PM Notebook

Forecasting

Forecasting Journal

A public track record. Every prediction, scored.

Total
47
Correct
7
Pending
20
Accuracy
46%
Brier score
0.075 +70%

Accuracy is over 27 resolved calls. The Brier score uses the 12 of them with a confidence level. One call is marked not scored and left out of these figures.

Status
Topic
Year

  1. Correct AI

    Layoffs accelerate as companies increasingly rely on AI to replace low/mid-level roles

    In the post
    “The wave of layoffs seen in 2024 will continue as companies increasingly rely on AI.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    85%, assigned March 2026
    Resolves when
    Multiple Fortune 500 tech companies announce layoffs explicitly citing AI automation
    Outcome
    Several Fortune 500 tech companies tied 2025 job cuts to AI. Salesforce’s Marc Benioff said in late August that AI agents let the company cut its support staff from about 9,000 to 5,000, and HP said in November it would cut 4,000 to 6,000 jobs by 2028 as it adopts AI. Microsoft framed its May round of about 6,000 cuts as removing management layers.

  2. Correct AI

    Countries race to set global AI regulations, with some adopting loose rules to gain competitive advantage while others (notably EU) fall behind

    In the post
    “Countries will compete to set global AI rules.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    75%, assigned March 2026
    Resolves when
    At least 3 major nations adopt meaningfully different AI regulatory stances within the year
    Outcome
    The US took a deregulatory stance, revoking the previous administration’s AI executive order in January 2025. The EU began applying its AI Act, with bans from February 2025 and general-purpose model rules from August, though the bloc stayed absent from frontier model leadership. Japan passed a promotion-first AI law in May 2025 with no penalties.

  3. Incorrect AI

    AGI is technically achieved in 2025, likely in Q1 or Q2, based on trajectory from o3 ARC-AGI results

    In the post
    “AGI is technically achieved in 2025”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    40%, assigned March 2026
    Resolves when
    A major AI lab publicly claims AGI has been achieved and the claim is credibly supported by benchmark evidence
    Outcome
    Not achieved by any accepted definition. Expert timelines moved out during H2 2025; the Forecasting Research Institute’s expert panel, surveyed in April and May 2026, puts the median at 2050 (2047 for superforecasters), and prediction markets sit in the 2030s. Resolved 21 Sep 2026. The 2026 predictions post listed AGI discourse among the calls that landed; this entry scores the call as written, that AGI would be achieved in 2025.

  4. Incorrect AI

    Nvidia announces a platform for decentralized computing that lets customers monetize idle GPU compute

    In the post
    “So, they create a platform to network Nvidia hardware that allows its customers to lend and monetize their idle compute.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    30%, assigned March 2026
    Resolves when
    Nvidia launches a product allowing consumer GPU owners to monetize idle compute capacity
    Outcome
    Nvidia announced no platform in 2025 for owners of its hardware to lend idle compute. Its DGX Cloud Lepton marketplace (May 2025) lists GPU capacity from Nvidia Cloud Partners such as CoreWeave and Lambda, which falls outside the criterion’s consumer GPU owners. The idle-GPU market grew through third parties such as Aethir, Vast.ai and Salad. The post itself called this more likely a 2026 initiative, and none had been announced by 30 September 2026 either. Resolved 21 Sep 2026.

  5. Correct Crypto

    AI crypto projects peak in Q1 2025 then crash as valuations hit unsustainable levels and market becomes saturated

    In the post
    “In Q1 2025, the AI craze will reach its zenith as a wave of infrastructure-focused projects is listed on major centralized exchanges (CEXs). However, the excitement will soon fade.”
    Made
    6 Jan 2025
    Resolves by
    30 Jun 2025
    Confidence
    70%, assigned March 2026
    Resolves when
    Top-10 AI crypto tokens peak in Q1 and are down >40% from peak by June 30
    Outcome
    AI-themed crypto tokens, led by AI-agent tokens such as VIRTUAL and AI16Z, topped out around the start of 2025. The AI-agent sector’s market value fell about 40%, from over $10B to $6.6B, by 7 February, and CoinGecko puts the AI narrative’s average 2025 return at -50.2%.

  6. Partial Crypto

    Meme coins stage a comeback as AI crypto narrative fades, driven by retail investors seeking asymmetric returns

    In the post
    “As AI projects lose their shine, meme tokens will slowly regain favor.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    65%, assigned March 2026
    Resolves when
    (a) Meme coins by market cap in CoinGecko top 50 exceed 2024 levels; (b) at least one new meme coin exceeds $5B market cap. Correct if both hold; partial if only one does.
    Outcome
    Partly right. TRUMP launched on 17 January 2025 and passed $5B in market value within days, which meets the new-coin half of the criterion. The sector then shrank: CoinGecko puts total memecoin market value at $150.6B in December 2024 and $47.2B in November 2025, and memecoins averaged a -31.6% return over 2025.

  7. Partial Crypto

    Solana ETF gets approved and Solana becomes a serious contender for flipping Ethereum in market cap

    In the post
    “Solana could make history with an ETF approval, marking the first significant push to challenge Ethereum’s dominance. With strong fundamentals, increased adoption, and institutional interest, Solana might become a serious contender for flipping ETH.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    55%, assigned March 2026
    Resolves when
    Correct if the SEC approves a spot Solana ETF in 2025 and Solana’s market cap reaches 70%+ of Ethereum’s; partial if only one happens.
    Outcome
    First half correct: after the SEC’s generic listing standards (Sep 2025), US spot SOL ETFs began trading 28 Oct 2025 (Bitwise, Grayscale, VanEck, Fidelity, Franklin, 21Shares, Canary). Second half missed: Solana did not become a serious contender to flip Ethereum by market cap. Resolved 21 Sep 2026.

  8. Partial Crypto

    At least one Tier 1 nation announces Bitcoin as a sovereign reserve asset, triggering a supply crunch

    In the post
    “Bitcoin’s adoption as a sovereign reserve asset could enter Tier 1 nations, sparking a supply crunch.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    50%, assigned March 2026
    Resolves when
    G20 nation officially adds Bitcoin to sovereign reserves via new purchases and BTC price hits new ATH within 60 days. Partial if a G20 nation formally adopts Bitcoin as a reserve asset without new purchases or the price condition.
    Outcome
    US signed executive order establishing a Strategic Bitcoin Reserve in early 2025, though it was seeded with seized BTC rather than new purchases. A meaningful step, short of the full supply-crunch catalyst predicted.

  9. Incorrect Gaming

    A mobile Web3 game hits $100M ARR, marking blockchain gaming’s first major commercial success

    In the post
    “this looks explicitly at commercial success, so a game hitting $100m ARR”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    35%, assigned March 2026
    Resolves when
    A blockchain-native mobile game publicly reports or credibly exceeds $100M ARR
    Outcome
    No public evidence shows a mobile Web3 game clearing $100M ARR in 2025. DappRadar data reported by CoinDesk put Q2 blockchain gaming funding 93% below a year earlier, with more than 300 web3 games shut down, and Decrypt called 2025 the year crypto gaming collapsed. Resolved 21 Sep 2026.

  10. Correct Gaming

    Unity or Unreal releases a generative AI co-pilot feature for real-time developer assistance inside the engine

    In the post
    “In 2025, one of the primary game engines (Unity or Unreal) will release a generative AI-powered co-pilot feature.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    65%, assigned March 2026
    Resolves when
    Unity or Epic releases a generative AI co-pilot inside the Unity or Unreal editor to all users during 2025; beta and experimental releases count, third-party plugins don’t.
    Outcome
    Both engines shipped a first-party generative AI assistant inside the editor in 2025. Unity 6.2 (19 August) added Unity AI as a beta that replaced Muse: an Assistant that answers questions, writes code and automates editor tasks, plus Generators for sprites, textures and animations. Unreal Engine 5.7 (November) added an experimental AI Assistant in an editor panel that answers questions and generates C++ code. Neither was labelled generally available by year end.

  11. Correct Gaming

    Gaming DAOs acting as publishers fail to gain meaningful traction due to experience gaps and chasing narratives over fundamentals

    In the post
    “Gaming DAOs as publishers will fail”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    80%, assigned March 2026
    Resolves when
    No gaming DAO reaches 100K paying players or $10M in revenue from games it has published
    Outcome
    No gaming DAO publisher emerged as a credible market force in 2025; most continued to struggle with execution and remained niche. The 2026 predictions post counted this among its misses, saying gaming DAOs faded into irrelevance. YGG Play, the most active DAO-run publisher, had made about $4.5M of lifetime revenue from LOL Land by October 2025. It had crossed $9M across all its games by the end of March 2026, so it stayed under the $10M bar through 2025.

  12. Incorrect Markets

    Universal Basic Income (UBI) moves from theory to active pilot programs in Tier 1 economies

    In the post
    “UBI will move from a theoretical concept to a pilot or policy in Tier 1 economies”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    40%, assigned March 2026
    Resolves when
    US, EU member state, UK, or Japan launches a government-funded UBI pilot covering 10,000+ participants
    Outcome
    No Tier-1 economy started a national UBI pilot in 2025. A privately funded German pilot (107 recipients, run by the NGO Mein Grundeinkommen) published its three-year results in April 2025, Wales’s care-leaver pilot continued, and the US Guaranteed Income Pilot Program Act was reintroduced but not passed. Resolved 21 Sep 2026.

  13. Partial AI

    AI tools redefine hybrid work, automating routine tasks while creating demand for more strategic, human-centered roles (fewer individual contributors, more operators)

    In the post
    “AI tools will redefine hybrid work environments, automating routine tasks while creating demand for more strategic, human-centered roles.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    By 31 Dec 2025, (a) large workforce surveys show AI in routine use among remote-capable (hybrid) knowledge workers, with frequent use clearly up on 2024, and (b) labor-market data show demand moving toward strategic, managerial or operator roles while routine individual-contributor roles shrink. Correct if both hold; partial if (a) holds and (b) is mixed or unsupported; incorrect if (a) fails.
    Outcome
    The augmentation half held: Gallup’s Q4 2025 survey found 66% of remote-capable US employees using AI at work and 40% using it a few times a week or more, and 46% of leaders in Microsoft’s 2025 Work Trend Index said agents already fully automate some workstreams. The demand half is mixed. A Stanford study found a 13% relative fall in early-career employment in AI-exposed jobs where AI automates tasks, which fits fewer individual contributors, but large employers such as Amazon also cut management layers, so evidence of rising demand for strategic or operator roles is thin. Evidence confidence is moderate.

  14. Partial Crypto

    Decentralized capital investment takes off in 2025: new funding options beyond angels, syndicates and VCs emerge for startups, especially in crypto

    In the post
    “Rather than going the traditional route of angel investors, syndicates, or VCs, new options will emerge to broaden the possibilities for the next startups.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    By 31 Dec 2025, (a) decentralized or on-chain routes for startups to raise money (public token sales, on-chain raise platforms, community rounds) grow markedly during 2025, with major firms launching or acquiring dedicated platforms, and (b) they become a widely used alternative to angel and VC funding, judged by annual volume relative to crypto venture funding. Correct if both hold; partial if only one does.
    Outcome
    New options did emerge: Coinbase bought the on-chain fundraising platform Echo for about $375 million in October 2025 and launched its own token sale platform in November, and CryptoRank counted over $2.3 billion raised in token sales across 2025. The scale stayed small next to traditional funding, with Galaxy counting roughly $20 billion of crypto venture investment in 2025, and five large sales (pump.fun, WLFI, Monad, MegaETH, Aztec) took nearly $1.5 billion of the token-sale total. The route widened without displacing VC funding for new startups; evidence confidence is moderate to high.

  15. Partial AI

    2025 sees an explosion of AI-driven marketing tools that hyper-personalize campaigns and tune ad spend

    In the post
    “2025 will see an explosion of AI-driven marketing tools.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    During 2025, AI-driven marketing tools multiply and spread: (a) an independent landscape count shows strong net growth in marketing tools led by AI products, and (b) the largest ad platforms (Google, Meta) ship AI-run campaign creation, personalization and budget allocation as standard products. Correct if both hold; partial if only one does.
    Outcome
    The ad platforms delivered: Google launched AI Max for Search, generative creative tools and agent-style Ads and Analytics assistants at Marketing Live in May 2025, and Meta, already running Advantage+ AI campaigns, set a target to automate ad creation and targeting end to end by 2026. The tool count grew more slowly: the chiefmartec 2025 landscape counted 15,384 marketing tools, up 9% on 2024, and described the growth as subdued and spread evenly across categories, though AI content categories added far more tools than they lost. Evidence confidence is high.

  16. Incorrect AI

    AI marketing tools automate and execute entire go-to-market strategies in 2025

    In the post
    “automate and execute entire go-to-market (GTM) strategies”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    By 31 Dec 2025, AI tools in general commercial use automate and execute complete go-to-market programs (strategy, creative, targeting and budget) as a standard product used at scale rather than in pilots.
    Outcome
    No mainstream tool did this in 2025. Meta’s plan for ads built and targeted from a product image and budget alone carried a 2026 target, and marketing leaders surveyed by Gartner from August to October 2025 put AI-driven automation at about 16% of marketing work in 2026. Gartner’s November 2025 guidance was that agents fail without people setting strategy and checking output. Evidence confidence is high.

  17. Partial Gaming

    AI embeds itself in gaming and revolutionizes gameplay and how games are made, marketed and monetized, with predictive analytics, automated user acquisition and personalized in-game experiences becoming standard toolsets

    In the post
    “AI will embed itself into gaming, revolutionizing gameplay and how games are made, marketed, and monetized.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    By 31 Dec 2025, (a) AI-driven automated user acquisition and predictive targeting are the standard way mobile games buy players, shown by the main mobile ad networks running on AI engines, (b) AI reshapes game production, shown by a majority of game developers using generative AI in their own production work in a large independent industry survey, and (c) AI reshapes gameplay, shown by personalized or AI-driven in-game experiences becoming a standard feature of widely played games. Correct if all hold; partial if one or two do.
    Outcome
    The marketing side came true: AppLovin’s revenue rose 70% to $5.5 billion in 2025, driven mainly by its Axon AI ad engine, and Unity moved its whole ad network onto the AI platform Vector by Q2 2025. Production changed less, since GDC’s surveys found 36% of developers personally using generative AI in both 2025 and 2026, only 30% at studios by the 2026 survey, mostly for research, email and code help, and 52% now calling it harmful. A vendor-sponsored Google Cloud poll of 615 developers put AI use at 90%, so production evidence is mixed and overall confidence is moderate. Gameplay changed least: AI characters reached widely played games only as one-offs, such as the conversational Darth Vader that Epic added to Fortnite on 16 May 2025, so leg (c) fails.

  18. Correct Crypto

    Base keeps eating into Layer 2 market share through 2025 and becomes a key player, helped by net inflows and its Coinbase integration

    In the post
    “Coinbase’s Base network will continue eating into L2 market share, fueled by consistent net inflows (+$2.4B over three months).”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    Base’s share of Layer 2 DeFi TVL (DefiLlama data, or Messari or The Block citing it) is higher at the end of 2025 than at the start of 2025, and Base ranks first or second among Layer 2 networks by DeFi TVL at the end of 2025.
    Outcome
    Base passed Arbitrum as the largest Layer 2 by DeFi TVL in January 2025, and its share of Layer 2 DeFi TVL rose from about 33% at the start of the year to about 46% by December (Messari). Messari also put Base at 62% of Layer 2 revenue for 2025 to date, and The Block found it took roughly half of Layer 2 DEX volume through the year. The evidence is strong.

  19. Partial Crypto

    DeFi protocols with better risk management, friendlier interfaces and institutional-grade products emerge in 2025 and revive interest in DeFi

    In the post
    “Enhanced decentralized finance protocols will emerge with better risk management, user-friendly interfaces, and institutional-grade products, reviving interest in DeFi.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    During 2025, (a) major DeFi protocols launch institutional-grade products (for example permissioned markets for institutions, real-world-asset collateral, or DeFi lending offered inside a regulated exchange app), and (b) total DeFi TVL on DefiLlama is higher on 31 Dec 2025 than on the publication date (about $125B on 6 Jan 2025). Correct if both hold; partial if only one does.
    Outcome
    Institutional-grade DeFi launched in 2025: Coinbase began offering bitcoin-backed USDC loans through Morpho on Base in January, and Aave Labs opened Horizon in August for institutions borrowing stablecoins against tokenized treasuries and funds. Total DeFi TVL on DefiLlama climbed to $171B on 7 October 2025, the highest since December 2021, then fell with token prices to about $113B on 31 December, below the roughly $125B at publication. Interest revived for most of the year and faded by December, so the call is partial on strong evidence.

  20. Incorrect Crypto

    AI is the most important part of the 2025 DeFi revival

    In the post
    “Most importantly, though, with AI.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    Year-end 2025 DeFi reviews (such as DL Research’s State of DeFi 2025 or CoinGecko’s annual narrative study) name AI or AI agents as a leading driver of DeFi’s 2025 growth, or AI-native DeFi protocols (agent-run trading, yield or lending) rank among the 50 largest DeFi protocols by TVL.
    Outcome
    AI-agent tokens, the DeFAI projects among them, crashed early: the sector’s market value fell from more than $10B to about $6.6B by 7 February 2025, and CoinGecko’s 2025 study put the AI narrative’s average return at -50% for the year. DeFi’s 2025 growth came from stablecoins, real-world assets and lending, the themes DL Research’s State of DeFi 2025 names, and AI is not among them. DefiLlama’s AI Agents category held under $40M of TVL when checked in September 2026, tiny next to the largest lending protocols. The evidence is fairly strong.

  21. Partial Crypto

    Interoperability protocols such as Cosmos and LayerZero gain greater adoption in 2025, easing cross-chain fragmentation and improving liquidity flow between chains

    In the post
    “As multi-chain ecosystems grow, interoperability solutions like Cosmos and LayerZero will gain greater adoption, solving fragmentation issues and improving liquidity flow between chains.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    (a) Interoperability protocols, including the named examples LayerZero and Cosmos IBC, carry clearly more cross-chain activity in 2025 than in 2024 (value moved, messages, or assets issued through them), and (b) cross-chain liquidity fragmentation is widely reported as reduced by the end of 2025. Correct if both hold; partial if only one does.
    Outcome
    USDT0, Tether’s cross-chain token built on LayerZero’s OFT standard, launched in January 2025 and passed $50B in cumulative transfers by November, and LayerZero bought the Stargate bridge for $110M in August. Cosmos lagged even after IBC Eureka linked it to Ethereum in April: ATOM fell to about $2 by December, a five-year low, and Cosmos Labs said Interchain Security had failed to find product-market fit. Fragmentation also persisted, with the Ethereum Foundation proposing an Interop Layer in November to fix fragmented liquidity and bridging across Layer 2s. Graded partial, with strong evidence of growth for LayerZero and weak evidence for Cosmos.

  22. Correct Crypto

    Tokenization of copyrights, patents and creative works emerges in 2025, enabling new funding models and fractional ownership for creators

    In the post
    “The tokenization of copyrights, patents, and creative works will emerge, enabling new funding models and fractional ownership for creators and innovators.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    During 2025, at least one platform built for registering, licensing or tokenizing intellectual property goes live on mainnet, and at least one set of copyrights, creative works (music, film, art) or patents is sold to the public as fractional tokens that raise money or pass through royalties.
    Outcome
    Story, an IP-focused blockchain backed by $140M in venture funding, launched its mainnet and IP token on 13 February 2025 so creators can register, license and get paid for their IP. The same month Aria raised $10.95M to buy partial rights to songs by Justin Bieber, Miley Cyrus, BLACKPINK and BTS and issued APL, a token representing those royalties, then added a $15M funding round in September. Both are live versions of the funding and fractional-ownership models the post described, though tokenized music deals like Aria’s stayed in the tens of millions of dollars. Evidence is strong that tokenized IP emerged and thin that it reached wide adoption.

  23. Partial Gaming

    AI-driven marketing dominates gaming go-to-market in 2025, with developers using AI tools to hyper-target audiences, refine messaging and optimize launch campaigns

    In the post
    “AI-based marketing will revolutionize gaming’s go-to-market strategies.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    Correct if 2025 evidence shows AI-driven tools became the main way games reach players across the industry: AI targeting and campaign-optimization platforms carry most game user-acquisition spend, and industry surveys show most game marketing and publishing staff using AI tools for targeting, messaging or campaigns. Incorrect if AI stayed a minor or experimental part of game marketing. Partial if the evidence falls between the two.
    Outcome
    AI ad targeting became the engine of mobile game user acquisition: AppLovin’s 2025 revenue rose 70% to about $5.5B, which it credits mainly to its Axon AI ad engine, and Unity rebuilt its ad network around its AI platform Vector. Beyond paid mobile UA the shift was shallower; GDC’s survey published in January 2025 found 39% of community, marketing and PR staff using generative AI, and its 2026 survey found 58% at publishing and marketing firms, mostly for research and email. Evidence is moderate: AI clearly reshaped mobile UA, but nothing shows it dominating go-to-market for PC and console games.

  24. Incorrect Gaming

    Play-to-earn evolves into a fairer, more sustainable second version that expands globally, with the next wave targeting Tier 2 geographies before entering Tier 1 markets

    In the post
    “The next wave will target Tier 2 geographies before entering Tier 1 markets.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    Correct if during 2025 a new wave of play-to-earn or blockchain games with sustainable economies grows its player base, launching or scaling first in Tier 2 markets (Southeast Asia, Latin America, South Asia and similar) and then moving into Tier 1 markets. Incorrect if blockchain gaming activity shrinks over the year or no such expansion appears.
    Outcome
    Blockchain gaming contracted through 2025 instead of expanding. CoinDesk, citing DappRadar, reported daily active wallets down 17% in Q2, funding down 93% year on year to $73M, and more than 300 web3 games shut down; Decrypt’s year-end review called 2025 the year crypto gaming collapsed as funding dried up. The Tier 2 first pattern did show up in the year’s biggest launch, Nexon’s MapleStory N, which went live in May 2025 while blocked in the US, Japan, South Korea and most of Europe, but that was a regulatory choice inside a shrinking market. Evidence is strong.

  25. Incorrect Crypto

    A new wave of blockchain-enabled wearables and devices arrives that protects users’ health data while tracking health and fitness, potentially influencing insurance premiums

    In the post
    “I anticipate a new wave of blockchain-enabled wearables and devices that protect your data while monitoring health metrics and fitness achievements, potentially influencing insurance premiums.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    Correct if during 2025 a visible wave of blockchain-enabled health wearables appears: several such devices ship and at least one reaches mainstream scale (hundreds of thousands of units sold), or a major wearable brand or insurer adopts blockchain-secured wearable data. A handful of niche crypto-native devices does not count.
    Outcome
    No wave arrived. The most visible blockchain wearable, the Solana-based CUDIS ring, reported just over 20,000 rings sold by May 2025 before launching its token in June, while Oura, which uses no blockchain, passed 5.5 million rings sold and projected $1B in 2025 revenue. There is no public evidence that a major wearable maker or insurer adopted blockchain-secured health data during the year. Evidence is strong for the mainstream players; small crypto-native devices are harder to count but stayed niche.

  26. Partial AI

    AI moves beyond research labs to become a staple of everyday healthcare, with predictive algorithms combining genetic, lifestyle and medical-history data for hyper-targeted diagnostics and treatment plans

    In the post
    “Artificial intelligence will move beyond research labs to become a staple in everyday healthcare.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    Correct if by end of 2025 AI is in routine clinical use (a majority of physicians report using AI in practice and major health systems deploy AI tools) and AI-driven personalization that combines genetic, lifestyle and medical-history data is part of routine diagnosis or treatment planning. Incorrect if AI in healthcare stays mostly in research and pilots. Partial if the evidence falls between the two.
    Outcome
    The first half happened: 81% of US physicians told the AMA in early 2026 that they use AI in practice, up from 38% in 2023, and Menlo Ventures found 27% of health systems had deployed domain-specific AI tools in 2025, with ambient scribes the largest category. The personalized-medicine half lagged: the leading physician uses were literature search and documentation, and genomics-driven care grew mainly inside oncology and hereditary testing, where Tempus’s genomics revenue rose 115% in Q2 2025 (helped by its Ambry Genetics acquisition). Evidence is strong for adoption and moderate for how little of it was genetic personalization.

  27. Incorrect Crypto

    As biohacking goes mainstream, decentralized platforms let individuals control and possibly monetize their health information, reshaping the wellness market toward consumers

    In the post
    “decentralized platforms will empower individuals to control and possibly monetize their health information.”
    Made
    6 Jan 2025
    Resolves by
    31 Dec 2025
    Confidence
    Not stated
    Resolves when
    Correct if during 2025 decentralized platforms that let people hold, share or sell their own health data become a meaningful part of consumer health, for example one reaching millions of users or a mainstream health or wellness company adopting user-owned, decentralized health data. Incorrect if such platforms stay niche.
    Outcome
    Consumer-directed health grew fast in 2025, but through centralized companies: Function Health raised $298M at a $2.5B valuation after more than 50 million lab tests. Decentralized health-data platforms stayed niche, with CUDIS, among the most visible, reporting about 200,000 app users. The year’s biggest consumer-data event cut the other way, as 23andMe’s genetic data on about 15 million customers was sold in bankruptcy for $305M. Evidence is moderate, since small decentralized projects are hard to count, but none reached mainstream scale.

  28. Pending AI

    AI goes local: desktop agents that run directly on the operating system become the AI story of 2026, displacing browser-based chat as the main way people work with AI

    In the post
    “The browser-based AI paradigm is about to get obliterated by local desktop agents that run directly on your operating system.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if both hold by 31 Dec 2026. First, year-end reviews in at least two major outlets (such as The Verge, Wired, Bloomberg, the FT or the New York Times) name desktop agents as the defining AI story of 2026 or one of its top stories; a desktop agent here has direct operating-system access to files, local apps and processes, outside a browser sandbox. Second, at least one desktop agent product reports 1M+ active users. Resolves incorrect if browser and chat-window assistants remain the main story and desktop agents stay niche.

  29. Pending AI

    Agentic teams become ubiquitous: by Q4 2026, at least 50% of software companies with 50+ employees have some form of agentic AI team deployed in production workflows, past the pilot stage

    In the post
    “by Q4 2026, at least 50% of software companies with 50+ employees will have some form of agentic AI team deployed in production workflows.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if a credible survey or analyst report published by 31 Dec 2026 finds that at least 50% of software companies with 50+ employees, or the closest segment it reports, run AI agents on multi-step work in production workflows past the pilot stage. Credible means a sample of 500+ companies or a major firm such as Gartner, McKinsey, IDC or Forrester. Chat assistants and code completion alone don’t count. Resolves incorrect if the best available survey puts the share below 50%.

  30. Pending AI

    The cognitive costs of heavy AI usage (context switching, reduced deep thinking, attention fragmentation, erosion of offloaded skills) surface in real research and mainstream discourse

    In the post
    “We’re going to start seeing real research and real discourse around the cognitive downsides of heavy AI usage.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    78%, assigned March 2026
    Resolves when
    Resolves correct if in 2026 at least one peer-reviewed study on the cognitive downsides of AI use (reduced deep thinking, attention, skill loss) is published in a major journal and covered by mainstream outlets. Resolves incorrect if no such study appears in 2026.

  31. Pending AI

    “Proof of Human” emerges as a real category: verified human-created content, authentication protocols for original work, and premium pricing for demonstrably non-AI output

    In the post
    “‘Proof of Human’ will emerge as a real category.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    82%, assigned March 2026
    Resolves when
    Resolves correct if in 2026 at least one major platform, marketplace or industry body in music, art, writing or photography launches a verified human-made label, certification or authentication system, and it sees commercial uptake (certified works on sale, paying members, or a reported price premium). Resolves incorrect if no such launch happens in 2026.

  32. Pending Markets

    Prediction markets integrate with daily life: at least one major news outlet (top 10 by traffic) natively integrates prediction market data into its editorial content by December 2026

    In the post
    “at least one major news outlet (top 10 by traffic) will natively integrate prediction market data into their editorial content by December 2026.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if by 31 Dec 2026 a news outlet in the top 10 by traffic (Similarweb or Comscore news rankings, US or global) shows live odds from Polymarket, Kalshi or a similar market inside its editorial content through an official partnership or product feature. Article modules, live pages and on-air graphics all count; quoting odds in a story now and then doesn’t. Resolves incorrect if no top-10 outlet does this by 31 Dec 2026. Only integrations announced between 3 April 2026, when the post went out, and 31 Dec 2026 count; CNN’s and CNBC’s Kalshi partnerships, announced on 2 and 4 December 2025, don’t.

  33. Pending Crypto

    X402 (HTTP-native payments) goes from interesting experiment to meaningful transaction volume in 2026, setting the stage for a bigger story in 2027 and 2028

    In the post
    “But I think it goes from ‘interesting experiment’ to ‘meaningful transaction volume’ and sets the stage for a much bigger story in 2027-2028.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if x402 payments settled during calendar 2026 total at least $100 million across all chains, counting agent and API payments and excluding bridge flows, other non-payment contract activity and anything the data provider flags as wash trading or self-dealing, per public on-chain data (x402scan, Dune, Bitquery or Artemis) or Coinbase or x402 Foundation reporting. Resolves incorrect below $100 million.

  34. Pending Crypto

    Omni banks rise as traditional finance bridges to blockchain: total real (adjusted) stablecoin transaction volume exceeds $20 trillion in 2026

    In the post
    “The specific call: total stablecoin real transaction volume will exceed $20 trillion in 2026.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if adjusted stablecoin transaction volume for calendar 2026 exceeds $20 trillion on the Visa Onchain Analytics dashboard (Visa and Allium), which strips out bot activity and other inorganic transfers; if that series is unavailable, use another published adjusted-volume series such as Artemis. Raw on-chain volume doesn’t count, because the post specifies real volume. Resolves incorrect at or below $20 trillion.

  35. Pending Gaming

    Arc Raiders remains a top-5 extraction shooter through 2026

    In the post
    “Arc Raiders remains a top-5 extraction shooter through 2026”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if Arc Raiders ranks among the five extraction shooters with the highest average concurrent Steam players (SteamCharts monthly averages) in every month of 2026. An extraction shooter is a game whose main mode is PvP or PvPvE extraction, such as Escape from Tarkov, Hunt: Showdown, Marathon, Arena Breakout: Infinite, Dark and Darker or Gray Zone Warfare. Resolves incorrect if it falls outside the top five in any month.

  36. Pending Health

    At least one AI-discovered drug enters Phase 3 clinical trials in 2026

    In the post
    “At least one AI-discovered drug enters Phase 3 clinical trials in 2026”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if during 2026 a drug candidate whose target or molecule was identified mainly through AI or machine learning, as stated by its developer, starts a Phase 3 trial (first patient dosed, or a Phase 3 trial registered as recruiting on ClinicalTrials.gov or an equivalent registry). Candidates that entered Phase 3 before 2026 don’t count. Resolves incorrect if none does.

  37. Pending AI

    Orbital data centers stay a narrative in 2026: prototypes and announcements, but no meaningful compute infrastructure in orbit

    In the post
    “Putting meaningful compute infrastructure in orbit is a few years away.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if by 31 Dec 2026 no company is selling compute from orbit to outside customers at data-center scale, meaning a dedicated constellation or station running production workloads rather than a demonstration satellite carrying a few GPUs. Prototypes, demo missions, test launches, announcements and licence filings don’t count against the call. Resolves incorrect if an orbital data center is in commercial service in 2026.

  38. Pending AI

    ASI discourse reaches fever pitch in late 2026, but none of it produces artificial superintelligence this year

    In the post
    “Just not to the point of ASI in 2026.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if no AI system released or announced in 2026 is broadly accepted, by independent evaluators and not only its maker, as artificial superintelligence, which the post defines as intelligence that exceeds all human capability across all domains. Beating humans on some benchmarks or tasks doesn’t count. Resolves incorrect if such a system is accepted as ASI in 2026.

  39. Pending AI

    The US Senate holds hearings on artificial superintelligence in 2026 as the ASI debate escalates

    In the post
    “The Senate will hold hearings.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if during 2026 a US Senate committee or subcommittee holds at least one hearing on superintelligence: “superintelligence” or “ASI” appears in the hearing title or official description, or its witnesses are called chiefly to testify on superintelligence risk. A general AI hearing where superintelligence comes up in passing doesn’t count. Resolves incorrect if no such hearing is held in 2026.

  40. Pending Markets

    The AI energy deficit is everywhere in 2026 but doesn’t close: an energy buildout big enough for AI’s appetite stays a late-2020s story

    In the post
    “The constraint will drive innovation at the margins, but the macro energy gap is not closing in 2026.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if NERC’s 2026 Long-Term Reliability Assessment (normally published in December), or the most recent NERC, IEA or PJM assessment available on 31 Dec 2026, still reports elevated or high risk of power shortfalls tied to data-center load growth in at least one major US region. Resolves incorrect if those assessments report that supply has caught up with AI-driven demand.

  41. Pending AI

    Robotics doesn’t go mainstream in 2026: impressive demos and limited commercial deployments in warehouses and manufacturing, with consumer and general-purpose humanoids in daily life no earlier than 2028

    In the post
    “Consumer and general-purpose humanoid robots in your daily life are for 2028 at the absolute earliest.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2027
    Confidence
    Not stated
    Resolves when
    Resolves correct if before 1 January 2028 no general-purpose humanoid robot reaches everyday consumer life, meaning none is sold through general retail (not a waitlist, pre-order or early-adopter programme) with more than 100,000 units delivered to homes. Warehouse and factory deployments, pilots and demos don’t count against the call. Resolves incorrect if a humanoid robot crosses that bar before 2028.

  42. Pending Markets

    A shockingly large arrest hits prediction markets in 2026: someone is arrested for trading on inside information on a market

    In the post
    “we’ll have a shockingly large arrest from someone wetting their beak from inside information on a market”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if in 2026 law enforcement arrests or criminally charges someone for trading on a prediction market (Polymarket, Kalshi or similar) using inside information, in a case that draws national news coverage for its size or the defendant’s position. A platform’s own fine or account ban doesn’t count. Resolves incorrect if no such arrest or charge happens in 2026. Cases made public before the post went out on 3 April 2026, such as Israel’s 12 February 2026 indictment of an air force reservist and a civilian over Polymarket bets placed with classified information, don’t count.

  43. Not scored Gaming

    Arc Raiders faces a meaningful retention dip in the early part of 2026

    In the post
    “faces a meaningful retention dip in the early parts of this year”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if Arc Raiders’ average concurrent Steam players (SteamCharts monthly averages) for any month from February to June 2026 fall at least 25% below the January 2026 average. Resolves incorrect if every month through June stays within 25% of January.
    Outcome
    Not scored. The post went out on 3 April 2026, when SteamCharts already showed February’s average player count (171,037) 29% below January’s (241,197). A draft from 30 March had placed the dip in the back half of the year.

  44. Pending Gaming

    Game industry layoffs in 2026 outpace 2024, the industry’s peak year, which would mean more than 15,000 jobs cut

    In the post
    “I predict layoff rounds in 2026 outpace totals we saw in 2024, the peak year in the industry.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if tracked video game industry layoffs for calendar 2026 exceed 15,000 and exceed the 2024 total from the same tracker (the Game Industry Layoffs tracker or gaminglayoffs.com, as cited by GamesIndustry.biz or Game Developer). Resolves incorrect if 2026 comes in at or below either bar.

  45. Pending Gaming

    Indie and small-studio titles continue to dominate critical reception in 2026 and increasingly compete on commercial metrics

    In the post
    “Simultaneously, indie and small-studio titles will continue to dominate critical reception and increasingly compete on commercial metrics.”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if indie or small-studio games (from developers not owned by a major publisher or platform holder) make up at least 11 of the 20 highest-rated games released in 2026 on Metacritic, the measure the post uses for 2025. Resolves incorrect at 10 or fewer. The commercial half of the sentence is too loose to score and is left out.

  46. Pending Gaming

    GTA 6 ships in 2026 with blockchain integration in its in-game economy (the post calls this a low-conviction call)

    In the post
    “GTA 6 Ships with Blockchain Integration”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if Grand Theft Auto VI is released by 31 Dec 2026 with blockchain rails in its game or online economy (on-chain items, tokens or crypto payments), confirmed by Rockstar or Take-Two. Resolves incorrect if it ships without them or doesn’t ship in 2026, the standard the post applied to last year’s version of this call.

  47. Pending Health

    The longevity economy (supplements, testing, optimization services) exceeds $50 billion in global market size in 2026

    In the post
    “The ‘longevity economy’ (supplements, testing, optimization services) exceeds $50 billion in market size globally”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves on the median of published estimates of the 2026 global longevity market (supplements, testing and optimization services) from market research firms: correct above $50 billion, incorrect at or below.

  48. Pending Health

    Wearable health tech adoption crosses 40% of adults in all Tier 1 economies in 2026

    In the post
    “Wearable health tech adoption crosses 40% of adults in all Tier 1 economies”
    Made
    3 Apr 2026
    Resolves by
    31 Dec 2026
    Confidence
    Not stated
    Resolves when
    Resolves correct if survey data from 2026 shows more than 40% of adults using a wearable health device (smartwatch, fitness tracker, smart ring or continuous glucose monitor) in every G7 economy: the US, UK, Canada, Japan, Germany, France and Italy. One multi-country survey or several national surveys may be used. Resolves incorrect if any G7 country is at or below 40%.

Methodology

Each entry is a call from one of my predictions articles. The large type is my summary of it, and “In the post” quotes the line it comes from, word for word. The site checks every quote against its article when it builds. I left out lines with no date or measure, and calls that hang on a condition that never came true. The 2026 article went out on 3 April 2026. Events from before that day don’t count toward its calls, and a call whose wording changed before publication to match what had already happened isn’t scored.

I score a call correct or incorrect where I can, and partial when the core of it happened but the details didn’t. Accuracy is (correct + 0.5 × partial) ÷ resolved, over every resolved call.

The articles gave no percentages. I assigned confidence levels to the original twelve 2025 calls in March 2026, after the year was over and about half of them had resolved, so the 2025 Brier score is a retrospective estimate. Nine 2026 calls got a confidence level that month too, before the article went out. Two of them still carry it; the other seven changed when I rewrote them to match the article, so they carry none, like the calls I added later. They count toward accuracy and stay out of the Brier score.

Brier score measures calibration: whether the confidence levels were honest, which matters as much as whether the calls were right. Formula: mean of (confidence − outcome)² across resolved predictions that carry a confidence, where correct=1, partial=0.5, incorrect=0. Lower is better: 0.00 is perfect, 0.25 is a random coin flip, 1.00 is perfectly wrong. The skill score shows percentage improvement over a random baseline (0.25).

I wrote the resolution criteria after the articles too. The first set dates from March 2026. In September 2026 I rebuilt the ledger against both articles: I added the calls it had missed, graded the new 2025 ones with sources linked under each, and rewrote the 2026 criteria to match what the article says. From my next predictions article on, confidence and criteria are fixed before it publishes. Every wrong call stays on the record.