Sources
Everything I have found, and how well I know it.
There are 42 badged rows on this page, covering 63 sources, and 70 entries in the full index. Each row carries a badge saying how well I actually know it: whether I read the bytes, got only a machine summary, hit a bot wall, found the host gone, or decided not to read it and said why. Those are different things, and a plain list of links hides the difference.
Nothing on this page is an unopened lead any more. As of 12 September 2026 every entry in the index has a disposition, and scripts/check-inventory-dispositions.sh fails the build if a new one doesn't. A row can cover more than one source - Chandler and its five postmortems are one row - which is why 42 rows and 63 sources are both true. Ten of the 70 index entries aren't here: the three emulator config files and Ormandy's two sample print files live on Downloads, the scan of the Agenda 1.0 disk is on history where the picture of it is, and four biographical sources sit in the private index because they are about the people, not the product.
Canonical
Read the CACM paper first. It specifies the engine completely and then tells you exactly where it stops.
| Source | Read |
|---|---|
| CACM paper - “Agenda: A Personal Information Manager” Belove, Drake, Kaplan, Kapor, Landsman. CACM 33(7):105-116, July 1990. The architecture document. This transcription removed the figure references, so the illustrations are missing - the ACM DL has the original. | RAW |
| US Patent 5,115,504 - “Information management system” Filed 1 November 1988, granted 19 May 1992. The five authors plus Stephen Zagieboylo, assigned to Lotus. The claims are the paper's storage invariant in legal form: item, category, and view objects kept independent and joined only by pointer networks, so the hierarchy can change without touching an item. | RAW |
| The six original manuals Uploaded by Tavis Ormandy, 2020. I've read Working with Definition Files in full, and Part IV of the Users Guide - the four chapters on conditions and actions, which turned out to be the specification I was looking for. The other four are on disk, unread. Direct links on the downloads page. | RAW SUM |
| Tavis Ormandy - “Lotus Agenda” The best modern operational account. Real filter syntax, macro examples, STF notes, DOSEMU2 setup. | RAW |
| Wikipedia - Lotus Agenda Dates, versions, commercial facts. It cites the patent, which is how I found the number, and takes its sales figures from Fallows. | RAW |
| stfjson Converts Agenda's STF export to JSON. Written against inferred structure - see the correction below. | RAW |
A correction, sitting next to the claim it corrects.
I had Working with Definition Files ranked as the most valuable thing here I
hadn't read. I expected the STF specification, a pattern tutorial, and maybe some
relief on the algorithms the paper withholds. I read all 356 KB of its OCR text
on 6 September 2026. It is the manual for TXT2STF, the text-import converter.
Across the whole manual: STF 650 mentions, match 314,
action 8, condition 7,
regular expression 0.
So it doesn't touch the inference gap. Chapter 4's patterns parse input text files, not items into categories. It does have a tag-level STF specification in Appendix B, which I had written down as missing entirely.
Getting it running
Agenda 2.0 is freeware and still runs, which makes it a live oracle for the behavior the paper won't specify.
| Source | Read |
|---|---|
| Official Lotus freeware FTP, via the Internet Archive The original host no longer resolves. This is the July 2021 snapshot, and it still serves all four files. | RAW |
| Preinstalled ZIP - Bob Newell Bypasses the floppy copy-protection. Also answers at agenda.bobnewell.net, byte for byte. The page describes the ZIP but no download link survived the fetch, so I never used it: the original disk images install under DOSBox Staging's floppy emulation, which is how the screens were made. | RAW |
| Agenda FAQ - Ormandy Copy protection, keybindings, troubleshooting. The key-rebinding recipe and the garbled-display fixes are DOSEMU2-specific. | RAW |
| DOSEMU2 Renders VGA text mode as terminal escape sequences. | N/A |
| DOSBox Staging What I run it under. Version 0.83 mounts a stack of floppy images and swaps them with a hotkey, which is all the copy protection needs, and its local REST API exposes the emulated machine's memory - the text screen included. | RAW |
| WinWorld 2.0 · 1.x · forum thread Mirrors, and a thread with people who ran it. | N/A |
| old-games.com · vetusware Further binary mirrors. | N/A |
| SCANCODE.COM TSR for keystrokes an emulator won't pass through. | N/A |
Bob Newell's archive
Bob Newell has been collecting this for decades. Much of it is nowhere else.
| Source | Read |
|---|---|
| Index of everything The same bytes as the preinstalled-ZIP page above and as agenda.bobnewell.net: one page, three addresses. | RAW |
| Agenda FAQ, 2006 revision - Tony Kan Carries the real scaling limits. | SUM |
| 1990-91 Lotus Support Q&A 78 support entries, June 1990 to May 1991. Ormandy singled it out as fascinating, and the 19 October 1990 entry states the write-back rule the emulator measured. | RAW |
| Agenda 2.0 help file, as HTML Michael Rourke. The in-product help; its Automatic Assignments page is the clearest statement of why conditions break and actions stick. | RAW |
| President's Planner documentation The major commercial third-party add-on. | RAW |
Also on that index and not yet fetched: the Agenda Tech Notes database, the “Beyond 2.0” manual, Tony Kan's preinstalled build, Phil Robinson's course, the Dez Ellis tutorial, Handling Damaged Agenda Files, Creating ASCII Files from Agenda, Partner's Planner, Supermacros, First Things First, and a bibliography of books about Agenda.
What came after
Two serious attempts, one abandoned and one alive, plus the product Lotus shipped instead.
| Source | Read |
|---|---|
| Beeswax Rust TUI, MIT. Abandoned 2008, restarted 2026, in daily use by April. The existence proof for the engine tier. | RAW |
| Beeswax project page Carries the 2026-04-22 note explaining the restart. | RAW |
| Chandler Kapor's own attempt, 2003-2009, and the reason the scope fence on this project is written down. Postmortems: Cantrill, Bernstein (Internet Archive), Tornhill, Building Real Software, and FOSS Academic. All five read in full, and they converge. Bernstein puts it best: Chandler "suffered from a baked-in incoherence - it was a successor to Kapor's agenda and so it was about agents and lexical properties of notes, but it was also supposed to leverage ubiquitous email and calendaring, and so wound up being about dates and places. That's too much to fit into a product." Cantrill refuses the generic excuse that software is just hard, calling it a tale of bad ideas and poor execution. Tornhill lands on the same root cause - nobody knew what they were building - and then warns that spotting it afterwards is easy. Three of the five never mention Agenda at all. Bernstein's host stopped answering over HTTPS on 11 September 2026, so that one is the February 2024 snapshot. | RAW |
| Dreaming in Code Rosenberg's book on why Chandler failed. | RAW |
| Lotus Organizer The Windows replacement. A paper-diary metaphor - a different product entirely, and not even a Lotus one: Organizer was built by Threadz, a small British software house, up to version 3.0, and Lotus bought it. It led the market for years until Outlook took it. The Lotus Organizer Museum is a tribute page that is also an advertisement for a 2026 successor product; read, and its only Agenda content is one sentence saying Organizer replaced it in 1992. | RAW |
| Slashdot, 2006 - “Lotus Agenda Returns as Open-Source Chandler” | RAW |
Commentary and context
People keep coming back to a program whose last release was in 1992.
| Source | Read |
|---|---|
| “Hidden Agendas, Lost Cities” August 2025. The best modern essay on why the paradigm mattered. Its closing image is Borges's “Tlön, Uqbar, Orbis Tertius”, which I opened; it's context, not a source. | RAW |
| Brad Feld - “Roam: My New Favorite Software Product” Ran his first company on Agenda, 1990-93. | SUM |
| Computer Chronicles, 1989 Live Agenda demo at the 21-minute mark. | N/A |
| Computer Chronicles Revisited #124, via the Internet Archive Tornado Notes, askSam, GrandView, and Agenda - the whole category, contemporaneously. The blog itself stopped responding, so this is the September 2025 snapshot. | RAW |
HN - “Rediscovering Lotus Agenda” · HN, 2015
October 2020, 74 points, and it is where the Org-mode-as-successor argument actually lives - read in full on 12 September 2026, and the claim is explicit: Org mode “has this functionality in its own Agenda features,” plus the third-party org-ql for saved-search views, and “Org mostly surpasses them.” One commenter's assessment, not a feature-by-feature comparison, but the claim now has an address. And a 2015 comment that Agenda and Ecco Pro “have never been replicated, despite multiple attempts to clone them on modern platforms.” One comment, but it's the thesis of this site in one sentence, from a stranger, eleven years ago.
|
RAW |
| DEVONthink - “An app ahead of its time” · MetaFilter · LinuxQuestions People looking for a replacement, repeatedly, for years. | RAW CHALLENGE |
| Building library catalogues in Agenda · Agenda and ConnectedText People actually using it, long after 1992. The ConnectedText post now 404s, so that one is the May 2020 snapshot. | RAW |
| James Fallows - “Zoot!”, The Atlantic, August 1997 The sales figures everyone quotes come from here: roughly 200,000 copies of Agenda at about $195, and 500,000 of Magellan, before Lotus cancelled both. Fallows was still running Agenda in 1997 and reviewing Zoot as the first program to match it. | RAW |
| Palmtop Paper - “DOS Connection: Lotus Agenda” The HP palmtop era. | RAW |
| Howard Rheingold - Tools for Thought Full text online. The intellectual context this whole category sits in. | RAW |
Two Lonsdale Systems training decks survive and are the only structured curriculum found anywhere: the course and applied self-management. Both read in full on 12 September 2026. The self-management deck is the strongest independent confirmation of the data model found anywhere outside Lotus's own documents; the course deck teaches Agenda as a database hybrid and never as a PIM.
The modern field
What I'm competing with, and against. I record these because the question is why nothing since has reproduced it.
| Source | Read |
|---|---|
| taskwarrior Ormandy's daily driver, after Agenda. | RAW |
| taskwiki Ormandy's pick for the closest modern equivalent to Agenda - which is a qualified endorsement, given he had just spent weeks running the original. | RAW |
| Capacities vs Tana · Tana vs Capacities vs Logseq The current generation, compared against each other. All three rebuilt promiscuous tagging; none rebuilt automatic assignment. | RAW |
Emacs Org mode recurs across Hacker News as the strongest living descendant. I couldn't find a canonical Agenda-comparison URL for it - the claim lives in comment threads, which is why it has no row above.
Bibliographic
These are for finding the paper of record with its figures intact.
| Source | Read |
|---|---|
| CACM July 1990 issue index Behind a Cloudflare challenge, so an automated check can't reach it. I loaded it by hand on 6 September 2026 and it's alive. | CHALLENGE |
| dblp - Richard A. Landsman · dblp CACM vol. 33 · Utah TeX bib, CACM 1990 TOC (Internet Archive) | CHALLENGE N/A DEAD |
| Kapor oral history - Computer History Museum, 2004 Recorded 19 November 2004, William Aspray interviewing; 23 pages. Mostly VisiCalc and 1-2-3, but Agenda gets a paragraph, and it's the one where Kapor calls it his favourite product and names Kaplan and Belove as his co-designers. Catalogue record. | RAW |
| Kapor oral history - IEEE History Center, 1993, via ETHW Andy Goldstein interviewing, 20 May 1993. Kapor on Lotus's product experiments, Agenda among them: the reviews and the devoted users, and not the sales. The IEEE reserves quotation rights, so I'm paraphrasing; read it there. | RAW |
What isn't on the web
I write these down so nobody spends a day looking. Three have closed - one on 6 September 2026 and two on the 12th - and a fourth narrowed on the 11th. The last of the three closed as a decision rather than a find.
The inference algorithms NARROWED 2026-09-12
The paper withholds them explicitly: string matching is “beyond the scope of
this article,” and update-through-views is credited to unspecified
“formal methods and heuristics.” Half of that turned out to be in the
Users Guide. Chapter 19 gives the string-matching rules in full: how text is
cut into words, numbers, dates, and proper names, the suffix list, and the score,
which is the fraction of a category's words present in the item, thresholded at
100%, 50%, or 2%. Chapter 20 documents the one-category form of view-update
inference as a setting, and chapter 11 gives the filter half as a plain rule: a new
item gets the filter's category, or the start of its date or number range. The
ambiguous case, a filter across an exclusive family, was put to the running 2.0b
binary across seven runs on 2026-09-11 and 12: a new item copies the item directly above
it, and with nothing above it takes whichever filter was entered first. That is the paper's
“look at the vicinity” heuristic in its narrowest form, and the only rule that survives all
seven. A Lotus support note from October 1990, read after the first six, states the same
rule and adds that Agenda 1.0 did the opposite; the 1988 patent claims it too, as “data
immediately adjacent to or preceding” the new item in the view, and names the item
below as an alternative, which the seventh run then tested: with nothing above
and a High item directly beneath, the new item took the first filter, so the shipped
program didn't use the item below in that case.
The other two write-back paths went to the binary on 12 September. Column entry turns
out to do three things at once: type a name into a view's People column and Agenda creates
that category under People, assigns the item explicitly rather than by rule, and drops the
conditional “Unassigned” it had been carrying. Section moves answered differently:
there is no such operation. The only command that moves an item is
Reposition, which says “within this section” and refuses outright when the section sorts
itself. Sections aren't containers you drag things into; membership is just what the
assignment implies, and assigning “High” relocates the item the instant you press ENTER.
Query synthesis, the last piece, turned out not to be a missing algorithm at all. The paper's
one rule for it is that nobody means a query that must come back empty, which is exactly why
picking categories that can't share an item gives you either one rather than both, and the
running binary showed on 14 September that this holds even for categories that
only exclude each other at a distance. What's still open is narrower: what a view does when a
category it names is deleted.
A complete STF specification CLOSED 2026-09-06
Recorded as missing for years; stfjson was written in 2020 against what its author could infer. Appendix B of Working with Definition Files is a tag-level specification - tag tables, a category-character table, date categories, the escape character, item tags - followed by worked examples pairing a text file, its definition file, and the resulting structured file. It was in a manual the same author had uploaded himself.
The Lotus patent numbers CLOSED 2026-09-12
US 5,115,504, “Information management system”. Filed 1 November 1988, granted 19 May 1992, assigned to Lotus and now held by IBM. Six inventors: the five authors of the paper and Stephen Zagieboylo. Wikipedia's Agenda article cites it; Google Patents has the full text. Its first claim is the paper's first invariant written as law: items, categories, and views are independent objects joined only by networks of pointers. Lotus filed no siblings. An assignee sweep of Google Patents for “Lotus Development”, priority 1986 to 1992, returns thirteen patents, and 5,115,504 is the only one with an Agenda-team inventor. Agenda is a single patent.
Agenda source code OPEN
Never released. The freeware release was binaries only.
Box art and manual covers with a licence CLOSED 2026-09-12
Settled on 12 September 2026, and settled as a decision rather than a discovery. Nothing rights-clear exists: all six Lotus Agenda items on archive.org declare no licence, Wikipedia carries only the corporate logo, the Computer History Museum requires written permission it could only grant for its own photographs and not for Lotus's artwork underneath them, and the 1998 freeware release turns out to have contained four files - disk images, an update kit, legal-profession templates, and icons - with no manual, no box, and no licence document at all. The manuals are under copyright until 2084.
None of that makes reproduction unlawful. No licence exists is a different finding from you may not, and fair use sits outside all four. The closest case is Bill Graham Archives v. Dorling Kindersley, where concert posters reduced into a book's timeline were held transformative as historical artifacts - and that use was commercial, where this one isn't.
So one image ships, small: the Release 1.0 program disk on the history page, scaled down from the archive's scan. It's a disk label rather than box art, because no box photograph turned up anywhere - and it carries the release and part numbers, which a box wouldn't. Everything else stays a link: the manuals on Downloads, no full-page scans, nothing at a size that could stand in for the original.