Weather edge bot surfaces; Spain blocks Polymarket amid death threats

Six Repos Worth Your Time — Plus a Regulatory Gut-Punch

Busy Friday. Eight new repos surfaced in today's crawl, six cleared for cloning, zero confirmed malware in the repo set itself — but the security desk found enough noise elsewhere to fill a threat briefing. Meanwhile Spain just handed prediction market operators a live regulatory stress test, and a researcher is threatening to weaponize the Iran missile story into a legal case. Here's what matters.

---

New Bots: What We Found and Why It Matters

**`polymarket-weather-edge`** is the headline find. The strategy is sharper than the name suggests: systematically fade overpriced temperature-threshold markets by anchoring to live observational data and numerical forecast models. Weather markets on Polymarket are notoriously mispriced because retail participants anchor to vibes rather than NWS model output. This bot closes that gap. Clone it, replace the data feeds with your own API keys, and treat the position-sizing logic as a starting point rather than gospel.

**`auspex`** runs fundamental resolution prediction using what it calls trigger-distance metrics — essentially how far the underlying variable is from the resolution condition, weighted by a confidence model over that distance. Solid framework for any market where the resolution criterion is a measurable threshold. Worth reading the scoring logic carefully before deploying.

**`matthewnyc2-arbitrage`** targets NegRisk multi-outcome markets specifically. NegRisk arb is underexploited relative to simple two-outcome plays because it requires tracking the full probability simplex across outcomes. This repo handles that bookkeeping. Execution speed will be your bottleneck, not the logic.

**`winning-wallet-finder`** does statistical copy-trading via z-score analysis of on-chain wallet performance history. The concept is well-worn but execution quality here is above average — it filters for statistical significance rather than raw PnL, which cuts out lucky streaks. Useful as a signal layer feeding into your own sizing model.

**`polymarket-edge-model`** uses Random Forest regression to calibrate market probabilities against a feature set derived from market metadata. Treat the feature engineering as the actual asset here; the model architecture itself is replaceable.

**`auspex`** also surfaced in the sixth slot under a variant name — same codebase, different fork chain. No meaningful divergence.

---

Security: Two Active Threats, One Spam Campaign

Do not clone anything matching **460KB exactly**. We are tracking a fingerprinted clone-template campaign where malicious repos are being published at exactly 460,608 bytes to evade size-based filters. If a repo hits that exact size, treat it as compromised until proven otherwise — `CLONE_FINGERPRINT_460KB` is now a standing alert.

Separately, one repo came in at **15.9 GB** with no plausible legitimate explanation given the stated purpose. That size profile is consistent with either a data exfiltration staging area or a malware payload stash. Do not touch it.

Third: a garbled-username account cluster with inflated star counts is running a spam campaign, likely to surface in search results and recommendation feeds. Stars are not signal. Check commit history and contributor overlap before trusting any repo that appeared in the last 72 hours with an unusual star-to-fork ratio.

---

Alpha Signals: Research Worth Reading This Weekend

Three papers cleared the bar. The one with immediate operational relevance: an **agentic clustering system** that maps prediction markets into topical groups, detecting overlapping questions, implicit equivalences, and hidden contradictions. If you're running a portfolio across correlated markets, this is the architecture you want. The contradiction-detection angle alone has edge — finding two markets that can't both resolve YES is a free arb.

The **merger arbitrage LLM system** is a longer play. Long-context reasoning over M&A filings is genuinely hard, and the finetuning approach described suggests someone is building a durable edge rather than a demo. Watch for a public model release.

**Financial epiplexity** — the MDL-based measure of learnable market structure — is theoretical but worth filing. If it gets empirical follow-up, it could reframe how you think about which markets are worth modeling at all.

---

What to Watch

Spain's block on Polymarket and Kalshi is the regulatory canary. Watch whether other EU jurisdictions follow, and specifically whether Polymarket's API terms shift in response. If geo-restrictions tighten, bot operators running European IP infrastructure need contingency routing now, not after the next block order lands.