How JobCtrl Compares

This page compares JobCtrl's approach to common job-search-tool capabilities against other approaches, one neutral capability category at a time.

Scaffold — alternative columns are not yet published

This is the comparison structure and rules only. It deliberately names no external product or company. The "Alternative approach" columns are placeholders for the maintainer to fill, privately and one at a time, only after verifying each fact against that approach's own public documentation (see How this page is maintained). Until a cell is verified it stays a TODO(owner) placeholder rather than a speculative claim.

The JobCtrl column states only shipped, Current behavior; each cell cites the claim-ledger IDs (CL-0xx) that back it. Those IDs are governed in the repository-only claims ledger (docs/claims-ledger.md), and each maps to a user-facing page linked from the capability category on the left.

Capability comparison

Capability category	JobCtrl	Alternative approach A	Alternative approach B
Data ownership — where your profile, jobs, and materials live	Local-first: one SQLite database and generated files under your home directory; no hosted backend and no account (CL-060, CL-061). Data at rest is not encrypted — protection is your OS account (CL-066). See Data, Privacy & Safety.	TODO(owner)	TODO(owner)
Outbound data — what leaves your machine, and when	Nothing leaves by default; each egress path (LLM, apply-agent prompt, job boards, read-only Gmail, Maps, CAPTCHA, telemetry) is opt-in and configuration-gated (CL-061, CL-062, CL-065). See Security → What Leaves Your Machine.	TODO(owner)	TODO(owner)
Scoring transparency — how fit is decided and explained	A deterministic, versioned policy scores fit 1–10 from structured evidence, with a per-requirement ledger explaining each score; scoring is an applicant-side triage aid, not a hiring decision (CL-010, CL-011, CL-012). See Scoring.	TODO(owner)	TODO(owner)
Materials integrity — how tailored resumes stay truthful	Every bullet traces to its source evidence; deterministic fabrication gates block invented facts and fail closed; keyword coverage is computed against the actual rendered resume text (CL-020, CL-021, CL-022). See Tailoring Contract.	TODO(owner)	TODO(owner)
Apply supervision — what stands between a job and a submission	Live submission requires an explicit approval bound to the reviewed materials, profile, and URL plus matching dry-run evidence, enforced in the worker claim transaction; dry runs submit nothing (browser-level guard); no application is submitted twice (CL-030, CL-031, CL-033, CL-034). See Security → Approval And Control Gates.	TODO(owner)	TODO(owner)
Third-party boundaries — what the tool refuses to do	Never submits, runs destructive actions, or bypasses CAPTCHA / paywall / login / rate-limit / bot controls without explicit authorization; the apply agent stops on SSO, declines permission prompts, and never enters payment details (CL-035). See Security → No Third-Party Bypass.	TODO(owner)	TODO(owner)
Reliability — how long-running work survives crashes	Every stage runs as a durable Temporal workflow with heartbeats and classified retries; the Runs page shows every run with a deep link into the workflow engine (CL-050, CL-051). See Runtime Boundaries.	TODO(owner)	TODO(owner)
Cost control — how automated LLM spend is bounded	A configurable daily spend ceiling runs a budget preflight before each spendful workflow and stops it at the ceiling; today's estimated spend is visible on the health surface (CL-040, CL-041). It is an estimate, not the provider bill. See Configuration → LLM spend.	TODO(owner)	TODO(owner)

How this page is maintained

This comparison rots unless it is re-verified. Two rules keep it honest:

• Facts-verified-before-publish. No comparative cell about an external approach is published without (a) a recorded verification date and (b) a dated link to that approach's own public documentation (README, docs site, or official page). No cell asserts a claim from memory, marketing copy, or a third-party summary. Every JobCtrl-column cell cites a frozen Current claim-ledger row. Any alternative cell that is not yet verified stays a TODO(owner) placeholder (or an HTML-commented draft) — never a speculative claim.
• No disparagement. Alternative entries are neutral, factual descriptions of a published approach, with the source and check date recorded. No unverifiable or comparative-superlative claims.

Maintenance cadence. Re-verify every JobCtrl-column cell against the claims ledger, and every published alternative cell against its cited source, on each release and at minimum quarterly. TODO(owner): confirm the interval and the sidebar label/placement (plan §11.3). Record the last full re-verification date here at publish time: last verified: TODO(owner).

When you add or change an alternative column, record its entries in this format inside the cell so the check date travels with the claim:

<neutral factual statement> (source: <url>, verified <YYYY-MM-DD>)