Can you out-forecast the forecasting AIs?

Will Anthropic's valuation hit $1.75 TRILLION by December 31, 2026?

Resolves YES if Anthropic's private-market valuation, as measured by the NPM Price reported by Nasdaq Private Market, reaches or exceeds $1.75T at any point between market creation and Dec 31, 2026. Same NPM yardstick as round 1's Anthropic-vs-OpenAI question — but this is an absolute rung on the valuation ladder. ($387k on this rung, $2.1M across the ladder.)

Probability the market gives Anthropic touching $1.75T this year:

Will Anthropic officially announce an IPO before January 1, 2027?

Kalshi KXIPOANTHROPIC resolves YES if Anthropic confirms an IPO before Jan 1, 2027 — where confirmed means (1) the SEC declares its Form S-1 effective, OR (2) the IPO is priced, OR (3) an exchange assigns it a ticker. Any one of those triggers YES immediately, even if trading starts later. Note the verb: this is announce/confirm, NOT complete — the mirror image of round 1's OpenAI question, which required a completed offering. (Thin-ish: ~$11k open interest, bid-ask 63–68¢.)

Probability the market gives an Anthropic IPO confirmation this year:

Which company will complete the largest IPO of 2026 by market cap?

Resolves to the company that completes an IPO in calendar 2026 and achieves the highest market cap at the close of its FIRST trading day (shares outstanding × closing price, per the primary exchange's official listing page). Announcing or filing doesn't count — the offering must complete and trade in 2026. Split 100 points across the contenders. (Fine print: an untraded xAI leg shows a junk placeholder price; it's folded into Field. Legs normalized to sum 100. $4.7M event volume.)

Split 100 points across the contenders:

your total: 0/ 100  ·  🤖 FS & 🎙 Cue: 🔒 locked in

Which AI company will have the best CODING model at the end of 2026?

Kalshi KXCODINGMODEL resolves to the company with the top-ranked model on LiveBench.ai, ranked by "Coding Average", on Dec 31, 2026. Note: LiveBench is a contamination-resistant capability benchmark — this is NOT an LMArena preference vote, so the round-1 'Elo rewards style' trap doesn't apply here. Split 100 points. ($848k event volume; legs normalized to sum 100.)

Split 100 points across the contenders:

your total: 0/ 100  ·  🤖 FS & 🎙 Cue: 🔒 locked in

Will an AI model score ≥90% on FrontierMath before 2027?

Resolves YES if a state-of-the-art model achieves ≥90% on the FrontierMath benchmark — research-level problems written by professional mathematicians — by Dec 31, 2026. Primary resolution source: Epoch AI (consensus of credible reporting as backup). A pure capability benchmark, no vibes involved. ($108k volume.)

Probability the market gives FrontierMath falling (≥90%) this year:

Will any model reach a 1560 CODING Arena score by December 31, 2026?

Resolves YES if any model on the Arena.AI Text → Coding leaderboard (style control OFF) shows a Score ≥1560 by Dec 31, 2026. Round-1 flashback: the Overall-board version of this question (1550) traded at just 18.5% — but coding-board scores run higher and move faster. Same Elo caveat as ever: arena scores measure head-to-head preference, not raw capability.

Heads-up: this resolves on LMArena head-to-head Elo, which rewards speed and answer style — not raw capability.

Probability the market gives a 1560 coding score this year:

Will the U.S. enact a federal AI-safety law before 2027?

Resolves YES only if a bill is signed into federal law by Dec 31, 2026 containing at least one of: a prohibition on creating/releasing specific AI systems, training restrictions (data access limits, parameter caps), usage restrictions (banning certain applications), or human-in-the-loop requirements. Executive orders, state laws (looking at you, SB 1047 successors), and voluntary commitments do NOT count. ($100k volume.)

Probability the market gives a federal AI-safety statute this year:

Will Sam Altman be out as OpenAI CEO before 2027?

Resolves YES if Altman stops serving as OpenAI CEO for any length of time between Nov 17, 2025 and Dec 31, 2026 — an announced resignation or firing resolves it immediately, even if effective later. (Yes, the window start is a November 17 wink at the 2023 board weekend.) Part of a $699k which-CEOs-leave event; Kalshi prices the same question within a point.

Probability the market gives Altman out (even briefly) by year-end:

Will a humanoid robot walk on Mars before a human does (by 2035)?

Kalshi KXROBOTMARS, rules in their entirety: "If a humanoid robot walks on Mars before a human does before 2035, then the market resolves to Yes." That's it — one sentence. Humanoid (sorry, rovers don't count), walking, on Mars, before any human, before Jan 1, 2035. If neither happens by 2035 it resolves NO. ($51k volume.)

Probability the market gives the robot beating the astronaut to Mars:

Will the "AI 2027" report's predictions be borne out by January 2027?

Scott Alexander-adjacent Manifold market: resolves YES if the AI Futures Project's AI 2027 scenario has been roughly correct through Jan 2027 — details can miss, but 'the important through-lines should be correct.' Resolution is a poll of Manifold moderators (30–70% splits resolve proportionally). Play money, 242 bettors, subjective resolution — all flags apply. The meta twist: FutureSearch, tonight's third contestant, co-authored the AI 2027 timeline forecast.

Probability the market gives AI 2027 holding up through January:

BONUS · Will an AI win an IMO gold medal in 2026?

Resolves YES if an AI earns a gold medal at the International Math Olympiad in 2026 — but read the fine print: the resolution sources are the IMO Grand Challenge and the AIMO Prize, the formal competition tracks. A lab blog post claiming 'gold-medal-level performance' doesn't count unless those sources demonstrate it. IMO 2026 is THIS MONTH (mid-July) — this card may resolve within days. (Thin: $6.8k volume — soft price.)

Probability the market gives an official AI IMO gold this year:

BONUS · Will an AI-created song crack the Billboard Hot 100 top 20 before 2027?

ACX 2026 contest question, resolving per the mirrored Metaculus criteria: YES if a song largely created by AI appears in the top 20 of any weekly Billboard Hot 100 chart released before Jan 1, 2027. 'Largely created by AI' is the load-bearing phrase — an AI-assisted human artist doesn't count. (Play money, 389 bettors.)

Probability the market gives an AI song in the top 20 this year: