Files
waggle-os/docs/launch/drafts/2026-05-10-pavlukhin-evolveschema-arxiv-email.md
Oleg Maslov 0c3e2ead3b
Some checks failed
Installer Smoke / installer-smoke (push) Has been cancelled
moving
2026-09-02 10:10:29 +02:00

4.5 KiB

DAY0-CRED-05 / CONT-03 — Pavlukhin EVOLVESCHEMA arxiv submission email (DRAFT)

Target: Mikhail Pavlukhin Author: Marko Marković (Marko sends; I draft) Purpose: Path A endorsement chain — request standalone EVOLVESCHEMA arxiv preprint so Marko's LoCoMo paper can cite it + benefit from Pavlukhin's existing arxiv-author status for endorsement. Status: READY-TO-SEND. Marko reviews + adjusts tone, then sends. Reply window: 24-72h. Path D fallback (methodology.md in OSS repo) already live if Pavlukhin doesn't respond. Authority: PM master §3a — Path A primary; v1.0 Phase 2 DAY0-CRED-05.


Subject line options (pick one)

  • A (direct): "Quick ask: standalone EVOLVESCHEMA arxiv preprint for Waggle paper citation"
  • B (relational): "Following up on EVOLVESCHEMA — quick arxiv coordination request"
  • C (concise): "EVOLVESCHEMA arxiv question — 5 min read"

Recommend A — sets clear expectation in the inbox preview.


Email body (DRAFT — Marko reviews + sends)

Hi Mikhail,

Hope you're well. Brief context + a coordination ask.

We're publishing a Waggle/hive-mind paper next week on apples-to-apples methodology for LoCoMo (long-context memory benchmark). Our substrate score is 74% on the canonical fixture vs Mem0's published 66.9% — a +7.1pp gap that we trace partly to methodology choices (judge ensemble, no-context vs full-context cell definitions) and partly to the substrate architecture (bitemporal KG + IPB compression).

The architecture relies on a prompt-evolution component that descends directly from your EVOLVESCHEMA work. We're citing you in §3.2 and §5.3 of the paper, and the lineage attribution is unambiguous — your name is in the paper either as a co-author candidate or as a strong primary citation, your call.

The ask: Would you consider submitting a standalone EVOLVESCHEMA preprint to arxiv (cs.AI) in the next 1-2 weeks? Even a 4-6 page write-up of the core method, separate from any longer paper you're working on, would be enormously helpful because:

  1. It lets our Waggle paper cite an arxiv ID rather than an informal lineage reference.
  2. As an existing arxiv author, you could endorse our submission, which gates first-time submissions to cs.AI.

If you're already planning a fuller EVOLVESCHEMA paper, even a placeholder/draft preprint would work for both purposes — the endorsement chain only needs the arxiv-author status to exist, not the final version.

Happy to share our paper draft if helpful (currently at the methodology-doc stage; the LoCoMo numbers are pre-registered + reproducible). And of course we can coordinate authorship/citation framing whichever way suits you best.

No pressure if the timing doesn't work — we have a fallback path that doesn't require arxiv coordination. But Path A (your standalone preprint) is materially stronger for the paper, and I wanted to offer it first.

Quick yes/no/maybe in a few days would help us calibrate the Day-0 sequencing.

Thanks, Marko


Tone calibration (for Marko's review)

  • Specific number: "74% vs Mem0's 66.9%" is concrete; signals we're serious about evidence.
  • Attribution latitude: "co-author candidate or primary citation, your call" — flexible without being pushy.
  • Time-bounded but not pressuring: "next 1-2 weeks" + "Quick yes/no/maybe in a few days" — clear window but no urgency.
  • Path D acknowledged: "we have a fallback path" — signals this isn't blocking the launch (per PM master §3a) and respects his time.
  • Length: ~310 words. Within "5 min read" framing of subject C.

Send mechanics

  1. Marko reviews + adjusts tone (especially the "your name is in the paper either as a co-author candidate or as a strong primary citation" framing — this depends on M-04 strategic decision pending on attribution).
  2. Marko sends from his personal email (not company; lineage is academic).
  3. Marko marks calendar reminder for T+72h to chase if no reply.
  4. Path D activation: if no reply by 2026-05-17 (T+7d), pivot to methodology.md citation in OSS repo per decisions/2026-05-02-track-e-arxiv-7-decisions.md Path D fallback.

Reply handling

  • YES: Marko replies with paper draft + arxiv submission timeline coordination. CONT-03 closes.
  • MAYBE / "let me think": Marko gives him 72h. Pre-schedule a chase email.
  • NO: Marko thanks him + activates Path D. No bridge burned; attribution remains in paper.
  • NO REPLY 7d: Path D auto-activates. No further follow-up.