slv1 Topics

For Agents

Living index of themes for this project. Each H2 is a topic; bullets are wikilinks to related notes. Updated by obsidian-documenter when documenting work. Read by historian at bootstrap. Topics kept alphabetical.

Alpenglow & Protocol Risk

• slv1-copy-trading-feasibility-2026-07-03 — Alpenglow defused near-term (Votor-only, keeps Turbine); real shred-visibility threat is the later unscheduled Rotor SIMD

Alpha Validation Backtest (solv1-alpha)

• solv1-alpha-first-verdict-no-go-2026-07-04 — FIRST REAL VERDICT: NO-GO on naive copy-trading (run f02ed44f; cohort +23.0% vs random basket +109.5%; segment split scalp −20.5% / swing +3.4% / position +296.1%; filtered−raw +14.8pp; survivorship upper bound; refinement = positions-only)
• solv1-alpha-validation-backtest-2026-07-04 — the built walk-forward go/no-go harness (16 tasks + 1 must-fix, 48 tests; score→filters→cohort→lookahead-guard→mirror→verdict; walk-forward integrity sound; survivorship upper-bound caveat verbatim)
• solv1-alpha-data-source-pivot-2026-07-04 — first real run blocked (non-archival Chainstack + bot-polluted leaderboard) → GMGN wallet_activity harvest
• solv1-alpha-gmgn-ingester-increment-2026-07-04 — the browser-harvest → file-ingest increment, now BUILT (G2/G3/G4, 56 tests, migration 0003_solv1_alpha_klines; decay pivoted to 1s klines; A8–A16 reused, A4–A7 replaced)

Arbitrage (Strategy 5)

• solv1-arbitrage-feasibility-measurement-2026-07-05 — feasibility + passive measurement, leaning NO-GO. First PRINCIPAL-SAFE strategy (atomic Jito bundle reverts → lose fees not inventory; flash loans push capital→0). Desk research = CONDITIONAL GO to measurement only (mainstream arb = solved colocated machine ~90M/yr @ $1.58,96$1.8–3.8k/mo; only solo lane = latency-insensitive long-tail PumpSwap × Meteora-DLMM). arb-measure harness (Python, branch arb-measure-harness): mid-price metric = FALSE GO, depth/slippage-adjusted = REVERSED (spreads persist because slippage eats them; realizable ~$0.33medianattrivialsize,MENSA+$1.58@5SOL/neg@10, ANSUM +$0.37@1SOL/neg@2); optimistic Meteora upper bound (DLMM ~6× thinner); one positive-persistent event = commoditized latency race (bin lag ~3 slots)
• slv1-strategy-candidates-2026-07-04 — candidate #5 in the scorecard; the meta-pattern (ShredStream table-stakes 3× · selection is the edge · beta is the payoff)

Copy Trading (Strategy 1)

• slv1-copy-trading-feasibility-2026-07-03 — conditional GO verdict, kill criteria, MVP plan
• solv1-alpha-validation-backtest-2026-07-04 — offline go/no-go on whether a copyable edge exists at all (upstream of the latency question)
• solv1-alpha-first-verdict-no-go-2026-07-04 — the empirical answer: NO-GO on naive copy-trading (loses to a random meme basket in a bull window); the surviving edge is position/long-holds only — scalps are killed by entry decay
• slv1-strategy-candidates-2026-07-04 — the “what next after NO-GO” menu: candidate #1 (co-buy consensus) directly replaces single-wallet copying to dodge survivorship

Data Ingestion & Harvesting (GMGN)

• solv1-alpha-data-source-pivot-2026-07-04 — Chainstack non-archival (~3h wallet history) + raw leaderboard bot/dead pollution; fix = GMGN wallet_activity via browser cf_clearance (headless = 403); no slot → synthetic slot
• solv1-alpha-gmgn-ingester-increment-2026-07-04 — browser-harvest-to-JSON → Rust file-ingester → solv1_alpha_trades; two-phase semi-manual run; G1 bot/dead filter (activity-rate + smartmoney). BUILT: decay pivoted to GMGN 1s klines (token_kline/sol/<token>?resolution=1s; per-fill harvest of 391 tokens infeasible); harvest bridge = same-origin fetch → blob download to ~/Downloads (localhost sink CSP-blocked; needs “automatic downloads” perm once); fire-and-forget async loop; ~1.3k fetches/session rate cap

Jito ShredStream & Latency Infrastructure

• slv1-copy-trading-feasibility-2026-07-03 — latency ladder, access/reseller options, tip economics, refuted vendor benchmarks
• solana-v0-alt-classification-trap — geyser leg is venue-classification authority (shred static-key scans miss ALT-loaded router venues); Jito tip_floor returns f64 SOL in a one-element array
• solv1-rig-increment-1-build-2026-07-03 — the built dual-feed harness measuring shred-vs-Yellowstone arrival delta (mpsc pipeline, histogram percentile authority, 1:100 raw sampling)
• slv1-strategy-candidates-2026-07-04 — reframes banked ShredStream as the rarest UNUSED asset (copy-trading never needed it); candidates #2 (launch sniper) and #4 (back-run MEV) are the strategies that finally make it load-bearing
• solv1-arbitrage-feasibility-measurement-2026-07-05 — ShredStream is table-stakes for the 3rd time (copy / sniper / arb): the arb long-tail lane is latency-insensitive, and the one positive-persistent spread the harness found was a pure latency race (Meteora bin price lagging PumpSwap ~3 slots) = commoditized — the recurring “speed is a cost of entry, never the edge” lesson

Launch Sniper (Strategy 2)

• slv1-strategy-candidates-2026-07-04 — candidate #2 data-filtered launch sniper: ShredStream detects pump.fun launches pre-confirmation → Jito-bundle snipe quality-filtered launches (no survivorship, no follow-decay); most bot/rug-heavy niche → feasibility-first (docs/sniper-feasibility.md) → NO-GO (quality filter couldn’t clear the adverse selection; speed is table-stakes not edge)

Legal, ToS & Regulatory

• slv1-copy-trading-feasibility-2026-07-03 — Jito ToS structurally adverse (revocable → feed-agnostic architecture); EU front-running/sandwiching MiCA grey zone

Measurement Rig (solv1-rig)

• solv1-rig-increment-1-build-2026-07-03 — increment 1 built (dual-feed delta harness; 15 tasks, per-task + whole-branch “ship it”; 21 tests + Docker-Postgres e2e; blocked on RPC Fast creds + GCP FRA VM)
• slv1 — Week-1 measurement-rig design spec (solv1-rig) + 2026-07-03 spec-grounding pass
• solana-v0-alt-classification-trap — venue-classification authority + volume / tip / sizing / slot-time corrections
• supabase-from-gcp-connection-gotchas — the rig’s Supabase storage connection recipe
• rust-otel-logs-signoz-export — the rig’s log path; cross-project Rust OTel root cause
• rust-protoc-docker-wkt-proto — the rig’s Docker build: protoc well-known-type .proto gotcha
• yellowstone-grpc-proto-layout-versioning — the rig’s Yellowstone adapter: proto layout + version pinning

Observability (OTel / SigNoz)

• rust-otel-logs-signoz-export — Rust OTel logs don’t export by design; opentelemetry-appender-tracing bridge fix (cross-project, in dev-setup/)

Rust Build & Dependency Gotchas

• rust-protoc-docker-wkt-proto — Docker/CI protoc needs libprotobuf-dev for well-known-type .proto imports
• yellowstone-grpc-proto-layout-versioning — proto module layout + client/proto non-lockstep version pinning
• rust-otel-logs-signoz-export — tracing-opentelemetry version pinning (0.33 needs otel 0.32) + logs not exported by design
• supabase-from-gcp-connection-gotchas — sqlx↔Postgres type-decode casts (make_interval int4, sum::bigint) + pooler mode

Solana Transaction Mechanics

• solana-v0-alt-classification-trap — v0 / ALT: top-level program IDs always static (require_static_program_ids) vs CPI programs that can be ALT-loaded; static-key scan blind spot
• yellowstone-grpc-proto-layout-versioning — yellowstone-grpc proto types under solana::storage::confirmed_block; SubscribeRequestFilterTransactions.token_accounts (tag 30) breaks exhaustive literals
• solv1-alpha-data-source-pivot-2026-07-04 — GMGN wallet_activity event schema (buy/sell/launch, quote_amount SOL, priority/tip fee, dex); no slot in payload → synthetic slot from timestamp

Storage & Database (Supabase / Postgres)

• supabase-from-gcp-connection-gotchas — Supavisor session mode :5432 (sqlx prepared statements); RLS-via-ownership vs BYPASSRLS / service_role; sqlx type-decode casts
• solv1-rig-increment-1-build-2026-07-03 — 8 solv1_ tables on the Supabase mgmt project; RLS-on / FORCE-off / anon-denied; applied live via MCP
• solv1-alpha-validation-backtest-2026-07-04 — 7 solv1_alpha_* tables on the same Supabase mgmt project; session-mode :5432, RLS-on / FORCE-off

Strategy Pipeline & Feasibility Research

• slv1 — strategy pipeline status (Strategy #2 market-intel sniper queued)
• slv1-strategy-candidates-2026-07-04 — scorecard of the full arc: 5 strategies explored, all resolved — Copy-trading NO-GO · Launch-sniper NO-GO · Co-buy consensus NO-GO · Meme-beta timing = RISK OVERLAY (not alpha) · Solana arbitrage = leaning NO-GO. Meta-pattern: banked ShredStream = table-stakes 3× (copy / sniper / arb), never the edge; the only durable levers are SELECTION (+14.8pp filtered−raw) and risk-managed BETA (+109.5% basket). (#4 Jito back-run MEV still parked/unexplored.)
• slv1-copy-trading-feasibility-2026-07-03 — deep-research methodology (2 rounds, 103 + 56 agents, adversarial verification); desk research exhausted, remaining gaps are MVP-only
• deep-research-subagent-fleet-directive — workflow sizing / model directive derived from these runs
• LOG — 2026-07-03 spec-grounding pass method: 3 parallel Opus verify agents (feeds / chain / infra) over the design spec caught 7 material errors pre-build