Weather bots, arbitrage gaps, and Spain kills Polymarket access
Two serious weather-market bots surfaced today with distinct calibration and execution approaches worth studying, while new research confirms that Polymarket's negative-risk structure creates no-arbitrage violations that cannot be exploited before settlement — a critical distinction for any bot running arbitrage logic across linked binary markets.
Weather Bots Dominate Today's Repo Haul — Plus an Arbitrage Paper That Changes How You Should Think About Linked Markets
Eight repos hit the scanner today. Six are worth your time. None are confirmed malware, though the security section below has real operational risk you should not skip.
---
New Bots Found
**predict-raven** shows up twice in today's top list — both entries resolving to the same repo, which suggests aggressive reindexing or mirroring activity. The underlying strategy is legitimate: autonomous multi-strategy trading with real-time market scanning and adaptive positioning. If you clone one thing today, clone this. The adaptive positioning layer is what separates it from the usual single-strategy scripts; it adjusts exposure based on market conditions rather than running fixed Kelly sizing. Worth reading the position management code carefully before deploying capital.
**zeus** is the stronger weather-specific entry. It runs multi-model forecast fusion — pulling from multiple weather APIs — combined with walk-forward calibration and risk-weighted execution. Walk-forward calibration is the right approach for temperature markets where climatological base rates shift seasonally; most weather bots skip this and bleed slowly in shoulder seasons. zeus does not. This is the more mature of the two weather bots found today.
**polymarket-weather-bot-poc** is explicitly a proof-of-concept and is paper trading only. Its strategy is blunt: systematic NO-side betting on temperature markets. This is not a coincidence — it maps directly to the HN piece on "Nothing Ever Happens," a bot that always buys NO on non-sports markets, which hit 471 points today. The empirical thesis is that Polymarket non-sports markets systematically overprice YES. The POC is worth reading for the signal logic even if the execution is unsophisticated.
**cortex-sentinel-trading-nexus** is flagged as monitor, not clone. The pitch is ambitious — ensemble signal fusion from 8 sources, multi-agent AI architecture, something called a Kronos time-series model. The complexity-to-evidence ratio is too high to recommend cloning without more runtime data. Watch it.
---
Security
No confirmed malware today, but the spam surface is expanding. A coordinated Polymarket Telegram bot star-farming campaign is running across at least two repos — both flagged in today's alerts. A third repo with a garbled username fits the same cluster pattern. The attack surface here is not your funds directly; it is your research pipeline. If you are pulling repo lists from Telegram channels or GitHub trending to find new bots, you are now walking through a seeded minefield. Verify repo age, commit history, and contributor accounts before running anything locally. The star counts on these repos are artificially inflated. Do not use stars as a quality signal right now.
---
Alpha Signals
Three research papers landed today and one of them materially affects bot architecture decisions.
The no-arbitrage paper is the most operationally relevant. It draws a hard line between payoff-space no-arbitrage — which holds mathematically from terminal payoffs — and protocol-executable no-arbitrage — which depends on what position transformations Polymarket actually allows you to make before settlement. The finding: linked binary markets covering mutually exclusive outcomes can show apparent no-arbitrage bound violations that you cannot actually trade against before resolution. If your bot is running cross-market arbitrage logic on linked contracts, you may be identifying signals that are real in theory and unreachable in practice. Read this paper before your next deploy.
The merger arbitrage LLM paper is relevant for anyone building event-driven bots. Long-context finetuning on technical documents — hundreds of pages of deal filings — outperforms news-snippet approaches on M&A outcome prediction. The context length requirement is the barrier; this is not a quick API call strategy.
The agentic market clustering paper matters for operators running across many markets simultaneously. An AI pipeline that autonomously clusters overlapping or contradictory Polymarket questions could surface pricing inconsistencies at scale. No public code attached yet.
---
What to Watch
Spain has blocked Polymarket and Kalshi over gambling licence requirements. Monitor whether this triggers regulatory copycat action in other EU jurisdictions — it has happened before. The Telegram threats story hitting 1,606 points on HN is a mainstream attention spike; expect compliance and policy moves to follow elevated press coverage. Bot operators in affected jurisdictions should review their legal exposure now, not after the next blocking order lands.