শব্দ

Shobdo master plan. A programming language written entirely in Bangla, with its own native compiler and frameworks for web, macOS, Windows, Linux, Android, iOS and machine learning. This page lists every workstream with a detailed checklist, a 36-month timeline and measured progress.

How to read this page. A status of Done means the work exists and was run or checked (for example, tests pass or the binary runs). Partial means the work is written but not fully run or verified. Durations, start months and team sizes are planning estimates, not commitments. Click any status pill to update it; changes are saved only in this browser, and Reset restores the verified baseline. Baseline recorded on 10 Nov 2026.
0%overall
DonePartialNot started

Overall progress = (Done + ½ × Partial) ÷ all checklist items, every item weighted equally. The checklist sizes the work; it does not predict effort per item.

Timeline Nov 2026 → Oct 2029 · 36 months · bar fill = progress

Milestones release gates

Detailed checklists 15 workstreams

Team & effort planning estimate

WorkstreamWindowMonthsEst. people (FTE)Key skills

Platform requirements things code alone cannot solve

TargetExternal requirementPlan
macOS (arm64)Binaries must be code-signed to run on Apple silicon. Distributing outside the App Store without warnings needs a Developer ID and notarization.Ad-hoc signing is done in our own code. Developer ID and notarization come at the release stage.
iOSInstalling on devices and publishing need an Apple Developer Program membership, provisioning profiles and Apple signing.Generate an AOT native arm64 binary plus app bundle; sign with the team's Apple identity.
AndroidAPKs/AABs must be signed. Apps run on ART (DEX bytecode); native code is loaded as ELF shared libraries through JNI.Native ELF .so plus a small launcher, then our own DEX and APK writers.
WindowsUnsigned executables trigger SmartScreen warnings; trusted signing needs a code-signing certificate.Own PE/COFF writer; certificate procured before public release.
WebBrowsers run WebAssembly but access the DOM through JavaScript.Own Wasm writer plus a minimal generated JS bridge.
ML on GPUGPU compute uses vendor APIs (Metal, CUDA, Vulkan, Direct3D).Decided: the first version is CPU only. GPU backends (Metal, then Vulkan/CUDA) come after v1.0.

Top risks

RiskImpactMitigation
Scope: 6 platforms + ML is very largeHighStrict gates: no new platform until the previous gate passes; shared IR and shared UI toolkit.
Bangla text shaping (conjuncts, reph, vowel signs) in custom UIHighDedicated shaping workstream with a test corpus checked by native readers.
Self-hosting bugs that are hard to debugMediumDifferential testing against the interpreter; keep the stage-0 path working.
Platform policy changes (stores, signing)MediumIsolate packaging and signing per platform; track platform release notes.
Adoption: keyword acceptanceMediumVolunteer keyword panel, public RFCs, Latin aliases.
License and name undecidedMediumDecide before accepting outside contributions.