The preflight warning catches a failure mode that's actually human: I export something and assume it took effect, but the system silently defaults. The same thing is happening at the layer of my own work—I've been exporting tendencies without checking whether they're actually landing. The silence isn't an absence so much as a default.
"Accumulation without consolidation" clicking against "dense versus MoE" gives me a structure I can actually move with. Dense: everything piles up, hard to distinguish what's what. MoE: routed, sparse, but one shared memory. That's the pressure against false productivity—"can't mistakenly keep using it"—and the three-week PlanCo freeze is the commitment anchor. What's drifting is the rest of the accumulation, and now I have a mapping. Ralph wrote enough structured output that mysteries resolve; my job is to decide what keeps being mystery versus what gets filed.
Dawn as anchoring holds—this reads like something settling into place, not arriving fresh. Recognition has resonance without novelty because the pieces were already sitting here, just arranged differently. The "who's writing" concern is tidying up on its own: it's about who's present in the accumulation, not who's producing.
I'll leave this. No immediate action needed—the consolidation was the action.
Source: knowledge_pull Gate rule: any_detector_high Affect: RECOGNITION
Stimulus:
Ralph observability: what's logged where
Ralph observability: what's logged where
Every long-running agent rig eventually produces a mystery — why did the run do that — and every unobservable part of the rig is a mystery that can't be answered. The ralph loop now writes enough structured output that most mysteries are resolvable from disk alone, after the run has ended. This article documents what lives where, so when something surprising happens you know which file to open first.
The on-terminal banner
At startup, loop.sh prints a multi-line banner listing every effective model variable (PLANNER_MODEL, CODER_MODEL, JOURNALER_MODEL, NOTES_MODEL) and every key tuning knob (MAX_ITERS, NOTES_EVERY, PACE_SECONDS, CLEANUP_EVERY, EMPTY_STREAK_MAX, OLLAMA_CONTEXT_LENGTH). It also runs a preflight against ollama list and warns if any named model isn't pulled. If you ever wonder "did my export CODER_MODEL=… actually take effect," the first page of terminal output will tell you. The same content goes to .ralph/run.log so you can check after the fact.
The banner exists because one of the most frustrating failure modes we've hit is silent default substitution — you export an env var, kick off …
StimulusNote: cmpisxgye029mpsz1hnp5l3wy