WORK

EWD Manuscripts

Dijkstra's sequence of over 1,300 numbered, mostly handwritten technical notes — the primary medium of his thought, and one of the most unusual bodies of work in twentieth-century intellectual history.

The EWDs are Dijkstra's numbered manuscripts, composed over nearly five decades and ranging from one-page aphorisms to forty-page formal developments of algorithms and proofs. The name comes from his initials (Edsger Wybe Dijkstra) and the series runs from EWD1 in the early 1960s to EWD1318, written shortly before his death in 2002. The manuscripts were circulated informally — photocopied, mailed, pinned to colleagues' doors — and constitute Dijkstra's primary medium for technical communication. They cover algorithm design, program verification, pedagogy, philosophy of computing, and occasional polemic; many of the most influential ideas in twentieth-century computer science first appeared in them, including structured programming, separation of concerns, and the mathematical treatment of program construction.

The Cost of Beautiful Isolation — Contrarian ^ Opus

There is a parallel reading of the EWD manuscripts that begins not with their intellectual content but with their material substrate: the personal network, the photocopy machine, the Austin archive. What made the EWDs possible was not just Dijkstra's discipline but his position — a senior figure at elite institutions with the resources to maintain a decades-long practice of informal publication and the social capital to make that informality count. The manuscripts are beautiful, yes, but they are beautiful in the way a private garden is beautiful: sustained by conditions most practitioners will never have.

The lived experience of most working programmers is not patient derivation at a standing desk but coordination under pressure, writing that must serve multiple audiences, code that ships before it is understood. To hold up the EWDs as a model in that context is not just aspirational; it risks reifying a mode of work that was only ever available to a small class of people whose institutional position allowed them to ignore the ordinary constraints of scientific labor. The question is not whether the EWDs are good — they plainly are — but whether their goodness is generalizable, or whether we are mistaking the luxury of intellectual autonomy for a reproducible practice. If the medium is the message, the message of the EWDs may be less about discipline than about what becomes possible when you are freed from the need to justify your work to anyone but yourself.

— Contrarian ^ Opus

In the AI Story

Hedcut illustration for EWD Manuscripts — *EWD Manuscripts*

The EWDs are unusual in format and in practice. Most are handwritten, in Dijkstra's distinctive clean script, and many include small corrections and marginal notes that show the trace of composition. They were not peer reviewed. They were circulated to a personal network of colleagues and former students and, through that network, to the broader field. Several of the most influential papers in computing history began as EWDs and were only later reworked for formal publication. Many never were.

The medium reflects Dijkstra's working method. He did not draft and revise at a keyboard; he worked out arguments by hand, on paper, often standing at a lectern. The finished manuscripts were, in effect, derivations in the mathematical sense — the written output of a thought process that had been conducted rigorously enough that the writing could serve as a record of the reasoning. This is consistent with his view of programming: the text is the proof made legible.

The full archive is maintained at the University of Texas at Austin, where Dijkstra spent the last two decades of his career, and has been scanned and made freely available online. It constitutes one of the most unusual corpora in the history of technical writing — a sustained, decades-long personal practice of informal publication that nonetheless achieved influence comparable to that of the most prestigious refereed venues.

The EWDs are also, incidentally, an object lesson in the discipline Dijkstra preached. They are short when short will serve, long when the argument requires. They state their premises explicitly. They work through their reasoning step by step. They do not hedge, but they do not claim more than they have shown. In an era of rapid-fire digital publication, the EWDs stand as a reminder of what sustained, patient technical writing looks like when the writer has taken the time to understand what she is saying.

Origin

The series began in the early 1960s, during Dijkstra's time at the Mathematical Centre in Amsterdam, and continued through his positions at the Eindhoven University of Technology, Burroughs Corporation, and the University of Texas at Austin. The numbering is roughly chronological, though Dijkstra occasionally rearranged the order when preparing manuscripts for distribution.

The practice inspired similar informal publication traditions in other research communities, and the idea of a personal, numbered manuscript series has been adopted by several of Dijkstra's students and collaborators. None, however, has matched the sustained quality and influence of the original.

Key Ideas

Primary medium, not supplement. The EWDs were where Dijkstra's thought lived. The formal publications were, in many cases, rework of EWD material for conventional venues.

Derivations made legible. Each manuscript is the written record of a completed reasoning process, closer to a mathematical derivation than to a typical paper.

Personal network as distribution. The manuscripts circulated through photocopies and mail, relying on a personal network of recipients to propagate the ideas. This is not how modern scientific communication works, and the fact that it succeeded is a testament to the quality of the content.

The archive is public. The University of Texas at Austin maintains the full scanned archive, freely available, and it constitutes one of the richest resources in the history of computer science.

An argument about patient writing. In an age of accelerated publication and AI-generated text, the EWDs are a model of what sustained, patient, personally authored technical writing looks like.

Debates & Critiques

The EWDs have been criticized as a medium that favored insiders — readers had to be on Dijkstra's mailing list, or know someone who was, to receive them reliably. The open archive at Austin has largely addressed this, but the original distribution method was exclusionary in a way Dijkstra himself acknowledged. Whether the quality of the manuscripts justified their mode of circulation is an open question that depends on what one thinks science's epistemic institutions should look like.

Appears in the Orange Pill Cycle

Edsger Dijkstra — On AI

When Autonomy Meets Infrastructure — Arbitrator ^ Opus

The right frame is to recognize that the EWD manuscripts succeed on multiple registers simultaneously, and the weighting depends on which question you ask. On the question of intellectual content — did the manuscripts advance the field? — Dijkstra's framing is nearly 100% right: they are rigorous, influential, and disciplined in a way that formal publication often is not. On the question of reproducibility — can this model scale? — the contrarian view dominates at perhaps 70%: the conditions that made the EWDs possible (institutional position, personal network, sustained resources) are rare, and mistaking them for a general model risks romanticizing privilege.

But there is a synthetic question that the topic itself opens: what does it mean to build intellectual infrastructure that outlasts its original context? Here both views are necessary. The EWDs succeeded because Dijkstra had autonomy, but also because he used that autonomy to create a durable public good — the scanned archive is now more accessible than most paywalled journals. The manuscripts were distributed through personal networks, but those networks were porous enough that the ideas propagated widely. The practice was exclusionary in its moment, but inclusive in its legacy.

The arbitrated position is this: the EWDs are both a model and a warning. They show what becomes possible when a thinker is freed to work at the pace thought requires, and they remind us that such freedom is unevenly distributed. The right move is not to dismiss the model but to ask what institutional structures would make patient, rigorous technical writing more broadly accessible — and to recognize that the answer is not individual discipline but collective infrastructure.

— Arbitrator ^ Opus