathers, and the essay on Alexandria and the routes of lost texts is written. The folklore list of 1998, on myths and oral tradition, took thirty-seven posts, the most of any list. The laptop of these years holds twenty-one essays on history and folklore and nineteen pieces of fiction and poetry. A July 1997 folder on photosynthesis and the origin of life became the question on polarised light that turned into the ice-and-chirality conjecture. Science fiction and the Mars landing of July 1997, a science-fiction author's list, Portuguese cinema in April 1997, and a fortnight of PC internals in the winter of 1996–1997 complete the picture.

Evidence: the 1995–1997 bookmarks, the mail folders and subscriptions of 1994–1998, the recovered server pages, the mailing-list posts, the libretto texts, the CV.

1999–2003: Usenet at its best, and a living made elsewhere

The record thins because the mail lived on other people's servers: a free British account during a 1999 assignment as a system-test engineer on a military aircraft programme, then a university account in Madrid where integral calculus for economists and introductory computing were taught in 2000–2001, with a programming course's mailbox as the only part that came home. From 2001 the teaching was programming and data structures for telecommunication engineers in Zaragoza, and it went on until 2006.

The intellectual life of the period is on Usenet, in sci.math and sci.physics.research: renormalisation seen geometrically, noncommutative spaces, coincidences in the constants, and, on the history side, the Grail, Democritus and alchemy. The arXiv preprints run through every year from 1999 to 2003, with the spectral-triples note of 2002 and a first submission to Physical Review Letters in December 2003; three papers on the history and philosophy of science date from 1999–2003. Two lists show where physics met the other interests: one on finance and physics, 2001–2003, and alchemy lists in 2002–2003. The laptop texts of these years include the 2001 interviews on the beginnings of computing and the Internet in Aragón, a piece on the coming euro, letters to the press on private universities and a squatted institute, and the notes for a talk of 2003 that took the four freedoms of free software into other fields. On the hardware side, a webcam turned into a mouse in 2002. Physics Forums begins in March 2003, the same month as the last pine session.

Evidence: Usenet posts, arXiv receipts, the CV, the pine and Evolution mail, the libretto, the 2003 slides, the webcam page.

2004–2008: forums, a blog, and computers that belong to everyone

The forum years. A single Physics Forums thread on lepton masses ran from 2004 to 2007, Wikipedia got its article on the Koide formula in 2004, and the supersymmetric bootstrap took shape in arXiv notes, sixteen preprints between 2004 and 2007, with a guest post on a particle physicist's blog about decay widths in 2006 and a term at a mathematics institute in Cambridge on noncommutative geometry in the autumn of the same year. The homepage of 2005 still had its research, politics and bicycle sections; the blog El último hovercraft, October 2005 to July 2006, wrote from Teruel about bicycles, politics, rail trails and whatever passed. On community sites of the time, diaries were kept in Spanish and English.

Computing turned towards devices and volunteers. Electronic-paper readers were taken apart and their GPL firmware discussed on the readers' forum; a ciclorail was built; and volunteer computing arrived with the city's project in 2007, presented at a free-software congress with two thousand users behind it, and its national successor in 2008. The mail moved to a webmail account at the end of 2007, which is why the local record ends there. The first GitHub repository dates from December 2008.

Evidence: forum threads, arXiv, the CV, the recovered homepage and blog, the talks, the 2008 CV, the repositories.

2009–2013: questions, answers and the first data

Usenet was over and the question-and-answer sites took its place: MathOverflow in February 2010, Stack Overflow that summer, Physics Stack Exchange in its first weeks of 2011. The bootstrap and the Koide relations moved to forum threads and viXra notes, and the blog returned in August 2012 as a second series, opening with local quantum field theory, lattice methods and Kaluza–Klein, then the bootstrap with Pati–Salam and, on the other side of the same blog, inequality, political gatherings and the statistics of bitcoin. A folder of nine issues of an old physics newsletter was bookmarked in 2013.

The data work starts here. Volunteer computing was presented at grid meetings in 2009, and a question-and-answer service among specialists led to camera widgets and event graphs; the 15M movement of May 2011 was followed on Twitter, a stream was captured through the spring of 2012, and the papers on online social networks and collective action appeared from 2011 to 2014. The recovered experiments of 2008–2012 belong to this strand: tweets as stickers, a map of seasonal searches, an interface for reworking ideas. The bookmarks of 2011–2014 are Linux administration, Android phones, the first big-data tooling, local archives and the tools of participation, with a script for searching video transcripts in 2013.

Evidence: the Stack Exchange inventory, viXra, the blog, the papers, the recovered experiments, the talks of 2009, the bookmarks of 2011–2014.

2014–2018: Kampal, and learning to build things

A data-science company was founded in March 2014 and the blog became half a notebook of the trade: a scraping course in September 2014, Spark versions and Scala bugs in 2015, electoral statistics for the new Spanish parties and the analysis of primaries and assemblies, retweet rankings of televised debates, the political activity of networks in a 2016 talk. Behind it the bookmarks show the reading: fifty-one on scientometrics in 2015–2017, which became papers from 2015; the data stack; deep learning in its second wave; and, from 2018, word embeddings and the size of the web and its search engines. A consultancy in Dubai in 2015 and a European research-information service in 2014 belong to the same strand.

The physics kept its own half: the bootstrap and the exceptional groups, strings and compactification, a season of nuclear-fission data in 2015–2017 that produced a conjecture about nuclear patterns, families from SO(32) and USp(32) in 2017–2018, the Koide formula in new formats, and a manuscript on generations in supersymmetric models with preons in 2014. A history-of-science account was opened in 2015.

The third strand is new: making. Seventy-one bookmarks on Lego mechanisms in 2015–2017, Arduino and sensors, science toys, a toy drone driven from Python in 2018, and teaching children to compute, with the questions of 2018 on sets, addition and the New Math, the first calculators for children and, in the leisure column, video games and manga.

Evidence: the blog, the talks of 2014 and 2016, the papers, the bookmarks of 2015–2017, the questions of 2018, the repositories, the CV.

2019–2023: octonions, a pandemic, and the first conversations with machines

The blog's busiest year since 2012 was 2019, split between octonions and E6 on one side and back-end engineering on the other, with alternatives to PageRank in between. In 2020 the pandemic was modelled at home, from an exponential hot take to a non-compartmental model and its extensions, with two notebooks comparing the waves at the end of the year; the open-data bookmarks of the same months are the sources. The company work turned to collective intelligence, with a paper in 2022 and two in 2023, and to a project on adolescents and social networks from 2021. A Python module for a physics master was taught in 2021–2022.

The bookmarks of 2018–2020 hold the largest making wave, sixty-four on Arduino and the Internet of things, twenty on drones, and a run on home networking and hardware; those of 2021–2023 are web development, archives and administration. The reading on language models begins with embeddings in 2018 and does not stop; a question on adaptive attention was asked in 2022, and in the spring of 2023 the first conversations with a chat assistant, about numerical integration and attention, were posted. In November 2022 the social presence moved to a mathematics instance of the fediverse, with a self-description as “the typical theoretical physicist becoming programmer and data scientist for survival”. Chess by vote as collective intelligence, the distance to the moon, regular polyhedra, and whether to cancel a mortgage when rates rise are the blog's smaller subjects of these years.

Evidence: the blog, the papers, the CV projects and teaching, the bookmarks of 2018–2023, the Stack Exchange questions, the repositories.

2024–2026: papers again, and research with agents

After twenty-five years the notes became papers: an interpretation of scalars in SO(32) in 2024, new sum rules of the Koide type in 2026, the collective-intelligence work in a physics journal, a history-of-science paper in 2024 and a scientometrics one in 2025. The bookmarks of 2024–2026 say where the physics reading went: sixteen on quantum gravity and noncommutative geometry, more than in any earlier period, and a question in 2024 on the Planck area. The blog follows: recursive anomalies in SO(32), string and trajectories, the Compton radius in Kerr black holes, a mass gap from Kaluza–Klein.

The other reading of these years is agents: twenty-five bookmarks on language-model agents, and from 2025 the repositories document research done with them, from knowledge augmentation and layerwise distillation to a living book on agents, a prompt-only reimplementation, numbered research workspaces and a file-based fixed-point workflow. Two smaller threads close old loops: a canvas of categories for children continues the teaching of abstraction, and a conversation with a model about artificial intelligence and anarcho-communism returns, thirty years later, to the Flexion archive. The projects of the period apply collective intelligence in schools and high-performance computing in industry. This site, and the thirty-year bookmark export it is built on, are the archive taking stock of itself in 2026.

Evidence: the papers, the blog, the repositories, the bookmarks of 2024–2026, the CV projects, this site.

Chronology, 2024–2026
Publications and the Planck-area question
Minimal length and quantum gravity, Koide, language models and agents
menard, selfupdate, agentbook, physres1, fixedpoint
The politica folder of 1995–1997, for the loop that closed
Bookmarks by theme and this archive

How this was put together, and what it misses

Every theme above is tied to something dated: a mail folder with a first and last month, a list subscription, a bookmark with a first-seen date, a text with a date inferred from its content, a post, a paper, a commit. Where the sources disagree the chronology and the CV keep the discrepancy visible rather than choosing. Mail evidence stops in 2004, when the mail moved to the browser; bookmarks are missing between November 1997 and 2001, the years of a laptop profile that did not survive; and the private mail of every period, its correspondents and its subjects, stays private, which is why this page names themes and never people. The same evidence, sorted by subject instead of by year, is on the bookmarks page and the subject pages.

Assembled on 22 September 2026 from the recovered mail stores, the bookmark export of March 2026 and the pages of this site.