Roadmap

As of May 16, 2026, the latest tagged release is v0.1.10. The stable public surface remains tubebrain: a local MCP stdio server for structured YouTube VOD transcripts, metadata lookup, caption-language discovery, opt-in PoToken support, and opt-in local Whisper fallback.

The active product direction is TubeBrain: evolve the local-first MCP tool into a stream-capable, multi-surface product with a FOSS core and an optional hosted tubebrain.ai convenience layer.

Shipped

  • v0.1.10 published as the first TubeBrain-named release line with tubebrain-* assets, tubebrain.rb, and tubebrain-npm-package.tgz
  • v0.1.9 published with po-token compiled into the release binaries so default installs work against current YouTube BotGuard traffic
  • GitHub release tarballs for Apple Silicon macOS, Linux x86_64, Linux ARM64, checksum publishing, install.sh, Homebrew formula, and npm wrapper asset
  • authoritative CI, docs build, nightly fuzz, and release publication now run on self-hosted Tinyland lanes, with only the Pages deploy OIDC step left as a bounded hosted-by-policy exception
  • authenticated private-repo install support in install.sh and the npm wrapper bootstrap
  • public Pages mirroring of the latest release assets for anonymous installs
  • serialized honey Linux release jobs for gnu targets with CARGO_BUILD_JOBS=1 to avoid release runner OOM and artifact-upload flake
  • npm registry publication as tubebrain@beta on npmjs.com (SLSA provenance) with a GitHub Packages mirror as @jesssullivan/tubebrain
  • MCP tools for transcript extraction, language listing, and metadata
  • opt-in po-token and whisper feature paths
  • client-specific integration guides and compatibility matrix
  • quickstart walkthroughs for common user stories
  • sample outputs and troubleshooting pages
  • install lifecycle docs for upgrade, uninstall, and platform support
  • contributor and agent scaffolding with CONTRIBUTING.md, AGENTS.md, and llms.txt
  • agent-focused usage guidance for in-session MCP workflows
  • comparison and positioning docs
  • GitHub Pages docs deployment with generated llms.txt, robots.txt, and sitemap.xml
  • release history and public release documentation
  • launch-operation templates for release notes, announcement copy, demo flow, and post-launch triage
  • GitHub release-note categorization via .github/release.yml
  • issue templates for bug reports and feature requests

Landed On Main

  • TubeBrain type foundation: generalized media URIs, platforms, VOD/live media descriptors, and bridge compatibility for the existing YouTube VOD source
  • stream session contract: StreamTranscriber, SessionManager, polling cursors, bounded buffers, health reporting, and MCP lifecycle tools
  • YouTube Live resolver and HLS ingestion: watch-page live detection, HLS manifest parsing, byte fetch, n challenge solving, fMP4 init-map handling, byte-range normalization, MPEG-TS/ADTS decode, and Whisper live-window path
  • direct HTTP audio-stream ingestion with bounded chunking and MP3 frame-aligned flushes where possible
  • first hosted stream-session API slice in tubebrain-hosted: authenticated start, poll, minimal SSE event, stop, and list endpoints using the same StreamSession and StreamChunk shapes as local MCP
  • first internal STT service slice in tubebrain-stt: bearer-authenticated GET /v1/health and POST /v1/stt/chunk, remote STT client wiring in tubebrain-hosted, Nix app/image packaging, and protected-preview manifests for cluster-local hosted-to-STT traffic
  • Linear project checkpoint: TubeBrain is in progress, phases A/B/C are done, and live transcription hardening is tracked separately from the core implementation

Now

  • run the GTM MVP sprint tracked in Linear milestone GTM MVP: Live/Radio Stream API
  • prove the buyer-facing data flow: radio/HTTP audio or YouTube source -> stream session -> buffered structured transcript/context chunks -> agent polling or SSE
  • land the self-hosted STT path tracked by TIN-1243 and TIN-1244, then use TIN-1245 and TIN-1246 to move the protected hosted preview onto the Blahaj/Tinyland remote-STT shape
  • keep the GStack timestamp workflow as the deterministic marketing demo while using live/radio streams to prove the broader product wedge
  • define and document paid-pilot boundaries: API-key auth, usage dimensions, retention, no raw audio persistence by default, no public managed PoToken endpoint, and redacted diagnostics
  • use the paid-pilot operator runbook as the pre-call go/no-go checklist for protected-preview design partners
  • use the paid-pilot package as the quotable pricing and billing-evidence surface for the first design-partner pilots
  • package the paid-discovery offer through TIN-1009, the TubeBrain GTM sprint issues (TIN-1204 through TIN-1212), and the current execution issues (TIN-1240 and TIN-1242 through TIN-1248)

Next

  • execute the first paid-discovery outreach actions from TIN-1233 through TIN-1237
  • record and publish the hosted plus MCP demo for Bluesky, Mastodon, and a short FOSS adoption post using the GStack demo script
  • implement persistent usage events and quota enforcement from the hosted pilot policy before charging for multi-customer traffic
  • follow the documented pilot STT posture: self-hosted Whisper first, managed fallback only with buyer opt-in and an explicit fallback-hour cap
  • consolidate the duplicate public surfaces so tubebrain.ai is the SaaS/buyer surface and the MkDocs site remains the FOSS/operator truth surface

Later

  • broaden platform adapters: Twitch, Kick, microphone/CPAL, and more direct audio streams
  • add audio fingerprinting and a recognize_audio MCP surface
  • split the workspace into tubebrain-* crates once adapter weight justifies it
  • add billing, rate limiting, abuse controls, persistence, and operational policy for the hosted layer
  • revisit extra packaging channels such as nixpkgs and long-term crates.io posture after the rename/productization push

Planning And Operator Artifacts

Internal planning docs live under docs/plans/ and docs/spikes/. The paid pilot runbook is a public-safe operator checklist:

  • 2026-04-16-gtm-initiative.md
  • 2026-04-16-gtm-milestones.md
  • 2026-05-16-tubebrain-paid-discovery-one-pager.md
  • 2026-05-16-tubebrain-gstack-demo-script.md
  • docs/paid-pilot-package.md
  • docs/paid-pilot-runbook.md
  • docs/spikes/2026-05-16-managed-stt-fallback-cost-model.md
  • 2026-04-16-launch-checklist.md
  • 2026-04-29-gstack-research-demo.md