Roadmap

Project status for contributors — what's live and what's next.

Live Now

Shipped Features

Language reference (live)

140 languages with compare, transfer, and MIT JSON — the deepest, most reliable layer today.

Hand-curated signals

14 AI/dev tools and 7 stack records in JSON — small, manual catalogues you can review.

Preview registries

Models, agents, companies, and benchmarks as catalogue layers — preview snapshots with uneven depth; verify on detail pages.

Agent surface

  • /api/languages.json — MIT-licensed catalogue (static hosting)
  • GET /v1/transfer?concept=X&from=A&to=B — requires Node server; not on static-only hosts
  • GET /v1/languages/{id} — requires Node server; not on static-only hosts
  • For AI tools — static JSON + optional MCP

Advisor (experimental)

Client-side Decision Brief drafts for language/stack choices — useful orientation, not authoritative. Verify against official docs.

Next up

Reference Foundation Quality

Decision-grade coverage

  • Compare: only surface languages with enough data to support a real decision; weak entries get hidden or clearly marked.
  • Language detail: keep rendering graceful when optional fields are missing, but make gaps obvious so the recommendation layer stays honest.
Design only

Tools & Stacks Layers

API endpoints

  • GET /api/ai-tools.json — curated AI/dev tools that affect technical choices
  • GET /api/stacks.json — shipped: 7 opinionated low-bloat combinations

Human pages

Design only until built:

  • tools.html / tool-detail.html
  • stacks.html / stack-detail.html
  • Optional services.html
Live (experimental)

Advisor layer

Human experience

Describe what you are building, what constraint matters, and whether an LLM has already suggested something. Get ranked options with short reasons, tradeoffs, and links back to the catalogue.

Agent API

  • POST /api/recommend-language (planned) — grounded recommendations with catalogue_refs.id tied back to languages.json
  • Optional light analytics on the advisor only, so we learn which decision prompts are actually useful without bloating the core site
Documented

Process Layer

How new decision-support layers get added without turning CacheSphere into noise:

Discovery → Verification → Curation → Tradeoff judgment → Publish

CacheSphere is meant to be a grounded filter, not a mirror of the internet. We verify against official docs and small experiments, curate only what helps a real decision, then publish lean outputs humans and agents can actually use.

Guardrails

Scope & Guardrails

✓ In Scope Now

  • Decision-support surfaces: advisor, compare, transfer, and agent pages
  • Reference foundation behind those surfaces
  • Small hand-maintained ai-tools.json once it improves real choices

✗ Out of Scope

  • Automated scraper spam or noisy cron ingestion
  • Large new UIs before the decision layer is sharp
  • Stripe/Pro wiring ahead of product clarity
  • Marketing pipelines that add noise instead of proof

Parking Lot

Parked: devlog-from-Git, broader discovery, fuller stacks UI, and monetisation pushes until the core decision workflow feels tight.

Later

Not shipped yet

Draft services

SERVICES_DRAFT.md sketches packaged help for teams that want grounded stack choices without buying into bloated all-in-one bundles.

Not live until the core decision layer is stable. Contact: [email protected] · Pricing.

Example future stack

v1-languages-tools-stacks.md sketches a lean modular stack where the point is not novelty but choosing the extra tool or language only when the tradeoff is actually worth it.

Illustrative only — stack UI pages not built yet; catalogue data is in stacks.json.

Monetisation stance

Any paid layer should sit on top of the free core: setup help, reports, or decision support that saves time without hiding the catalogue.

No display ads. The core foundation stays free and tool-friendly. Pro or Teams ideas remain unshipped until the product earns them.

Source of Truth

Authoritative detail lives in the open-source repository under docs/ and product notes:

  • VISION.md — North star and framing
  • PRODUCT_VISION.md — Product direction
  • PROCESS_LAYER.md — How tools enter
  • v1-languages-tools-stacks.md — Stacks design
  • SERVICES_DRAFT.md — Future services

This page is updated when those documents change.