nendlabs

We build experimental software systems with new technology and proven engineering methods.

Packages

Typed command definitions that compile into bounded CLIs, traces, and generated agent skills.

Config-first project pages with typed sections, code tabs, MDX slots, and bundled styles.

Notion-connected tooling for research logs, docs, and structured workspace data.

Environment handling for small services that need explicit configuration boundaries.

Fast source-tree measurement for codebases, experiments, and build surfaces.

Bounded batch work should leave the main thread only when worker fanout, concurrency control, or keyed resume pays for the boundary.

A benchmark is useful only when the run itself is an artifact: planned, leased, resumable, and backed by per-case evidence.

Remote race engineering becomes more useful when the shared object is structured telemetry instead of an expiring screen share.

Model-written operational summaries only become useful after events are constrained, scoped, aggregated, and turned into bounded context.

Generated visual work becomes publishable only after seeded rendering, artifact receipts, curation, validation, and public inspection agree.

Computed object fields become useful when the runtime and the TypeScript result type agree on what is data and what is derived.

Cross-channel message delivery becomes accountable when provider-specific sends sit behind shared authentication, limits, formatting, and validation.

Small operational JSON state becomes safer when every record has an address, expiry behavior, scoped authority, and mutation evidence.

method

We prefer narrow prototypes with real interfaces over speculative architecture.

output

When a tool survives contact with our own work, it becomes a package, note, or public surface.

bias

Artifacts should expose what is active, what changed, and where the next edge is.

We work best where research, infrastructure, and product shape are still moving.