From 83315de52cba9482df8578d0a3a1d70eae65ad42 Mon Sep 17 00:00:00 2001 From: pyr0ball Date: Wed, 25 Feb 2026 11:40:55 -0800 Subject: [PATCH] =?UTF-8?q?docs:=20backlog=20=E2=80=94=20Ultra=20tier=20ma?= =?UTF-8?q?naged=20applications=20concept?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- docs/backlog.md | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/docs/backlog.md b/docs/backlog.md index bb13018..04b57a5 100644 --- a/docs/backlog.md +++ b/docs/backlog.md @@ -22,6 +22,18 @@ Unscheduled ideas and deferred features. Roughly grouped by area. --- +## Ultra Tier — Managed Applications (White-Glove Service) + +- **Concept** — A human-in-the-loop concierge tier where a trained operator submits applications on the user's behalf, powered by AI-generated artifacts (cover letter, company research, survey responses). AI handles ~80% of the work; operator handles form submission, CAPTCHAs, and complex custom questions. +- **Pricing model** — Per-application or bundle pricing rather than flat "X apps/month" — application complexity varies too much for flat pricing to be sustainable. +- **Operator interface** — Thin admin UI (separate from user-facing app) that reads from the same `staging.db`: shows candidate profile, job listing, generated cover letter, company brief, and a "Mark submitted" button. New job status `queued_for_operator` to represent the handoff. +- **Key unlock** — Browser autofill extension (above) becomes the operator's primary tool; pre-fills forms from profile + cover letter, operator reviews and submits. +- **Tier addition** — Add `"ultra"` to `TIERS` in `app/wizard/tiers.py`; gate `"managed_applications"` feature. The existing tier system is designed to accommodate this cleanly. +- **Quality / trust** — Each submission requires explicit per-job user approval before operator acts. Full audit trail (who submitted, when, what was sent). Clear ToS around representation. +- **Bootstrap strategy** — Waitlist + small trusted operator team initially to validate workflow before scaling or automating further. Don't build operator tooling until the manual flow is proven. + +--- + ## Container Runtime - **Podman support** — Update `Makefile` to auto-detect `docker compose` vs `podman-compose` (e.g. `COMPOSE ?= $(shell command -v docker 2>/dev/null && echo "docker compose" || echo "podman-compose")`). Note in README that rootless Podman requires CDI GPU device spec (`nvidia.com/gpu=all`) instead of `runtime: nvidia` in `compose.yml`.