Remake, Not Remaster

"Keep the spirit. Throw out the machinery. Build the good version once, from what's true now."

What the Forge Was

Before Flint there was TextForge: a command-line text tool with a local model runtime, a browser admin panel on one port, an API hub on another, and hotkeys wired through a third-party automation app. It did the job. But over time it had accreted machinery — model caching, freshness checks, a whole web UI, two long-running servers. None of that was the point; the point was always "select text, transform it." The rest was scaffolding that had quietly become the building.

The Decommission

The remake did not port that machinery forward. It retired it. The servers, their background jobs, their ports, their wrappers, and their environments were all shut down and left off — and crucially, nothing replaced them server-side. The legacy source was moved to a read-only archive: present as evidence, but with its symlinks stripped so it cannot be launched, restored as a service, or quietly used as a fallback when something breaks. Decommission means the old thing is gone as a runtime, not merely deprecated in a comment.

SURVIVED (the spirit)          RETIRED (the machinery)
---------------------          -----------------------
editable operations            model caching
ordering, enable/disable       freshness / canary scoring
per-op execution settings      the admin web UI
keyboard shortcuts             the API hub server
clipboard compatibility        the two server ports
                               the automation-app hotkeys

Why retired: cwkPippa already owns the model boundary,
so Flint never needs a model runtime of its own.

The distinction is exact. A remaster improves the surface — better textures, higher resolution — while the old content stays underneath. A remake rebuilds from the spirit of the original using current ground truth: the better schema, the cleaner boundary, the honest name, all in one pass. TextForge-to-Flint is a remake. The reason it could shed so much is that the world changed around it: cwkPippa now owns brain, soul, model routing, and inference. The old tool carried a model runtime because it had to. The new one doesn't, so it doesn't.

Remake from current ground truth; don't migrate the old data forward. When you rebuild something, the trap is treating the old version as the source you must preserve. Ask instead what's true now — what another system already owns, what boundary already exists — and build the good version once against that. Migration keeps yesterday's assumptions alive inside today's code.

A decommissioned dependency is not a fallback. The most dangerous form of "keeping the old thing around just in case" is a silent runtime fallback to it. Flint's legacy is archived symlink-free precisely so no code path can revive it. If the new system fails, it should fail honestly — not quietly resurrect the machinery you spent the remake removing.

Code

The legacy runs nowhere — the decommission guard, as a concept·bash

# After the remake, these must all come back EMPTY. If any of them
# returns a live process, port, or job, the old machinery has crept back
# in as a shadow runtime — exactly what the decommission forbids.

pgrep -f 'textforge|8510|8520'      # no legacy process
launchctl list | grep -i textforge  # no legacy launch job
lsof -i :8510 -i :8520              # no legacy port listening

# The archived source stays read-only and un-launchable. Present as
# evidence, never as a service. 'Gone as a runtime' is the whole point.

Exercise

Think of a tool of yours that grew machinery around a small core — extra config, a service, a dashboard, a plugin — that you no longer really use. List what would survive a remake (the spirit) and what you'd retire (the machinery). Then ask the harder question: is there anything you're keeping only as a 'just in case' fallback, and what would break if you deleted it outright?

Hint

The spirit is usually one or two sentences of intent; the machinery is everything you built to support an assumption that may no longer hold. The 'just in case' fallback is the real test — if you can't delete it, you haven't finished the remake, you've only added a new layer on top of the old one.

Remake, Not Remaster

What the Forge Was

The Decommission