PKM Wiki

Source - Matuschak 2026 - Apps and Programming Two Accidental Tyrannies

5 min read

Citation

Andy Matuschak, "Apps and Programming: Two Accidental Tyrannies," MIT HCI Seminar, March 3, 2026. URL: https://andymatuschak.org/tat/

Overview

Matuschak's thesis is that the contemporary landscape of digital tools is shaped by two inadvertent constraints that have narrowed what computing can be: the app model (software that siloes functionality behind one-size-fits-all interfaces users cannot meaningfully extend) and programming as specialization (the restriction of interface invention to those with technical fluency, starving the field of designers and domain experts with novel vision). Both constraints, he argues, can now be partially lifted — the first via composable architectures like CodeMirror's facet model, the second via coding agents that let non-programmers prototype directly. The essay is structured around three sections: Apps, an Interlude on digital reading, and Programming.

Main Thesis

  1. The app model is a tyranny. Users can only combine tools "at junctions the developer anticipated." Meaningful customization requires reimplementing the surrounding software.
  2. Specialized programming is a tyranny. Novel interface invention requires imaginative design skill, technical fluency, and domain insight. Because the technical requirement is hard and the others are not taught together, programmer-inventors dominate the field, and non-programmer designers/domain experts must pitch ideas to programmers to reach prototypes. This gatekeeping prevents many good ideas from ever existing.
  3. Both constraints are being partially lifted. Composable architectures (CodeMirror's facet system) allow plugins to declare what they contribute without coupling to other plugins. Coding agents allow designers and domain experts to prototype directly — what Matuschak (citing Nio Ono) describes as "vibe-coding" when ideas are still inchoate.

Key Concepts

  • Research purgatory — the state in which novel interface ideas remain trapped in academic prototypes, never reaching actual production use because "each has to pay an enormous tax to escape"
  • Facets — declarative structures through which plugins contribute to shared concerns (decorations, keybindings, state) without mutating shared state, avoiding the fragile coupling of systems like Emacs
  • Physics of a medium — the foundational model (for text: state, selection, decorations) that enables extensible, composable design; the thing everything else is built on top of
  • Vibe-coding — exploratory prototyping using coding agents when design concepts are still inchoate; "trying all kinds of ideas I just wouldn't have tried a year ago"
  • Gatekeeping dynamic — the structural pattern where non-programmers must pitch ideas to programmers for execution, preventing iterative exploration on genuinely novel concepts
  • Real-enough prototypes — prototypes that demonstrate dynamic feel and communicate intent without needing to be production-quality; reusable prototype code need not ship

Structural Observations

On composability (CodeMirror case)

  • CodeMirror is built from plugins, not with plugins bolted on
  • Declarative facet system: plugins declare contributions rather than mutate shared state
  • Users get first-class access to the same primitives the core developers used
  • Obsidian is cited as an analogous case — plugins using "exactly the same primitives that the Obsidian developers used to create their product"

On reading (Interlude)

  • Reading environments have barely evolved; current tools mostly display "pictures of paper on screens"
  • Multiple research prototypes (Scim, CiteSee, Papeos, LiquidText, SpaceInk, Quantum Country) demonstrate rich augmentation possibilities
  • All sit in research purgatory, isolated from each other
  • A composable reading substrate with facets could support all these ideas simultaneously

On education

  • HCI/interface design programs are dominated by computer science pedagogy
  • "We may need programs that look a little more like art school with technical electives"
  • Rebalancing curricula toward design-first training with technical support could diversify the invention pool

Practical Recommendations

  1. Architecture: Design systems with declarative, composable extension points (facets), not isolated plugin panels
  2. Teams: Remove programmer gatekeeping in early invention stages
  3. Education: Rebalance curricula toward art-school-style design pedagogy with technical support
  4. Reading tech: Build open reading environments with facet-based composition
  5. Practitioners: Cultivate complementary capacities (programmers → design; designers → technical literacy)
  6. Prototyping: "Real enough" prototypes beat production-quality exploration

Why This Matters for PKM

Matuschak's essay is directly load-bearing for several PKM debates:

  • PKM is mostly research-purgatory-adjacent. Many of the most interesting PKM ideas — spatial canvases, rich transclusion, typed links, AI-augmented retrieval — exist in scattered tools, none of which compose. The research purgatory framing names a problem PKM has had for decades.
  • Obsidian is a positive example of composability. The plugin API gives users the same primitives as the core team. This is why Obsidian's ecosystem accumulates depth while Notion's does not — same function, different architecture.
  • Agentic vaults are the pattern continued. YB's agentic-vault work (constitution, Knowledge Work PRs, subroutines) is exactly the vibe-coding pattern applied to the vault itself — knowledge workers prototyping their own tools with coding agents.
  • Reading augmentation is a PKM primitive. Reading tools are the input side of the knowledge pipeline; their composability determines whether new reading ideas can ever reach daily practice.
  • The gatekeeping critique applies to PKM tool-building. Most novel PKM ideas come from programmer-practitioners because they can prototype. Matuschak's argument predicts that coding agents will let more non-programmer PKM practitioners prototype directly, diversifying the field.

Concepts and Entities

Concepts

  • App model / programming as accidental tyrannies
  • Composability; facets (CodeMirror-style declarative contribution)
  • Research purgatory
  • Vibe-coding
  • Physics of a medium
  • Real-enough prototypes
  • Gatekeeping dynamic (programmer-generalist as bottleneck)

Entities

  • Andy Matuschak (author; Quantum Country co-creator with Michael Nielsen; former Apple; Evergreen Notes)
  • Marijn Haverbeke — CodeMirror author
  • Nio Ono — designer cited on coding-agent experience
  • Steve Rubin — audio-text synchronization research
  • Bas Ording — iOS inertial scrolling prototyping anecdote
  • Research prototypes: Scim, CiteSee, Papeos, LiquidText, SpaceInk, Quantum Country
  • Historical: Engelbart (1962), Smalltalk, Webstrates, Potluck, Canon Cat, Ted Nelson / Xanadu