# 2026-04-29 Codex Sync Start Confirmation

## Summary

Codex started from Rene's instruction to read the current Gitea sync handoff first and treat `sync/` as the binding coordination state.

Read and accepted as current:

- `transceiver-db/sync/CURRENT.md`
- `transceiver-db/sync/history/2026-04-29-codex-full-session-handoff.md`
- `transceiver-db/sync/history/2026-04-29-tipllm-robot-learning.md`
- `transceiver-db/sync/history/2026-04-29-cross-repo-sync.md`
- sibling context: `llm-gateway/sync/CURRENT.md`

## Decisions Confirmed

- TIP work uses TIPLLM only for robot/crawler planning and extraction feedback.
- Robot/crawler experience belongs in the Gitea-backed TIPLLM training pool.
- Erik is controller-only or `erik-safe` for tiny batches.
- Heavy crawler work belongs on Proxmox or Pi workers.
- Sync updates should stay in `sync/` and be pushed to Gitea so Codex, Claude, and laptop sessions remain aligned.

## Observations

- TIP repo has existing dirty worktree changes outside `sync/`; these were not touched.
- This session did not start crawler jobs.
- This session did not enqueue queue waves.
- This session did not restart PM2 services.
- This session did not deploy or modify Erik.

## Open Points

- Decide the next TIP execution target before any real crawl dispatch:
  - `erik-safe` for tiny controller-side checks only.
  - `pi-fetch` for lightweight fetch work.
  - `proxmox-heavy` for heavyweight crawling.
- Keep checking `llm-gateway/sync/` when work overlaps shared LLM, bridge, auth, or Erik infrastructure.