Discovery runbook: GitHub topics + arXiv + vLLM issue + HF back-links + Papers with Code + 2 × DEV.to

[FluffyAIcode]

Summary

Six-task discovery runbook + one ready-to-execute artifact per task.
Each artifact is self-contained, targets a specific search-engine /
AI-answer-engine surface, and mirrors the NexusQuant launch pattern
that currently dominates "E8 KV compression" query results.

PR content is entirely documentation + copy-paste kits. Nothing
runs code. No benchmark numbers are introduced that are not already
backed by JSON under reports/.

Files added

docs/announce/
  discovery_runbook.md                    — top-level 6-task runbook
  arxiv/
    SUBMISSION.md                         — full arXiv submission checklist
  dev_to/
    post_1_theory.md                      — DEV.to post 1 (~1200 words, theory)
    post_2_practice.md                    — DEV.to post 2 (~1000 words, practice)
  hf_space_backlinks.md                   — Space + model-card back-link PR templates
  papers_with_code/
    SUBMISSION.md                         — PwC paper + 5 benchmark rows
  vllm_integration_issue.md               — pre-written vLLM RFC body (NexusQuant-style)

Task → owner → status

#	task	owner	artifact	status
1	GitHub topics (8 terms)	you — repo admin	discovery_runbook.md §1	ready — one gh repo edit ... --add-topic ... line away
2	arXiv submission	you — paper author	arxiv/SUBMISSION.md	ready — paste metadata into arxiv.org/submit
3	vLLM RFC	you — GitHub account	vllm_integration_issue.md	ready — paste body + title into vllm-project/vllm Discussions
4	HF Space + model-card back-links	you — HF token holder	hf_space_backlinks.md	Space side done (2026-04-25); 6 model-card PR drafts ready
5	Papers with Code	you — paper author	papers_with_code/SUBMISSION.md	blocked on arXiv ID; form fill-in ready
6	DEV.to × 2	me (wrote) + you (publish)	dev_to/post_1_theory.md + post_2_practice.md	ready — paste into dev.to/new, front-matter is correct

Why the order 1 → 6 matters

1. GitHub topics first — GitHub topic pages index within hours.
   Cheap win, makes the repo show up on
   github.com/topics/e8-lattice etc.
2. arXiv second — arXiv ID is the highest-authority anchor in
   ML and unlocks Task 5. Expected processing: 1 business day.
3. vLLM RFC third — discussion-thread format, needs the arXiv
   ID to reference in the opening post for maximum credibility.
4. HF back-links fourth — depends on arXiv ID for the Space
   badge refresh.
5. Papers with Code fifth — requires the arXiv ID.
6. DEV.to sixth — posts should link forward to arXiv + PwC + vLLM
   discussion; posting them before those exist means editing later.

All artifacts include clear timing guidance, pre-prepared fallback
content for the most likely reviewer asks, and explicit "done when"
criteria.

GEO rationale (why this PR complements PR #54)

PR #54 set up the local GEO surfaces (README, FAQ, blog,
launch kit, CITATION.cff, ACKNOWLEDGMENTS.md, DEPLOYMENTS.md).
This PR sets up the external GEO surfaces (GitHub topic pages,
arXiv, vLLM, HF, PwC, DEV.to) that push traffic toward the local
surfaces. Credit-wise, cross-source consistency of naming and
citation form is preserved: all external artifacts name the same
peer methods (TurboQuant, KIVI, SmoothQuant, HQQ, Quanto, SnapKV,
H2O, Scissorhands) with the same arXiv / DOI anchors as the README
and ACKNOWLEDGMENTS.md.

What's not in this PR

• No actual submissions (I cannot file on your behalf — all 6 tasks
  are owner-operated by design).
• No changes to the existing README / FAQ / blog / announce
  templates (those are PR GEO + Credit: README hero + FAQ + landscape-survey blog + CITATION.cff + ACKNOWLEDGMENTS.md + DEPLOYMENTS.md + launch kit #54's scope).
• No new benchmark numbers or re-runs.
• No vLLM code — the RFC is explicitly a discussion-first filing
  to align on integration path before coding.

Follow-up (after this PR merges)

Once the arXiv ID lands, a small follow-up PR will:

• Replace the DOI — pending badge in README with an arXiv badge.
• Wire the arXiv ID into CITATION.cff (identifiers: list) and
  ACKNOWLEDGMENTS.md.
• Push an updated SPACE_README.md to the HF Space via
  huggingface_hub.HfApi.create_commit so the Space landing page
  carries the arXiv badge.
• Enable Zenodo's GitHub integration and tag v1.5.0-arxiv to mint
  a DOI for the exact commit the arXiv abstract references.

That follow-up is ~5 minutes of automated work. Leave a message on
this PR (or start a new cloud-agent session) with the arXiv ID once
it mints and I'll wire everything up.