Safe code. Built for machine speed.

Build broadly, constrain later

General-purpose program dey start with ambient semantics. Sandboxes, IAM, containers, policy, and signing dey add around am to recover di intended boundary.

Grant small, then compile

Effects and capabilities na part of the admitted computation. If the target no fit prove and bind the grant, e no go emit or run the artifact.

Code as readable data

Immutable values, ordinary functions, explicit data, and composable syntax easy for humans and models to produce and inspect.

Talk wetin fit happen

Effects, capabilities, resources, dependencies, and targets na visible inputs to admission—not surprise wey dem discover after deployment.

Less language, harder boundary

No ambient interop, runtime code loading, unrestricted mutation, guest-defined macros, or unbounded concurrency for admitted component surface.

Declarative intent

Small, Clojure-shaped surface dey keep programs readable and e no allow ambient escape hatches.

Checked KIR

Types and transitive effects don turn to target-independent, inspectable representation.

Intersect authority

Requested, delegated, local-policy, resource, and target grants fit only narrow.

Address the artifact

Code, dependencies, policy, compiler contract, and target ABI bind di computation identity.

Bind for di host

Di runtime and provider bind only admitted capabilities, enforce finite budgets, and emit receipts.

Wetin be code?

The CID dey select hash-verified checked-KIR definition and e CID-only dependency closure.

E fit run here?

The exact interface, complete effect row, current allowance, fuel, and decreasing eval depth dem bind before execution.

Wetin come back?

Di typed result dey persist as content-addressed evidence. E hash no fit retroactively authorize effect.

Build faster. Run faster. Keep di boundary.

Kotoba dey publish compiler-startup, developer-loop, native-runtime, and workload-domain measurements wit exact-result checks. Current speed ranks still dey withheld until quiet-host gates pass; security admission no dey remove to win timing.

Storage wey no get language ceiling.

Kotobase dey use content identity, range reads, immutable history, and provider-neutral storage. Physical capacity, tenancy, retention, cost, replication, and execution budgets dey explicit; dis no be infinite-disk claim.

Post-quantum cryptography by default.

Every new Kotoba cryptographic boundary must name ML-KEM or ML-DSA evidence and reject classical-only downgrade. Existing Passkeys, transport, implementations, and key custody remain separately qualified boundaries.

Authentication dey. Authority denied by default.

Passkey identity gats dey for control boundary. Verified identity no get filesystem, network, storage, model, secret, payment, or GPU authority until explicit scoped grant pass local policy and host checks.

Flexible delegation wey fit only narrow.

Requested, delegated, local-policy, resource, and target scopes dey intersect. Delegation fit compose and reduce, but e no fit mint ambient authority or widen issuer grant.

Web3 ready, chain-neutral for root.

Stable Kotoba Principal and Passkey controller na di main ones. CAIP-10 accounts, ERC-1271, and ERC-6492 na explicit linked-account proofs; wallet address no go silently turn storage or execution authority.

Zero-copy where ownership allow; one copy where boundary require am.

Columnar byte views dey keep vector, direct ByteBuffer, and Uint8Array backing. Arrow projection fit keep uncompressed buffers columnar through the authorized Kotobase lake path. Network ingress, decompression, GPU upload, and immutable persistent updates still remain named copy boundaries.

Arrow-shaped data. Explicit CPU SIMD and device-native GPU kernels.

For Apple M4, uncompressed, non-nullable Arrow float32 column keep one WebAssembly linear-memory backing while Num run explicit v128 f32x4 kernel over borrowed values slice with zero Arrow-to-SIMD copies; scalar tail cover remaining rows. For three qualified runs of same 262,147-element scale workload and artifact, that SIMD kernel finish 3.66-3.72x faster than scalar Wasm. Na kernel-and-host result, no be general runtime claim. Same bounded column path still keep one ArrayBuffer through CPU views, cross GPU ownership boundary with one measured WebGPU upload, run for Metal, and return one four-byte scalar. Nullable columns, other Arrow dtypes, unified-memory upload removal, broader kernels, and universal CPU/GPU qualification still dey wait.

AI first. Agent-safe by default.

Kotoba na for programs wey AI agents and bots write or operate. Di stronger di model, di more important explicit effects, finite resources, capability confinement, receipts, and host enforcement dey.

AGI-ready boundaries, no be AGI claim.

Di architecture na to keep authority clear as models dey get more power. Kotoba no dey claim say AGI dey here, say generated programs dey trustworthy, or say confinement fit remove di compiler, runtime, provider, key-custody, and OS trusted computing base.

33 cores

Internal production dogfooding

Di wider Kotoba stack dey run 33 inference cores inside. Dis one prove say di team dey operate im own stack; e no be customer traction, paid adoption, or revenue.

8 claims

Boundaries dey machine-readable

Safety claims dey name their trusted computing base, negative evidence, and residual risk instead of just dey talk 'unhackable' slogan.

deny by default

No grant, no host effect

Empty policy no grant filesystem, network, process, clock, model, or secret authority. Providers must validate concrete resource scope.

4 toolchains, 21 runs each

Kotoba 40.998 ms · Rust 126.422 ms · C 146.324 ms · JVM 961.248 ms median.

21 rotating process-cold samples · load1 30.79 → 39.79 · required ≤ 1 · 2026-08-29 · Apple M4

6 workloads × 5 comparators

Amu native dey exercise against Rust, Clang / C11, Zig, Go c-shared, Swift through one common native call boundary.

30/30 comparator/workload pairs · exact answers verified

19 of 30 pairs

Amu native win 19 of the 30 comparator/workload pairs by at least 5%, separate from arms own spread. The bounded fastest claim need every pair, so e still dey unqualified — the count na the informative half.

At least 2 of those pairs no fit win at all. For narrow-arithmetic, amu, Apple clang -O3 and rustc -O3 dey compile the kernel to the same 61-instruction sequence — clang and rustc byte-identical, amu only different for register numbers. 5% margin over identical code no dey, so the bounded claim no fit happen, e no just unmet.

Median of 5 host-qualified runs; di score range 19–20 and 19 of di 30 pairs qualify for every one. One noisy sample fit disqualify plenty pairs at once, so one-run score no precise for one pair.

busy-CPU 0.090 → 0.069 → 0.076 · required ≤ 0.10 · 2026-09-07

11 toolchain paths

Dependency resolution, checking, clean and no-change builds, and process-cold first result dem record separately.

7 samples per measured stage · load1 20.84 → 24.49 · required ≤ 1

8 source sizes

Di same program from one function to 2048, build by every toolchain for host then run. Released binary get lowest cold-start cost here and e correct ceiling above 128 functions.

artifacts wey dem check after clock stop · load1 2.73 → 2.73 · required ≤ 1 · 2026-08-31

6 domains × 6 runtime paths

Strings, collections, allocation, file I/O, four-worker concurrency, and request-admission policy application kernel dey correctness checked.

7 samples for both process-cold and amortized lanes · load1 13.87 → 12.38 · rank withheld

Narrow arithmetic

Wide register pressure

Deep spill pressure

Call preservation

Branch + call control flow

Loop call back edge

String

Collection

Allocation

I/O

Concurrency

Real application

Admitted components no fit handle runtime/native memory and component memory operations gats dey bounded or trap.

Trusted computing base

bounded reader · frontend admission · artifact verifier · Wasm/native runtime

Residual risk

• runtime-engine vulnerabilities still dey inside di TCB
• native loaders need second OS isolation boundary

Every transitive component effect dey declared and admitted before emission, including effects wey Kotoba-written providers dey use.

Trusted computing base

effect inference · capability catalog · frontend call graph

Residual risk

• kotoba and compiler grammar/effect parity gats dey compared steady steady

Ungrant capability no dey or e no bind, e no fit reach provider or native handler.

Trusted computing base

policy intersection · compiler import emission · tender import binding · host guard

Residual risk

• provider and native implementations must independently validate resource scope
• production effective grants must forbid wildcard scope

The same admitted source, target, policy and lock go produce the same observable pure result and artifact bytes.

Trusted computing base

canonical reader · deterministic lowering · pinned toolchain

Residual risk

• host effects na deterministic only where their capability contract talk am so

Source, admission, execution, memory and output dey use explicit finite bounds.

Trusted computing base

admission limits · fuel meter · runtime quota · supervisor timeout

Residual risk

• platform supervisors never get equal production isolation evidence yet

Release admission dey bind artifact identity, trusted signer, validity and reproducible evidence.

Trusted computing base

signature verifier · trusted signer configuration · clock · revocation set

Residual risk

• key custody and external revocation distribution still dey operational TCB

One shared portable component get equal acceptance, result and effect trace across qualified backends.

Trusted computing base

shared conformance manifest · backend adapters · comparison runner

Residual risk

• compiler-only features no dey portable and dem must reject by portable profiles

Component import reach im provider or native handler only with concrete post-intersection resource scope and e emit receipt.

Trusted computing base

capability intersection · host guard · provider handler · receipt sink

Residual risk

• provider-specific path, redirect, symlink and tenant checks need Q5 kits

Profile 6

package contract 1

v0.7.0

profile binding: verified

RELEASED

:docs/release-bound-profile

Install and self-check

brew tap kotoba-lang/kotoba
brew trust kotoba-lang/kotoba
brew install kotoba
kotoba selfhost check --json

Accept valid response wey get empty problem list.

First program

(defn main []
  (+ 40 2))

Dis program no request host imports; di emitted module no get imports.

Docs by intent

Start with installation, learn the admitted language, or inspect the normative semantics and conformance data.

One source, one answer

The example wey dey below na the exact source wey dem compile into the browser demo—not be JavaScript reimplementation.

Run for dis page

Load same-origin, digest-bound WebAssembly artifact and call its exported Kotoba main function.

Libraries and contracts

Browse bounded core names, foundational libraries, package rules, and dia current maturity boundary.

;; W1 pure representative: ordinary Clojure-shaped values/functions only.
(ns examples.w1-pure)

(defn double [n]
  (+ n n))

(defn main []
  (double 21))

Highlighting authority: kotoba-lang/grammar → kotoba.grammar.highlight/tokenize → build-time HTML. Editor scope contract: source.kotoba. Browser highlighter dependency: none. Inspect dependencies

Compile locally: kotoba compile double-21.kotoba --target wasm32-browser --output double-21.wasm

Generated symbol reference

Search di names wey di current bounded standard-library contract admit.

Data, effects, I/O, tooling

Start with coll, spec, json, text, wit, async, time, fs, http, test, fmt, lint, and LSP contracts.

Content-addressed dependencies

Inspect exact dependency CIDs, identity layers, GitHub provenance, and di current publication boundary.

One versioned contract

Keep grammar, effects, checked KIR, target adapters, qualification, and first-run documentation aligned.

Close provider gaps

Expand typed request/result conformance, adversarial testing, receipts, revocation, and reproducible release operations.

Earn wider deployment

Make production use big after provider, host-isolation, rollback, and soak evidence—plus grow inspectable declarative libraries.

Language issues

Ask design question, propose documentation improvement, or report reproducible language-contract problem.

Compiler and CLI issues

Follow implementation work, releases, target support, and runtime integration for the installable implementation.

Report privately

Use di published security policy for vulnerabilities; no make you talk exploitable details for public issue.

GitHub Sponsors

No sponsorship payment fit happen through kotoba-lang.org while di GitHub profile no dey live.

See sponsorship status and principles

Support no be authority

Sponsorship no dey buy feature, roadmap priority, support SLA, private access, or security exception.

Evidence before slogans

Read short engineering notes wey connect product claims to measurements, authority files, and remaining gates.

Controlled execution

Kotoba Cloud dey connect identity and deploy control to the execution environment. Discovery dey live; hosted apply no dey offered yet. Compute still dey provided by separately governed services.

Trusted graph state

Kotobase na di content-addressed graph database for AI state and knowledge: explicit relationships, identifiable history, and scoped access.

Compute and inference plane

Fleet compute and model-serving infrastructure. Availability and route qualification still service-specific.

Agent work plane

Continue agent work across workspaces, goals, evidence, tools, approvals, and governed effects.

kotoba-lang/kotoba-lang

Grammar, semantics, capability contracts, safety claims, CLI contract, documentation, and conformance fixtures.

kotoba-lang/kotoba

CLI, host integrations, providers, runtime adapters, integration tests, and target-specific qualification evidence.

Learn, build, or evaluate

Separate paths for first use, language reference, backend implementation, security boundaries, and maturity evidence.