serde-rs/serde
Dimensions
Interface surface 4 exemplary An API reference is generated from the code (autodoc, mkdocstrings, typedoc, …).
Can another program use it through a documented, machine-readable surface?
library ladder · level 4
- ✓ L1 An importable, packaged API exists. serde/src/lib.rs Rust library crateserde_core/src/lib.rs Rust library crate
- ✓ L2 The public API is delimited (__all__, an exports map, exported identifiers). serde/src/lib.rs pub itemsserde_core/src/lib.rs pub items
- ✓ L3 The public API is statically typed (py.typed, type declarations, a typed language). serde/src/lib.rs statically typed languageserde_core/src/lib.rs statically typed language
- ✓ L4 An API reference is generated from the code (autodoc, mkdocstrings, typedoc, …). serde/Cargo.toml 1 matchserde_core/Cargo.toml 1 match
Contract stability 1 ad hoc Releases are versioned (semver tags or versioned changelog sections).
Can integrators rely on it not changing under them?
- next: L2 A changelog exists and marks breaking changes. (no changelog)
all ladder · level 1
- ✓ L1 Releases are versioned (semver tags or versioned changelog sections). 323 semver tags, latest v1.0.229
- ✗ L2 A changelog exists and marks breaking changes. no changelog
- ✓ L3 A deprecation mechanism is used in code (warnings, @deprecated, Deprecation/Sunset headers) or the API is versioned. serde/src/integer128.rs 1 match
- ✓ L4 Across tagged releases, no OpenAPI operation, CLI option or __all__ symbol was removed without first being deprecated. 20 tagged releases compared
Events n/a Not applicable: no service and no long-running CLI mode (batch CLIs, libraries and desktop apps have no events).
Can others react to state changes without polling?
Identity n/a Not applicable: uses no credentials and has no service surface.
Can a machine act with its own revocable, least-privilege identity?
Data portability n/a Not applicable: owns no persistent data.
Can its data get in and out, in bulk, in open formats, with an explicit schema?
Agent ergonomics 1 ad hoc Errors are specific exception types, and library code never calls sys.exit.
Can an automated caller use it without guessing, and recover from failure?
- next: L2 Inputs are validated with messages naming the bad argument. (no input validation errors)
library ladder · level 1
- ✓ L1 Errors are specific exception types, and library code never calls sys.exit. serde_derive/src/internals/case.rs 1 match no sys.exit in library code
- ✗ L2 Inputs are validated with messages naming the bad argument. no input validation errors
- ✗ L3 Timeouts and retries are configurable. no timeout parameter
- ✗ L4 Agent docs, and examples that run as tests (doctests). no agent docs
Embeddability 2 partial Installable as a package (a package manifest).
Can another system run and compose it without a human?
- next: L3 Published to a public registry, with at most 10 runtime dependencies. (not published to a registry)
library ladder · level 2
- ✓ L1 Importable from source. serde/src/lib.rs Rust library crateserde_core/src/lib.rs Rust library crate
- ✓ L2 Installable as a package (a package manifest). serde/src/lib.rs package manifestserde_core/src/lib.rs package manifest
- ✗ L3 Published to a public registry, with at most 10 runtime dependencies. not published to a registry
- ✓ L4 Optional features behind extras/features. serde/Cargo.toml 1 matchserde_core/Cargo.toml 1 match
Extensibility 0 absent Not yet: Internal seams: a registry, registration function or plugin manager.
Can behaviour be added without modifying the code?
- next: L1 Internal seams: a registry, registration function or plugin manager. (no registry or registration function)
all ladder · level 0
- ✗ L1 Internal seams: a registry, registration function or plugin manager. no registry or registration function
- ✗ L2 The docs show how to add behaviour through those seams (even in-tree). no registration functions to document
- ✗ L3 Out-of-tree extensions are discovered (entry points, a plugin directory, configuration) without editing the package. extensions must be added inside the package
- ✗ L4 The extension API is versioned, has lifecycle hooks, and has example extensions. the extension API isn't versioned
Observability 0 absent Not yet: Logs through a standard logger namespace and never prints from library code.
Can an operator, or an orchestrating agent, tell whether it works and why not?
- next: L1 Logs through a standard logger namespace and never prints from library code. (no standard logger)
library ladder · level 0
- ✗ L1 Logs through a standard logger namespace and never prints from library code. no standard logger
- ✗ L2 Log levels are used meaningfully (at least two levels). fewer than two log levels used
- ✗ L3 Optional metrics or tracing hooks. no metrics or tracing hooks
- ✗ L4 OpenTelemetry instrumentation available. no OpenTelemetry instrumentation
Change legibility 2 partial CI runs the tests; applications pin dependencies (libraries and zero-dependency projects are exempt).
Can an agent change it safely with limited context?
- next: L3 At least 90% of modules fit 32k tokens and no file exceeds 12k; test/source token ratio at least 0.3; at least 70% of the code is typed. (86% of modules fit 32k tokens; 6 files over 12k)
all ladder · level 2
- ✓ L1 A README and some tests. README.mdtest_suite/tests/bytes/mod.rs 150 test files
- ✓ L2 CI runs the tests; applications pin dependencies (libraries and zero-dependency projects are exempt). .github/workflows/ci.yml library or zero-dependency project: lockfile not required
- ✗ L3 At least 90% of modules fit 32k tokens and no file exceeds 12k; test/source token ratio at least 0.3; at least 70% of the code is typed. 86% of modules fit 32k tokens; 6 files over 12k
- ✗ L4 No import cycles, and agent docs (AGENTS.md, CLAUDE.md, llms.txt). no agent docs (AGENTS.md, CLAUDE.md, llms.txt)
Make or buy
Make with agents
950,483 output + 23,762,075 input tokens
on anthropic:claude-opus-5-5 ≈ $37.
10.6 agent-hours · 141.8 human days (9.5 steering + 132.3 rediscovering history).
$110,275 over 3 years
Buy (adopt as is)
15 min to running (via library) + 2 min to first use (via library).
240 min/year of upgrades; price $0/year.
$823 over 3 years
Buying wins below $36,484 per year. How this is estimated.
Classic COCOMO vs neosloc value
Classic COCOMO (sloccount)
18.26 KSLOC → 50.7 person-months over 11.1 months.
$570k
The effort to write it by hand, as COCOMO priced it in 1981.
neosloc value (neoCOCOMO)
Reproduce with agents: 10.8 PM (×0.21 of classic; 75% of behaviour captured by tests, contract and docs).
Rediscover uncaptured history: 6.3 PM (375 fixes in 3542 commits by 216 authors).
× leverage 0.93 (integrability) × obsolescence 0.95.
$170k (15.1 PM)
What survives
30% of the classic figure
37% of the replacement cost lives only in code and history.
History
| Evaluated | Commit | neosloc | Index | Make or buy |
|---|---|---|---|---|
| 2026-10-07 | 6693a89cca77 | 0.8.0 | 1.43 | buy |
| 2026-10-07 | 6693a89cca77 | 0.7.4 | 0.33 | buy |
| 2026-10-07 | 6693a89cca77 | 0.7.3 | 0.33 | buy |
| 2026-10-07 | 6693a89cca77 | 0.7.2 | 0.33 | buy |