What I chose to keep while reading. Browser bookmarks recovered from every
machine I could still open, from a Netscape file of January 1995 to the current profiles, merged, dated by their first appearance and sorted into the
subjects they belong to. A bookmark is a weaker trace than a paper or a post: it records
an intention to come back, and the folder it was filed in records what the intention was
about. Read together across years, they show which questions kept returning.
Browse the archive by subject
Where the bookmarks come from
The raw material is 41 bookmark stores: three Netscape bookmark files saved in the
document root of my university homepage between 1995 and 1997, Netscape and Mozilla-suite
profiles from a 2001 workstation, Firefox profiles from a 2007 Ubuntu laptop and a 2010
netbook, and Chrome and Firefox profiles from every machine since 2012, including the copies
kept inside older backups. Together they held more than 8,000 entries; after removing the
copies that the same bookmark left in successive profiles, 2,363 distinct addresses remain.
The date given to each one is the earliest timestamp any profile recorded for it, so it is a
latest-possible date of discovery. A batch of about 120 bookmarks first seen in October 2012
is a profile import, and many of those links are from 2009 to 2011. The three 1990s files
carry Netscape's own ADD_DATE stamps, from 12 January 1995 to November 1997.
Browser defaults and the Netscape 4 personal toolbar account for 153 entries and are set
aside. The remaining 2,210 were classified into some seventy themes by folder name, domain and
title, with a manual pass over everything. Themes about personal life, shopping, bookings,
household paperwork, contacts and internal dashboards at work stay in the private notes; the
page below covers the rest. About a quarter of all the addresses had stopped answering when
the 2001–2026 set was checked in March 2026, and for links first saved between 2012 and 2016
the share is closer to a third; of the 1995–1997 addresses almost none answer today. Links
here were re-checked when the page was written; where the original had gone, the nearest
surviving page or a Wayback Machine capture is given instead.
Classified bookmarks by area and period of first appearance
Area
1995–97
2001–10
2011–14
2015–17
2018–20
2021–23
2024–26
Total
Physics and mathematics
49
0
70
107
17
5
27
275
Data, networks and text
60
1
80
188
166
63
34
592
Systems and devices
47
5
74
78
54
9
3
270
Machines, sensors and learning to build
0
0
11
114
102
14
2
243
History, language and society
190
0
83
104
24
25
8
435
Reading, playing, travelling and daily life
65
1
46
136
100
39
8
395
All classified
411
7
364
727
463
155
82
2,210
The shape of the table is itself a record. The 1990s column is a single Netscape
profile, saved three times in 1995, 1996 and 1997; then nothing survives until 2001, because the
machines of the Telefónica, Carlos III and Teruel years were wiped. Bookmarking peaked in 2017, with 351 new
entries in a single year, and fell steadily afterwards as reading moved to feeds, search
and, since 2023, conversations with language models. The physics column revives in 2024
because a research question needed a reading list again.
1995–1997: the Netscape bookmarks of the university homepage
Three files named bookm1.html, bookm2.htm and
bookm3.html sat in the document root of dftuz.unizar.es/~rivero/, saved
from Netscape in 1995, 1996 and 1997 and recovered in September 2026 with the rest of that tree.
They are the only bookmarks that survive from before 2001, and they run alongside the
wwwlab experiments and the Flexion archive of the
same years.
The Web as a subject
1995 to 1997 · 60 bookmarks
The first folder, June 1995, is a survey of what it cost to put pages on the web: hosting
prices at XS4ALL, Cistron and NetLand in the Netherlands, Hiway and onLine in Britain, UniNet
in Italy, the WELL, and Netscape's guide to creating net sites. December 1996 is the reading
around the backlinking paper: Tim
Berners-Lee's design note on topology, the W3C note on describing and linking resources, the
Dublin Core, the Distributed Indexing workshop report, RDM, transparent content negotiation, the
WebDAV working group and a WWW5 paper on a distributed hyperlink database. February 1997 adds a
folder called ISoc et related: the IETF, the Internet Society, the EFF, the Tao of the
IETF and the POISED list; the January 1997 message to the
HTTP working group was written in those weeks. The Internet Archive appears in January 1997,
a few months after it started.
The folder politica is the reading list behind Flexion:
the IWW, Love and Rage, Infoshop, the Chomsky archives, the Research on Anarchism archive at
nothingness.org, Social Anarchism Online, the Anarchist FAQ on GeoCities, Biblioteca Arús, La
Baronata, the Siberian Confederation of Labor, and the Zapatista pages that also fed the
1995 consulta ballot. Marx's Civil War in France
arrives by gopher. In 1996 the Mondragón cooperative appears, and in February 1997 the FEEL
forum of Zaragoza, which the mail of those years shows me posting to.
Hardware, from a VGA-to-TV cable to the ATA-2 specification
1995 to 1997 · 47 bookmarks
Pen computing and the Newton in January 1995; the MBone, CU-SeeMe and a VIC-20 page in
1996; OCR software in September 1996. Then two dense weeks: converting VGA to television
(SCART, the sci.electronics.repair FAQ) in December 1996, and in January 1997 the internals of a
PC: the Enhanced IDE FAQ, Intel's PIIX chipset datasheets, the ATA-2 specification, Tom's
Hardware, pinout collections, a databook shelf. Perl and the GNU C++ library close 1997. The
same curiosity about what is inside a machine returns in the sensors and
devices themes of the 2010s.
Related reading: the VGA-to-television converter came from Tomi Engdahl’s pages at the Helsinki University of Technology, and his ePanorama has kept growing from them ever since; pinout collections live on at pinouts.ru and the Hardware Book; Tom’s Hardware is still publishing.
A folder named pilar e historia starts it in 1996 with Herodotus and the Seven
Wonders; in May and June 1997 the reading widens without a folder: Alexandrian scholarship at
Perseus, the legend of Alexandria, Euclid's Elements in Greek and in Joyce's Java edition,
Ptolemy's Geography, the Perseus Atlas, the history of mathematics at Brown, the Catalan Atlas
at the BnF, Oddens's cartographic bookmarks; then John Cassian, the Pachomius library,
Augustine, Byzantine and medieval links, the Avesta, Indology and the Buddhist canon, Gibbon,
the Gnosis pages and Asklepios at Epidauros. This is the material of the essays on
Alexandria and the routes of lost texts (May 1997) and
Democritus, and of the later
conjecture on Ptolemy's coordinates.
Photosynthesis, cyanobacteria and the origin of life
July and August 1997 · 33 bookmarks in a folder named fotosintesis
The most focused folder of the decade. Light-scattering codes, the optical properties of
cirrus, the colours of seaweeds, the Photosynthesis Center at Arizona State, CyanoNews and the
cyanobacteriologists' directory, biochemistry courses on glycolysis and carbohydrates, the
Glycoscience network, an interview with Stanley Miller on exobiology, the origins-of-life pages
at Santa Cruz and Scripps, and an AltaVista query for “origin of life Szostak Miller”. It is the
research behind the 1997 question on photosynthesis and polarised light that became the
ice and chirality conjecture, and the reason the
earth and climate reading of the 2010s keeps returning to ice.
A folder called SF, Espacio, Teoría de la Ciencia: David Brin, Greg Bear, Asimov's
Foundation, alternate-history travel guides, Gibbon, the Federation of American Scientists,
and the Mars Pathfinder news pages in the week of the landing, July 1997. The brin-l list
of the same years appears in the mail archive.
Only a handful: SPIRES at SLAC, the department's bulletin board, CERN preprints, the APS
e-print server, Baez's This Week's Finds in August 1996, his papers in March 1997,
Sprott's Wonders of Physics, the Metaphysical Review, TIPTOP's physics jobs, a
Mexican school on gravitation. Physics lived in the department's server and in the mail, not
in the browser; the bookmarks belong to everything else.
Related reading: This Week’s Finds is complete on Baez’s site and The Wonders of Physics is still running; both authors are on the reading trail, as is Cosma Shalizi, whose notebooks were bookmarked in May 1997 while they still lived at Wisconsin. SPIRES became INSPIRE, and the e-prints are all at arXiv now.
The undated remainder of 1997 tells its own story. Portuguese cinema listings and a
Lisbon cultural agenda in April; the IBM–Kasparov match in May; the Fura dels Baus and Spanish
film magazines through the year. From August to October, a long run of recruiters and careers
pages in quantitative finance and IT, in London, Paris and New York, together with RiskMetrics,
GARP and NumeriX, and, in September and October, rental listings in Madrid. The mail of those
months records the same search and its outcome: by November 1997 I was working at Telefónica
I+D in Madrid, and the Netscape profile there is where the next bookmarks would have been, had
it survived.
2013 to 2026 · 56 bookmarks, most of them in 2014 and 2017
The largest physics thread, and the one that turned into papers. It starts in January
2013 with the bosonic ancestor of open and closed fermionic strings, gathers momentum in
September 2014 around composite models of quarks and leptons, and returns in force in
June and July 2017 with a folder called leer, to read: Regge trajectories,
diquarks, the old dual models with isospin, quarks at the ends of strings, supersymmetric
quantum mechanics with singular potentials, and the superconformal meson–baryon symmetry
of light-front holography. The 2017 reading also took in the non-supersymmetric heterotic
strings and the E8 × SO(16) structure of eleven-dimensional supergravity. The last entries
are from September 2024, group theory from quarks at the ends of strings, and March 2026,
back to the 112 breakings of E8.
Few bookmarks, because most of this work happened in forums and on Wikipedia rather
than in reading. A March 2015 entry keeps my own tweet with the basic mass of the last
Koide triplet; February 2019 adds a geometrical interpretation of the quark and lepton
mass matrices and a paper linking conformal symmetry to the Fermi and Planck scales;
March 2024 is the Wikipedia edit page for the formula, and November 2024 the note on my
blog announcing the published article that joins the superbootstrap with Koide's relation.
The Physics Forums thread What is new with Koide sum rules was bookmarked at its
tenth page in January 2026.
Koide formula on Wikipedia, the article I have edited since 2004
2012 to 2024 · 25 bookmarks, concentrated in August 2014, October 2014 and July 2017
The representation-theory toolbox for the thread above. August 2014 opens a folder
named M y compactificación with the triality of Spin(8), Dynkin diagrams, Coxeter
elements and the projective planes over the division algebras. October 2014 collects the
software: LieART, the Weyl character rings of Sage, Slansky's tables. July 2017 adds the
Atlas of Lie Groups, Susyno, the branching rules of E8 through SO(16) and the decomposition
of E8 tensor products. In 2021 the Perimeter Institute course on octonions and the Standard
Model, and in 2024 a question on the existence of the exceptional groups.
Related here: the earliest bookmark in the group, from 2012, is Tony Smith's page on E6,
strings and branes; his site appears in the reading trail of other personal pages.
The preprints include the spectral-triple notes that
first needed this material.
Compactification, M-theory and strings without supersymmetry
2012 to 2018 · 26 bookmarks
A D-brane primer and David Tong's string notes in the 2012 import; then, in August 2014,
M-theory on manifolds of G2 holonomy and the flux-compactification reviews,
filed next to a Physics Forums thread about whether eleven dimensions can make almost
predictive statements. September 2017 is the densest month: the O(16) × O(16) heterotic
string, the dualities of type 0 strings, a Durham thesis on non-supersymmetric string
theories, Sagnotti's lectures and the heterotic strings in fewer than ten uncompactified
dimensions. The last entry, in November 2018, counts Standard Model fermions against
N = 8 supergravity.
Quantum gravity, minimal length and noncommutative geometry
2012 to 2024 · 21 bookmarks, sixteen of them in September 2024
The noncommutative-geometry strand is old here: the participant list of the 2006 Newton
Institute programme survives in the 2012 import, the NCG blog and the Dutch NCG site were
saved in 2014 and 2015, and a 2019 entry keeps a set of notes on supersymmetry in
noncommutative geometry. Seiberg's Emergent spacetime was bookmarked in September
2019. Then, in September 2024, a folder named gravity appeared in a few days: the
Living Reviews article on minimal length scales, eight arXiv papers on the same question,
and the pages of the journals that publish such work. It was the reading list for a
question and a draft.
September 2015, with two later additions · 15 bookmarks
A single month's folder, fisión: the JEFF evaluated nuclear data report, the
IAEA nuclear data for safeguards, the JANIS browser, the GEF fission model from Bordeaux, a
Tokyo lecture on nuclear physics and Preskill's course chapters. Two IAEA INIS documents
followed in 2017. The point was to test a numerical observation about nuclear patterns
against the tables.
Tong's Cambridge notes on string theory and on quantum field theory, Susskind's
Theoretical Minimum, Kleinert's chapters on path integrals and the Scholarpedia article on
their mathematical side. Beside them, the historians' material: Sommerfeld's letters in
the Pauli correspondence, the AIP exhibit on Heisenberg's doctorate, Einstein's Zurich
notebook, Newton's De motu and the annotated Principia in Cambridge. The
last entry is Kenneth Wilson's Nobel lecture, saved in August 2023.
In June 2013 a folder called baez gathered eight issues of John Baez's
This Week's Finds: weeks 59, 61, 90, 91, 104, 105, 106 and 253, the last one
reached through the n-Category Café. Week 64 followed in 2014, filed with the
compactification reading. The rest of the group is the physics web I kept returning to:
Jacques Distler's Musings (a 2012 comment on shrinking SU(2)), the Reference Frame,
Dispatches from Turtle Island, the n-Category Café's M5-brane model, Scott Aaronson,
the CERN morning meetings and the LHC announcer, the 2013 workshop Why mH = 126 GeV?,
and the particle-physics rumour mill in 2024.
A folder named primos from the 2012 import: the prime-counting function, prime
gaps, Artin's conjecture on primitive roots and the AKS primality test, together with the
Inverse Symbolic Calculator for recognising numbers. Later, the Perron–Frobenius theorem
(2016), the Cayley projective plane (2017), extremal graph theory (2017) and, in December
2020, Alcuin's Propositiones ad acuendos juvenes, the Carolingian puzzle collection.
The working half of the collection. It follows the move from grid
computing and citizen science at BIFI to social-network data, collective-intelligence
platforms and language models at Kampal, and it maps
onto the network,
collective-intelligence and
scientometric themes of the publications.
Social-network data and its APIs
2001 to 2026 · 56 bookmarks
The oldest surviving bookmark of my own is the Yahoo Groups list of August 2001. From
2012 the folder basicFacebook keeps the FQL and REST references of the time (friend,
mutual friends, page fans, likes) and Facebook's Python SDK; Topsy and Google Correlate are
saved in the same months. 2013 adds twtrland's map of who talks to whom in Zaragoza; 2016
the Instagram location pages of the city, Telegram and WhatsApp bot frameworks and the
Twitter keyword analytics of Followthehashtag; 2017 the developer-forum threads about what
automation Twitter still allowed. Botometer arrives in 2020 and a map of Mastodon instances
in November 2022, the month of the migration.
NetworkX, graph-tool and CFinder in the 2012 import; igraph from 2013, including a 2016
commit and the 2017 pull request that added fluid communities. May 2017 is a folder on
community detection: degree-corrected block models, robust detection, the KONECT
collection, graph-tool's inference tutorial. Then the Laboratory for Web Algorithmics
(2018), Snap.py (2020) and, in 2025, a simulation of the Deffuant consensus model on a
Barabási–Albert network.
HDFS architecture, Cassandra and the sort benchmark in 2012; a Spark and Scala winter in
2014–15, with a folder json scala full of JSON parsing questions; Cython, OpenMP,
OpenACC and AVX at the end of 2016; Apache Beam and a folder on parallelising tasks with
Celery and multiprocessing in late 2018; GraphX in 2019, Dask in 2021, Gremlin in 2023.
Each cluster is a project's infrastructure question, solved and left behind.
Word2vec tutorials in July 2016. In January 2020 a folder named ideas keywords
collected the Universal Sentence Encoder, its multilingual version, FAISS similarity search,
keyword extraction and Chrome's reader mode as a text extractor; spaCy and textacy follow
in 2020 and 2021. The Hugging Face hub appears in October 2021. In April 2024 a folder
called llm gathered llama.cpp, Ollama, function calling (Functionary, the
local-llm-function-calling package, the LangChain and Ollama threads) and Dify, alongside
TRL and the implementation details of RLHF. December 2024 adds RAFT and a survey of
knowledge editing; 2025 a critique of MCP, the Agents4Science venue and NeurIPS; 2026
a guide to which model to use, a comparison of open models, and Claude itself.
December 2015, a folder named neuron: LSTM tutorials, backpropagation through
time, learning to execute, character-level convolution, Torch benchmarks; March 2016 adds
DeepWalk and Swivel. Early 2018, the second wave: the information-bottleneck story in
Wired, the Quanta article on deep learning and renormalisation, permutation-invariant
networks, Sinkhorn optimal transport, momentum and adaptive learning rates, reinforcement
learning on CartPole. 2022 to 2024, the third: deep Q-networks in PyTorch, seq2seq and
attention, Keras masking and padding, the VGG paper, and policy optimisation.
2012 to 2024 · 61 bookmarks, fifty of them in 2017
Sideral, the university's research catalogue, is in the 2012 import; Teseo, TDX and the
Dialnet OAI-PMH instructions arrive in February 2016. March 2017 is a folder named
catálogos y repositorios: OpenAIRE and its CRIS guidelines, Crossref's funder
metadata and DOI content negotiation, DataCite, CORE, BASE, OpenDOAR, re3data, Recolecta,
Digital.CSIC, REBIUN, Scopus, Microsoft Academic, AMiner, DBpedia, the Webometrics ranking
of repositories. September 2017 adds the Shanghai subject rankings and the ministry's
university statistics, and July 2017 three arXiv papers on mining GitHub and a PLOS ONE
paper on the popularity law shared by science and Facebook. Later: Handle, Wikidata's item
for the university (2018), Dimensions (2019), a paper on identifying academic rising stars
(2020), NASA ADS and VOSviewer (2024).
The Kampal Research code repository in 2014, Kampal Social and the Comtrade
visualisation in 2015 and 2016, Thinkhub from 2019, and the 2025 manual of collective
learning. Between them, a 2017 post on online tools for collective decision-making, the
2018 final-year project that built an API of open social data for Aragón, and the
monitoring dashboards of the servers that ran it all. Most of these addresses are internal;
the public ones are below.
The national statistics institute and the LiDAR flight of Spain in 2013, the world top
incomes database in 2015 and its successor WID.world in 2016 and 2017, UN Comtrade,
GDELT, the Wikipedia clickstream dataset (2018), the public-procurement platform, the
electoral results service, the atlas of income by district (2019), mortality by any cause
during the epidemic (2020) and Gapminder's worldview upgrader (2021). The spring of 2020
has its own cluster: an SEIR model against healthcare capacity, Nicky Case's playable
simulations, provincial case data, the genomic epidemiology viewer Microreact, and a paper
on how generation intervals relate growth rates to reproduction numbers.
BuiltWith and W3Techs in 2013, when a transcript-search site needed to know what the
web was made of; Baidu's crawl-pressure page for that same site. In 2018 and 2019 the
question became how to run one's own search: public searx instances, Common Search's
attempt at an open PageRank, the Open PageRank of the top ten million domains from Common
Crawl, Startpage, SerpApi, the Bing API pricing, the top websites in Spain, and a fork of
pytrends published on GitHub Pages in 2019.
BOINC project options and the Ibercivis server monitor, the GStat view of the grid, the
national grid resources wiki and RedIRIS traffic graphs, all from the 2012 import; the
OpenStack dashboard at BIFI in 2013; the SCI-BUS gateway publications in 2014; the CESAR
cloud at the university from 2016 to 2020; the EGI applications database in 2019.
2012 to 2025 · 117 bookmarks on web development and Python
The least thematic and most revealing group. Rails and Vue questions cluster in 2019 to
2021, when Thinkhub was being built; PostgreSQL arrays and pivot tables, peewee and Django
migrations, Celery again. On the Python side, gnuplot demos in 2012, TiddlyWiki as a
personal notebook in 2013, Jupyter viewers, pandoc, ctypes and mixed Python–C++ debugging,
Dask, Scikit-HEP in 2021, and the standard-library documentation for regular expressions
as late as 2023.
A quarter of the classified bookmarks are about making physical
things move or sense. They continue the open-hardware
strand: the webcam mouse, the bicycle on rails, the electronic-paper readers.
Lego mechanisms and Mindstorms
2012 to 2025 · 92 bookmarks, seventy of them in March and April 2017
March 2017 is the month of linkages: the Tinkering Studio's Lego linkage experiments,
four-bar, Chebyshev, Klann and Hoeken mechanisms, a jig for drawing ellipses, hypotrochoids,
SolveSpace as a linkage solver. Around them, the mechanical literature of the Lego world:
Sariel's worm-gear casings, Isogawa Yoshihito's creation book, TechnicBRICKs on the Torsen
differential, Philo's measurements of motors and the internals of the NXT motor, JK
Brickworks' drawing machine, the LEGO Education machines-and-mechanisms instructions, ev3dev.
September 2018 adds a chain precession drive and a marble counter; January 2019 a pinwheel
escapement built in Lego and brickr, which turns 3D models into brick instructions; 2022
the Mindstorms hub API and Pybricks; 2024 a fischertechnik optics kit.
2013 to 2021 · 84 bookmarks, sixty in August and September 2020
A DIY cellphone from the MIT Media Lab in 2013; the nine-dollar CHIP computer, the
Raspberry Pi Zero and its display in December 2015; a January 2016 folder on hacking a
Roomba through its serial port. The big cluster is a folder called usb hub from
the summer of 2020: how to switch USB ports off programmatically (uhubctl, YKUSH, a hub
modified for per-port power), AC power control through a hub, then a drift towards
vibration sensors (the SW-420, the MPU-6050 and its motion processor without an interrupt
pin, the Aqara Zigbee sensor read through zigbee2mqtt), NodeMCU and ESP32 boards, Tasmota,
Sonoff and Shelly switches, the Arduino Nano 33 IoT, and speech synthesis from a web page.
A small solar panel in 2021 closes it.
The Ryze Tello programmed from Scratch, the Eachine E011 and its open Silverware firmware,
the BoldClash whoop, the RT-Thread operating system behind a Chinese Wi-Fi shield, a
drone-hacking reading list and a note on logging into an Eachine E50S. The same summer
produced a repository for controlling a JJRC toy drone from a PC.
A folder called codinggames, October and November 2017: CodeCombat, CodinGame,
Codewars, HackerRank's checkers leaderboard, Bebras, Programamos and its Scratch tutorials,
code.org, and the tools that port Scratch projects to the open web. Educational robotics in
early 2018 (Makeblock, Estalmat, the mathematics-talent programme), Cuisenaire-rod worksheets
in August 2018, GeoGebra and Twine along the way.
A Geiger counter board for Arduino (2014), a chemical garden and the nutmeg chemistry
of Erowid's archive (2014), an ionocraft (2015), a turn-key cloud chamber (2017), the
chemical chameleon demonstration (2018), Fusor.net and the Spanish amateur-scientists forum
with its scanned TBO comics (2019), a printable planisphere, a double-pendulum simulator,
the Roche biochemical pathways map, bean-to-bar chocolate (2021) and a fischertechnik
optics kit (2024).
Owning a device means being able to look inside it. This group
is the practical side of the free-hardware essay.
Linux, small machines and other people's operating systems
2007 to 2023 · 85 bookmarks
The oldest are the Ubuntu links of a 2007 laptop and, undated, the manual of the Links
browser in English and Czech. openSUSE is the constant from 2012 to 2017, with its default
folders reappearing in every profile. In April and May 2013 a folder named
interesantes tracks the effort to put Arch Linux, Debian or Fedora on the cheap
WM8650 ARM netbooks of the time; the Zipit Z2 messenger had received the same treatment in
2012. Later come RAID rebuilds, rsnapshot, mosh, and a long tail of one-off fixes. The
Apple side runs from a 68k Macintosh emulator and a hackintosh in a virtual machine
(2016) to Catalina on unsupported hardware (2020).
Phones, televisions, routers and how they are locked
2011 to 2025 · 103 bookmarks
Rooting a Galaxy Note in 2013 and repairing it in 2014; a Nokia 770 and Bluetooth
keyboards on Maemo; a smartphone as a dial-up modem, in 2017, for a place with no other
line. The summer of 2016 belongs to Pokémon Go: a map server on my own domain, then the
question of how the game pinned its certificates, which led through Frida, Substrate,
radare and the pgoapi project; Frida returns in 2019. Samsung and LG televisions were
opened up the same autumn with SamyGO and OpenLGTV. On the network side: a BT Openreach
modem in 2013, dynamic DNS on Route 53 in 2017, the admin password of a Vodafone router
and an IPv6 tunnel broker in 2018, VPN routing in 2020. The security reading is of the
same kind, from a 2012 Linux rootkit analysis and Shodan in 2014 to zip bombs as a defence
in 2025.
April 2013, a folder named tesoros: the Aragón archives portal DARA and the
Virtual Library of Bibliographic Heritage searched for treasure, next to the 1848 penal
code; the Old Bailey proceedings of 1695 and 1876 and the Anglo-American legal tradition
site had been saved the previous autumn, with the Newton Project's diplomatic edition of
De motu. The second wave is August to October 2022: the Zaragoza newspaper
library (issues of 1869, 1890 and 1910), the national newspaper library, the Gazeta of
1839 and the BOE of November 1938, the search engine of the Civil War repressed, the
history of Torrero cemetery, the university's historical archive, a family name in the
Prensa Histórica catalogue. In between: Migne's Patrologia Latina (2015), a
photograph of the Moncloa palace in the Royal Library (2015), the Census-Guide of Spanish
archives (2017), the Oracc cuneiform projects (2018), digitised Swiss journals, the CNIG
photo library.
Ladino through the Jerusalem magazine Aki Yerushalayim and an Aljamiado library
in 2013; Arabic self-study pages in January 2015; the Etruscan word for water on the
WordReference forums, saved twice, in 2014 and 2017; Philogelos, the ancient joke book, in
2016; the months of the Voynich manuscript in 2019; the genetic map of Europe; the
arabisms in Spanish playground rhymes, from Federico Corriente's academy speech (2024);
and in 2025 the Upper Palaeolithic proto-writing paper, the Old European Culture blog,
Yomi, and the ScanPyramids report. Beside them, the places: Pluto's gate at Pamukkale,
Naqsh-e Rostam, the underground cities of Cappadocia.
Participation, 15M and the tools for deciding together
2012 to 2022 · 61 bookmarks
The first anniversary of 15M in May 2012, the trend graphs of its hashtags and the guide
for the indignant; the Partido de Internet's Ágora on Rails. In 2013, Agora Voting, David
Ruescas on liquid democracy, Metagovernment's list of active projects, a sociological
analysis of a unitary citizen candidacy, and the Medialab-Prado pads. ListAbierta in 2014;
the collaborative programme of Ahora Madrid in 2015 and, in June 2015, a hand-collected set
of thirty tweets from a single controversy about a new councillor, kept as raw material
for a conversation graph. In 2016 apps4citizens and a study of participation through
Twitter; in 2017 a post on online tools for collective decisions and the account of the
cyberattacks during the Catalan referendum. The tail runs into 2022.
The bitcoin block explorer and the network graphs of 2013, the mining claims of the
American West, the world top incomes database (2015), Nada es Gratis on complexity and
financial regulation (2016), Bronte Capital and Mish's global trend analysis, graphs of the
Spanish property sector, the Ethereum Foundation (2017), the IPO calendar (2021).
Two small groups. On minds: the Small World of Words association study (2012), the Global
Consciousness Project (2013), online IQ tests and the Psychometrics Centre's profiling
demonstration in 2016 and 2017, the article on how a campaign might know a voter better
than the voter does, the Open-Source Psychometrics Project (2019), LessWrong and Kialo. On
the planet: the thirty years of the iron hypothesis of ice ages, how maximum rates of
climate change are underestimated in the geological record, Skeptical Science, the
crystals that record Earth's first magnetic field, the rocks of Mars, the Quaternary
glaciation, and wind maps at 300 hPa.
The remaining third of the collection is daily life, and most of it stays private. What
can be said in public is short. The webcomics were Gunnerkrigg Court, Schlock Mercenary,
Girl Genius and Vexxarr; the manga Dimension W; the series Doctor Who in its
2015 season and Girls und Panzer in 2022, with a Finnish news item about its
pilgrims. The games were Minecraft on a home server in 2015, Pokémon Go and Overwatch in
2016, 0 A.D., the graphic adventures from Monkey Island to Thimbleweed Park, and chess
watched rather than played: AlphaZero against Stockfish in December 2017, lichess and the
live ratings in 2019. A recovered copy of Maxis's SimRefinery was bookmarked in 2020.
The travel bookmarks document a stay in Dubai in early 2015, with the ferries to Iran
studied and never taken; a week of Ordesa and Monte Perdido in August 2019; London and
Oxford in October 2019, punting included. The recurring daydreams are older: round-the-world
tickets (2013), freighter travel (2012 and 2019), the underground cities of Cappadocia and
Edinburgh (2014), the water towns of China and Zhangjiajie (2019), Tokyo (2021). Several of
these feed the travel wish list, and the documented stays
belong with the places I have been.
Notes on method and limits
Classification was done by rules on folder names, domains and titles, then corrected by
hand; the private working files keep the full table with one theme per bookmark. Themes
overlap: a paper on E8 branching rules sits with the superbootstrap reading because that is
the folder it was saved in, and the same paper appears under exceptional groups when it was
saved again years later. Counts are of distinct addresses; a page bookmarked twice from
different profiles counts once, at its earliest date.
Two known gaps. Bookmarks from 1998 to 2010 barely survive, because the browsers of
those years lived on machines that were wiped: the Telefónica workstation, the Carlos III and
Teruel computers, two laptops. The chronology in Messages covers that
period from other sources. And bookmarking itself thinned out after 2020, so the absence of a
theme in recent years says little about whether the interest is alive: the 2024 gravity folder
shows how quickly a reading list can reappear.
This page was built on from
an export made on 4 March 2026, extended the same day with the three 1995–1997 files. New bookmarks will be folded in at the next export.