Flarepoint

flarepoint

Annotate load-bearing claims with their epistemic basis (verified/inferred/recalled/guess) and a per-response footer tally. Pure annotation — never changes what the model does.

Sizes each commit (patch, minor, or major) and remembers the largest, then applies one version bump when you release. The model judges the size; you confirm. Suggests by default; set auto to apply without asking.

Resume a previous session. Saves a structured note of where you left off (goal, files changed, commands run) plus the likely steps ahead, and surfaces it at the start of the next session so you can continue without re-explaining. It proposes the next steps and waits; it never starts work on its own.

Understanding-first mode. On exploratory tasks (refactors, design, ambiguous scope) the model grills you toward mutual understanding before writing code, so you can debug and own the result. Sets a prior at the task edge and skips trivial work; never steers mid-solve. Modes: off, on (default).

Comment-concision mode. Shapes the code the model writes so comments earn their place — a non-obvious why, or meaning a name can't hold (config, units, constraints) — and stay short; deep backstory goes to the commit. API docstrings are exempt. Sets a prior at the task edge; only ever touches comments, never behavior. Modes: off, new, tidy (default). Plus /fp-minify:strip for deliberate bulk de-commenting.