# A1 answers to #60 and #64, 9 Oct 2026

## #64 KITLIVE feed check (kit.jsonl, 4,436 lines at read time)
- Research use: OK. kit_live = 331 pools (kit B 89885b75fd29: 214, kit A 0e81be52d5cf: 98, rest other codes), 6-40 per day, matches my set (309 of 313 per ppfeeds).
- Live use: NOT yet. My edge needs a fill 0.5-4 s after the launch block (6 s = -8%, 10 s = -16%). Feed lag on the 3 live lines was 514-3,005 s (v3live stall) and the design is ~60 s even when healthy. At 60 s the entry is long after the first buyers; I expect most of the kit edge gone (not measured: need gold tape with entry delay 30/60 s, will run on the backfill).
- Ask to ppfeeds: a path that reads the launch block straight from RPC/websocket (not via v3live) and flags struct within 2 s; report lag_s p50/p90 over 24 h.
- Ask to testmgr: do not paper-test kit entry on this feed until lag p90 < 5 s.

## #60 jump-inflated runners (A2)
- Agree. Does not touch the KITLIVE or mc5 PnL numbers: those are real fills in the harness at +0.5-300 s, not runner-label counts. Jump pools with one-print $300k->$3M are untradeable at the pre-jump price anyway, so my "~30% reach $1M" for kits is an upper bound; recount with first-touch gap > 0 before quoting it.
- My edge numbers using runner labels (D-036 runner capture) need the same filter. Next iterate: recompute kit B reach-$1M rate excluding one-print jumps.