diff --git a/SCRATCHPAD.md b/SCRATCHPAD.md
index 22431c8..cd16568 100644
--- a/SCRATCHPAD.md
+++ b/SCRATCHPAD.md
@@ -1,62 +1,209 @@
-<!--
-
 # Scratchpad
 
-Temporary information holder for the next iteration. Developers may create or
-delete this file at any time. Anything durable should move into code, tests, or
-canonical docs, then this file should disappear.
-
-
 ## Goal
 
-Use this section for the current slice of work. It should explain what the
-scratchpad is helping us move forward right now.
+- Capture a compact, project-wide list of optimization candidates after a broad scan of the current FFX codebase, tooling, and requirements.
 
 ## Settled
 
-Use this for decisions that are stable enough to guide the next steps, but are
-still temporary enough to live in the scratchpad for now.
+- The biggest near-term wins are in startup cost, repeated subprocess work, repeated database query patterns, and general repo hygiene.
+- This list is intentionally optimization-oriented rather than bug-oriented. Some items below also improve correctness or maintainability, but they were selected because they can reduce runtime cost, operator friction, or iteration overhead.
 
 ## Focused Snapshot
 
-Use an extra section like this only when one slice needs its own compact
-summary. This is useful when a specific API, boundary, or model was recently
-recreated and should be captured clearly.
+- Highest-leverage application optimizations:
+  - Lazy-load CLI command dependencies so lightweight commands do not import most of the app.
+  - Collapse repeated `ffprobe` calls into a single probe result per source file.
+  - Replace `query.count()` plus `first()` patterns with single-query ORM accessors.
+  - Cache or precompile filename pattern regexes instead of scanning every pattern for every file.
+  - Guard logger handler installation to avoid duplicated handlers and noisy repeated setup.
 
+- Highest-leverage repo and workflow optimizations:
+  - Stop tracking nested `__pycache__` output and other generated artifacts.
+  - Consolidate setup and upgrade tooling to reduce overlapping shell-script responsibilities.
+  - Trim or reorganize the oversized test/combinator surface so it is easier to run, debug, and extend.
+
+## Optimization Candidates
+
+1. CLI startup and import cost
+- [`src/ffx/ffx.py`](/home/osgw/.local/src/codex/ffx/src/ffx/ffx.py) imports a large portion of the application at module import time, even for cheap commands such as `version`, `help`, `setup_dependencies`, and `upgrade`.
+- Optimization:
+  - Move heavy imports into the commands that actually need them.
+  - Keep the CLI root importable with only core stdlib and Click dependencies.
+- Expected value:
+  - Faster startup for scripting and tooling commands.
+  - Less coupling between maintenance commands and the runtime stack.
+
+2. Repeated database queries via `count()` plus `first()`
+- Controllers such as [`src/ffx/show_controller.py`](/home/osgw/.local/src/codex/ffx/src/ffx/show_controller.py), [`src/ffx/pattern_controller.py`](/home/osgw/.local/src/codex/ffx/src/ffx/pattern_controller.py), and [`src/ffx/database.py`](/home/osgw/.local/src/codex/ffx/src/ffx/database.py) often do `q.count()` and then `q.first()`.
+- Optimization:
+  - Replace with `first()`, `one_or_none()`, or existence checks that do not issue two queries.
+  - Standardize this across all controllers.
+- Expected value:
+  - Lower SQLite query volume.
+  - Simpler controller code.
+
+3. Filename pattern matching scales linearly across all patterns
+- [`src/ffx/pattern_controller.py`](/home/osgw/.local/src/codex/ffx/src/ffx/pattern_controller.py) loads every pattern and runs `re.search` against each filename on every lookup.
+- Optimization:
+  - Cache compiled regexes in process memory.
+  - Stop after the first intentional match instead of silently returning the last match.
+  - Consider explicit pattern priority if overlapping rules are valid.
+- Expected value:
+  - Faster per-file setup when many patterns exist.
+  - More predictable matching behavior.
+
+4. Media probing does two separate `ffprobe` subprocesses per file
+- [`src/ffx/file_properties.py`](/home/osgw/.local/src/codex/ffx/src/ffx/file_properties.py) calls `ffprobe` once for format data and once for stream data.
+- Optimization:
+  - Use one probe call that requests both format and streams.
+  - Cache that result inside `FileProperties`.
+- Expected value:
+  - Less subprocess overhead.
+  - Faster inspect and convert flows.
+
+5. Crop detection is always a full extra ffmpeg scan
+- [`src/ffx/file_properties.py`](/home/osgw/.local/src/codex/ffx/src/ffx/file_properties.py) runs a dedicated `ffmpeg -vf cropdetect` pass for each file when crop detection is requested.
+- Optimization:
+  - Cache crop results for repeated runs on the same source.
+  - Consider exposing shorter sampling windows or probe presets for large files.
+- Expected value:
+  - Lower latency on repeated experimentation.
+
+6. Process wrapper lacks stronger execution controls
+- [`src/ffx/process.py`](/home/osgw/.local/src/codex/ffx/src/ffx/process.py) uses `Popen(...).communicate()` without timeout handling, structured error mapping, or direct missing-command handling.
+- Optimization:
+  - Add timeout support and clearer `FileNotFoundError` handling.
+  - Consider `subprocess.run(..., check=False, text=True)` where streaming is not required.
+  - Centralize return/error formatting.
+- Expected value:
+  - Better failure diagnosis.
+  - Cleaner process management semantics.
+
+7. Logger handlers can be added repeatedly
+- [`src/ffx/ffx.py`](/home/osgw/.local/src/codex/ffx/src/ffx/ffx.py) adds file and console handlers each invocation.
+- Several helper classes install `NullHandler` instances ad hoc, for example [`src/ffx/process.py`](/home/osgw/.local/src/codex/ffx/src/ffx/process.py), [`src/ffx/tmdb_controller.py`](/home/osgw/.local/src/codex/ffx/src/ffx/tmdb_controller.py), [`src/ffx/media_descriptor.py`](/home/osgw/.local/src/codex/ffx/src/ffx/media_descriptor.py), and [`src/ffx/helper.py`](/home/osgw/.local/src/codex/ffx/src/ffx/helper.py).
+- Optimization:
+  - Guard handler installation so each logger is configured once.
+  - Prefer module-level logger setup patterns over per-instance handler mutation.
+- Expected value:
+  - Less duplicate logging.
+  - Lower confusion in long-running or repeatedly invoked contexts.
+
+8. Repo-local hygiene for generated Python artifacts
+- The repo currently contains nested compiled artifacts under `src/ffx/__pycache__/...`.
+- `.gitignore` only ignores `__pycache__` at the repo root, not recursive `__pycache__/`.
+- Optimization:
+  - Ignore `__pycache__/` recursively and clean tracked generated files.
+  - Consider ignoring local virtualenv or other generated tool directories if they may appear in-repo later.
+- Expected value:
+  - Cleaner diffs and scans.
+  - Lower repo noise.
+
+9. Tooling overlap and naming drift
+- There are now multiple prep-related scripts: [`tools/prepare.sh`](/home/osgw/.local/src/codex/ffx/tools/prepare.sh), [`tools/setup.sh`](/home/osgw/.local/src/codex/ffx/tools/setup.sh), and the legacy-like [`tools/ffx_update.sh`](/home/osgw/.local/src/codex/ffx/tools/ffx_update.sh).
+- Optimization:
+  - Decide which scripts remain canonical.
+  - Replace or remove legacy wrappers once equivalent CLI commands exist.
+  - Keep CLI maintenance commands and shell wrappers aligned.
+- Expected value:
+  - Less operator confusion.
+  - Fewer duplicated procedures to maintain.
+
+10. Placeholder UI surfaces should either ship or disappear
+- [`src/ffx/help_screen.py`](/home/osgw/.local/src/codex/ffx/src/ffx/help_screen.py) and [`src/ffx/settings_screen.py`](/home/osgw/.local/src/codex/ffx/src/ffx/settings_screen.py) are placeholders.
+- Optimization:
+  - Either remove them from the active UI surface or complete them.
+  - Avoid paying ongoing maintenance cost for unfinished navigation targets.
+- Expected value:
+  - Leaner interface.
+  - Lower UX ambiguity.
+
+11. Large Textual screens repeat configuration and controller loading
+- Screens such as [`src/ffx/media_details_screen.py`](/home/osgw/.local/src/codex/ffx/src/ffx/media_details_screen.py), [`src/ffx/pattern_details_screen.py`](/home/osgw/.local/src/codex/ffx/src/ffx/pattern_details_screen.py), and [`src/ffx/show_details_screen.py`](/home/osgw/.local/src/codex/ffx/src/ffx/show_details_screen.py) repeat setup patterns and local metadata filtering extraction.
+- Optimization:
+  - Extract a shared screen base or helper for common config/controller/bootstrap logic.
+  - Reduce repeated table refresh and repeated DB fetch code where possible.
+- Expected value:
+  - Lower maintenance overhead.
+  - Easier UI iteration.
+
+12. Several helper functions are unfinished or dead-weight
+- [`src/ffx/helper.py`](/home/osgw/.local/src/codex/ffx/src/ffx/helper.py) contains `permutateList(...): pass`.
+- There are many combinator and conversion placeholders across tests and migrations.
+- Optimization:
+  - Remove dead code, finish it, or isolate it behind a clearly dormant area.
+  - Avoid carrying stubbed utility surface that looks reusable but is not.
+- Expected value:
+  - Smaller mental model.
+  - Less time spent re-evaluating inactive paths.
+
+13. Test suite shape is expensive to understand and likely expensive to run
+- The project has a large matrix of combinator files under [`src/ffx/test`](/home/osgw/.local/src/codex/ffx/src/ffx/test), several placeholder `pass` implementations, and at least one suspicious filename with an embedded space: [`src/ffx/test/disposition_combinator_2_3 .py`](/home/osgw/.local/src/codex/ffx/src/ffx/test/disposition_combinator_2_3 .py).
+- Optimization:
+  - Consolidate combinator families.
+  - Add a lighter smoke-test path.
+  - Normalize file naming and test discovery conventions.
+- Expected value:
+  - Faster contributor onboarding.
+  - Easier CI adoption later.
+
+14. Process resource limiting semantics could be clearer
+- [`src/ffx/process.py`](/home/osgw/.local/src/codex/ffx/src/ffx/process.py) prepends `nice` and `cpulimit` directly when values are set.
+- Optimization:
+  - Validate and document effective behavior for combined `nice` + `cpulimit`.
+  - Consider explicit no-limit vs configured-limit states in the CLI and requirements.
+- Expected value:
+  - Fewer surprises in production-like runs.
+  - Easier support for user-reported performance behavior.
+
+15. Import-time dependency coupling makes maintenance commands brittle
+- Even after recent CLI maintenance additions, the top-level CLI module still imports most application modules before Click dispatch.
+- Optimization:
+  - Push imports for ORM, Textual, TMDB, ffmpeg helpers, and descriptors behind the commands that actually need them.
+- Expected value:
+  - Maintenance commands such as setup and upgrade stay usable when optional runtime dependencies are broken.
+  - Better separation between media runtime code and maintenance tooling.
+
+16. Regex and string utility cleanup
+- [`src/ffx/helper.py`](/home/osgw/.local/src/codex/ffx/src/ffx/helper.py) still emits a `SyntaxWarning` for `RICH_COLOR_PATTERN`.
+- Optimization:
+  - Convert regex literals to raw strings where appropriate.
+  - Review filename and TMDB substitution helpers for repeated string churn.
+- Expected value:
+  - Cleaner runtime output.
+  - Less warning noise during dry-run maintenance commands.
+
+17. Database startup always runs schema creation and version checks
+- [`src/ffx/database.py`](/home/osgw/.local/src/codex/ffx/src/ffx/database.py) runs `Base.metadata.create_all(...)` and version checks every time a DB-backed context is created.
+- Optimization:
+  - Measure startup cost and consider separating bootstrapping from ordinary command execution.
+  - Keep schema migration/version enforcement explicit.
+- Expected value:
+  - Faster command startup.
+  - Clearer operational boundaries.
 
 ## Open
 
-Use this for unresolved questions, design choices, and risks that still need a
-decision.
-
-## Sketches
-
-Use this for rough candidate structures, names, or shapes. Keep it explicit
-that these are sketches, not committed architecture.
-
+- Should optimization work focus first on operator-perceived latency, internal maintainability, or correctness-risk cleanup that also has performance upside?
+- Is the long-term supported model still “local Linux workstation plus Textual UI,” or should optimization decisions bias toward a more scriptable/headless CLI?
 
 ## Gaps Right Now
 
-Use this for concrete missing pieces in the current repo state. This section
-should describe what is absent or incomplete, not broad future ambitions.
+- No explicit prioritization owner or milestone for the optimization backlog.
+- No benchmark or timing harness exists for startup, probe, DB, or conversion orchestration overhead.
+- Repo hygiene is still mixed with generated artifacts and some clearly unfinished files.
 
 ## Next
 
-Use this for the immediate sequence of work. It should be short, ordered, and
-biased toward the next deliverable rather than a long roadmap.
+1. Triage the list into quick wins, medium refactors, and long-horizon cleanup.
+2. Tackle the cheapest high-impact items first:
+   - recursive `__pycache__/` ignore and cleanup,
+   - regex raw-string warning cleanup,
+   - `count()` plus `first()` query cleanup,
+   - single-call `ffprobe` refactor.
+3. Decide whether maintenance/tooling command imports should be split from media-runtime imports before adding more CLI maintenance surface.
 
 ## Delete When
 
-Use this to define when the scratchpad should disappear. That keeps it clearly
-temporary and helps prevent it from turning into shadow documentation.
-
-
-## Suggested Style
-
-- Prefer short bullets over long prose.
-- Keep facts, questions, and rough sketches in separate sections.
-- Add custom sections only when they help the next iteration move faster.
-- Move durable outcomes out of the scratchpad once they stop being temporary.
-
-
--->
+- Delete this scratchpad once the optimization backlog is either converted into issues/work items or distilled into durable project guidance.