| $20M | “twenty million dollars” | Prince: Fixed-Sum AI Licensing Deals Are a Mistake for Content Creators | Matthew Prince | Aug 7, 2025 | — |
| 15T | “15 trillion” | Prince: Cloudflare processes roughly 15 trillion requests per day | Matthew Prince | Aug 7, 2025 | Partly supported |
| $10M | “ten million dollars” | Chen: Many YC Founders Only Care About Fundraising Brags and Headlines | Edwin Chen | Jul 24, 2025 | — |
| 2% | “two percent” | Chen: Early Startups Should Not Hire Data Scientists as First Employees | Edwin Chen | Jul 24, 2025 | — |
| 5% | “five percent” | Chen: Early Startups Should Not Hire Data Scientists as First Employees | Edwin Chen | Jul 24, 2025 | — |
| 10× | “10 X” | Chen: Early Startups Should Not Hire Data Scientists as First Employees | Edwin Chen | Jul 24, 2025 | — |
| 80% | “80%” | Laskin: Engineers spend 80% of their time comprehending complex systems | Misha Laskin | Jul 17, 2025 | — |
| 20% | “20%” | Laskin: Teaching AI agents to take action is mostly solved | Misha Laskin | Jul 17, 2025 | — |
| 10% | “10%” | Laskin: Deploying superintelligence and reaching 10% GDP growth is multi-decade | Misha Laskin | Jul 17, 2025 | — |
| 80% | “80%” | Conrad: Over 80% of Rippling's engineering team focuses on existing products | Parker Conrad | Jul 10, 2025 | — |
| 1K | “a thousand” | Conrad: Rippling employs over 1,000 engineers, staffing new products with 5–7 people | Parker Conrad | Jul 10, 2025 | — |
| 20% | “20%” | Dent: Chai-2 hits nearly 20% wet-lab antibody binding success | Jack Dent | Jul 3, 2025 | Supported |
| 1% | “one percent” | Dent: Chai Discovery originally targeted only a 1% antibody success rate | Jack Dent | Jul 3, 2025 | — |
| 70% | “70%” | Chai filtered benchmark targets to under 70% identity from training data | Joshua Meier | Jul 3, 2025 | — |
| 25% | “25%” | Meier: Chai-2 maintains success rate on targets with 25% sequence similarity | Joshua Meier | Jul 3, 2025 | Supported |
| 50% | “50%” | Meier: Chai-2 achieved a 50% success rate on tested targets | Joshua Meier | Jul 3, 2025 | Supported |
| 70% | “70%” | Dent: Chai achieved 70% success rate on mini protein binder designs | Jack Dent | Jul 3, 2025 | Supported |
| 1K | “a thousand” | Gil: Applied Intuition Is Profitable and Has Not Spent Raised Capital | Elad Gil | Jun 17, 2025 | — |
| 16B | “a fifteen billion” | Younis: Applied Intuition is a profitable $15B AI company | Qasar Younis | Jun 17, 2025 | Supported |
| 15B | “fifteen billion” | Younis: Applied Intuition is a profitable $15B AI company | Qasar Younis | Jun 17, 2025 | Supported |
| 82% | “82%” | Younis: Applied Intuition is approximately 82% software engineers | Qasar Younis | Jun 17, 2025 | — |
| 50% | “50%” | Ben Mann defines transformative AI by the 'Economic Turing Test' | Ben Mann | Jun 12, 2025 | — |
| 51B | “a fifty billion” | Gil: No $50B+ biotech started since the 1980s, excluding Moderna | Elad Gil | May 29, 2025 | Contradicted |
| $40M | “forty million dollars” | Gil: Biotech VC incubations are designed to flip to pharma pipelines | Elad Gil | May 29, 2025 | — |
| 40% | “40%” | Gil: Biotech VC incubations are designed to flip to pharma pipelines | Elad Gil | May 29, 2025 | — |
| 1B | “one billion” | Jain: One of Glean's largest customers has over 1B documents | Arvind Jain | May 15, 2025 | — |
| 1B | “one billion” | Jain: The entire internet was one billion documents in 2004 | Arvind Jain | May 15, 2025 | Contradicted |
| 1B | “a billion” | Jain: Enterprise search requires transformer models due to scarce user signals | Arvind Jain | May 15, 2025 | — |
| 90% | “90%” | Jain: About 90% of enterprise knowledge is private within a company | Arvind Jain | May 15, 2025 | — |
| 90% | “90%” | Jain: Workers will have personal AI teams performing 90% of their work | Arvind Jain | May 15, 2025 | — |
| 2B | “two billion” | Von Ahn: About two billion people worldwide are learning English | Luis von Ahn | May 8, 2025 | Supported |
| 10M | “ten million” | Duolingo has 10 million active users with 365+ day streaks | Luis von Ahn | May 8, 2025 | Supported |
| 83% | “83%” | Giving users an 83% success rate on exercises maximizes learner enjoyment | Luis von Ahn | May 8, 2025 | — |
| 90% | “90%” | McKinzie: Over 90% of internet clock images show 10:10, biasing vision models | Brandon McKinzie | May 1, 2025 | — |
| 15% | “15%” | Fulford: RFT is only worthwhile for out-of-distribution or make-or-break tasks | Isa Fulford | Apr 24, 2025 | — |
| 1K | “a thousand” | Goldman: Mineral exploration yields 1,000x returns on a few million dollars | Josh Goldman | Apr 17, 2025 | — |
| 1K | “a thousand” | Goldman: Mingomba is the highest-grade undeveloped large copper deposit globally | Josh Goldman | Apr 17, 2025 | Open |
| 0.6% | “0.6%” | Goldman: Mingomba is the highest-grade undeveloped large copper deposit globally | Josh Goldman | Apr 17, 2025 | Open |
| 5% | “five percent” | Goldman: Mingomba is the highest-grade undeveloped large copper deposit globally | Josh Goldman | Apr 17, 2025 | Open |
| 5% | “five percent” | Goldman: 10x ore grade reduces rock processing and capital intensity tenfold | Josh Goldman | Apr 17, 2025 | — |
| 0.5% | “0.5%” | Goldman: 10x ore grade reduces rock processing and capital intensity tenfold | Josh Goldman | Apr 17, 2025 | — |
| 10× | “10 times” | Goldman: 10x ore grade reduces rock processing and capital intensity tenfold | Josh Goldman | Apr 17, 2025 | — |
| $1B | “a billion dollars” | Goldman: $1B in exploration yields less than one deposit discovery today | Josh Goldman | Apr 17, 2025 | Supported |
| 1B | “a billion” | Goldman: $1B in exploration yields less than one deposit discovery today | Josh Goldman | Apr 17, 2025 | Supported |
| $100M | “a hundred million dollars” | Goldman: KoBold targets $50M to $100M cost per mineral discovery | Josh Goldman | Apr 17, 2025 | — |
| 27% | “27%” | Rao: Two in five doctors and 27% of nurses want to quit | Dr. Shiv Rao | Mar 27, 2025 | Contradicted |
| $1M | “a million dollars” | Rao: Replacing a Clinician Costs Health Systems Close to $1 Million | Dr. Shiv Rao | Mar 27, 2025 | Supported |
| 60% | “60%” | Rao: Stanford survey shows Abridge cuts clinician burnout by 50% | Dr. Shiv Rao | Mar 27, 2025 | Supported |
| 50% | “50%” | Rao: Stanford survey shows Abridge cuts clinician burnout by 50% | Dr. Shiv Rao | Mar 27, 2025 | Supported |
| $500M | “five hundred million dollars” | Rao: Abridge has raised over $500 million | Dr. Shiv Rao | Mar 27, 2025 | Partly supported |
| 80% | “80%” | Rao: Abridge plans to allocate 80% of funding to R&D | Dr. Shiv Rao | Mar 27, 2025 | — |
| 100% | “hundred percent” | Dohmke: Improved model reasoning will push SWE-bench scores near 100% | Thomas Dohmke | Mar 13, 2025 | — |
| 25% | “25%” | Dohmke: GitHub Copilot preview generated 25% of code early on | Thomas Dohmke | Mar 13, 2025 | Supported |
| 1.8M | “1.8 million” | Dohmke: Copilot reached 77,000 organizations and 1.8M paid users | Thomas Dohmke | Mar 13, 2025 | Supported |
| 28% | “28%” | Dohmke: Copilot yields 25-28% end-to-end developer productivity gains | Thomas Dohmke | Mar 13, 2025 | Supported |
| 55% | “55%” | Dohmke: Copilot yields 25-28% end-to-end developer productivity gains | Thomas Dohmke | Mar 13, 2025 | Supported |
| 30% | “30%” | Hendrycks: Over 30% of Staff at Some Top AI Labs Are Chinese Nationals | Dan Hendrycks | Mar 5, 2025 | — |
| 50M | “fifty million” | Alidoust: Prior public single-cell data totaled only 45M to 60M cells | Nima Alidoust | Feb 25, 2025 | Supported |
| 60M | “sixty million” | Alidoust: Prior public single-cell data totaled only 45M to 60M cells | Nima Alidoust | Feb 25, 2025 | Supported |
| 60M | “sixty million” | Alidoust: Single-cell AI performance holds even after 99% data downsampling | Nima Alidoust | Feb 25, 2025 | — |
| 99% | “99%” | Alidoust: Single-cell AI performance holds even after 99% data downsampling | Nima Alidoust | Feb 25, 2025 | — |
| 1% | “one percent” | Alidoust: Single-cell AI performance holds even after 99% data downsampling | Nima Alidoust | Feb 25, 2025 | — |
| 2M | “two million” | Alidoust: Tahoe dataset expands public perturbational single-cell data fiftyfold | Nima Alidoust | Feb 25, 2025 | Partly supported |
| 100M | “a hundred million” | Alidoust: Tahoe dataset expands public perturbational single-cell data fiftyfold | Nima Alidoust | Feb 25, 2025 | Partly supported |
| 100M | “a hundred million” | Alidoust: 100 million single-cell data points equate to 200-300 billion tokens | Nima Alidoust | Feb 25, 2025 | — |
| 300B | “three hundred billion” | Alidoust: 100 million single-cell data points equate to 200-300 billion tokens | Nima Alidoust | Feb 25, 2025 | — |
| 30M | “thirty million” | Burke: Arc Institute built SC Basecamp with 230M single cells | Dave Burke | Feb 25, 2025 | — |
| 100M | “hundred million” | Burke: Arc Institute built SC Basecamp with 230M single cells | Dave Burke | Feb 25, 2025 | — |
| 10% | “10%” | Burke: Current virtual cell models have only around 10% predictive accuracy | Dave Burke | Feb 25, 2025 | — |
| 9.3T | “9.3 trillion” | Burke: Evo 2 trained on 9.3 trillion nucleotides predicts BRCA1 pathogenicity | Dave Burke | Feb 25, 2025 | Supported |
| 50% | “50%” | Vogt: AV labor costs hit diminishing returns at 25% remote assistance | Kyle Vogt | Feb 20, 2025 | — |
| 25% | “25%” | Vogt: AV labor costs hit diminishing returns at 25% remote assistance | Kyle Vogt | Feb 20, 2025 | — |
| 75% | “75%” | Vogt: AV labor costs hit diminishing returns at 25% remote assistance | Kyle Vogt | Feb 20, 2025 | — |
| 10× | “10 times” | Vogt: Robotaxis have no MVP without superhuman safety performance | Kyle Vogt | Feb 20, 2025 | — |
| 80% | “80%” | Vogt: Autonomous vehicle skepticism dropped from ~80% to ~20% after one ride | Kyle Vogt | Feb 20, 2025 | — |
| 30% | “30%” | Vogt: Autonomous vehicle skepticism dropped from ~80% to ~20% after one ride | Kyle Vogt | Feb 20, 2025 | — |
| $500M | “five hundred million dollars” | Guo: Harvey has raised over $500M from OpenAI, Sequoia, Kleiner, and others | Sarah Guo | Feb 14, 2025 | Supported |
| 50M | “fifty million” | Guo: Harvey serves over 250 clients and generates $50M+ ARR | Sarah Guo | Feb 14, 2025 | Partly supported |
| 20% | “20%” | Gil: 20% drop in NVIDIA stock after DeepSeek was unwarranted | Elad Gil | Feb 7, 2025 | — |
| 180% | “180%” | Gil: GPT-4 equivalent inference costs dropped 180x over 18 months | Elad Gil | Feb 7, 2025 | Supported |
| 180× | “180 X” | Gil: GPT-4 equivalent inference costs dropped 180x over 18 months | Elad Gil | Feb 7, 2025 | Supported |
| 98% | “98%” | Clerico: 98% of Climate Capital Goes to Mitigation, Only 2% to Resilience | Bill Clerico | Jan 23, 2025 | Partly supported |
| 2% | “two percent” | Clerico: 98% of Climate Capital Goes to Mitigation, Only 2% to Resilience | Bill Clerico | Jan 23, 2025 | Partly supported |
| 6× | “six times” | Clerico: Permitting conservation logging on 170-acre parcel took three years | Bill Clerico | Jan 23, 2025 | — |
| $250M | “two hundred fifty million dollars” | Clerico: U.S. Forest Service spends 40% of budget, $250M, on planning | Bill Clerico | Jan 23, 2025 | Partly supported |
| 40% | “40%” | Clerico: U.S. Forest Service spends 40% of budget, $250M, on planning | Bill Clerico | Jan 23, 2025 | Partly supported |
| 3M | “three million” | Clerico: Florida conducts 2 to 3 million acres of prescribed burns annually | Bill Clerico | Jan 23, 2025 | Supported |
| 46% | “46%” | Clerico: WUI housing units have increased by ~46% since 1980 | Bill Clerico | Jan 23, 2025 | Partly supported |
| 1M | “one million” | Clerico: Palisades Water Failed From Plumbing Depressurization, Not Supply Shortage | Bill Clerico | Jan 23, 2025 | — |
| 11% | “11%” | Clerico: Utilities Cause 11% of Ignitions but 50% of Wildfire Damage | Bill Clerico | Jan 23, 2025 | Supported |
| 50% | “50%” | Clerico: Utilities Cause 11% of Ignitions but 50% of Wildfire Damage | Bill Clerico | Jan 23, 2025 | Supported |
| $4M | “four million dollars” | Clerico: Undergrounding Power Lines Costs $3M to $4M per Mile | Bill Clerico | Jan 23, 2025 | Supported |
| $5B | “five billion dollars” | Clerico: Regulated utilities profit by marking up CapEx 15% to ratepayers | Bill Clerico | Jan 23, 2025 | — |
| 15% | “15%” | Clerico: Regulated utilities profit by marking up CapEx 15% to ratepayers | Bill Clerico | Jan 23, 2025 | — |
| 10× | “10 times” | Bernhardsson: Developer productivity gains historically increase the number of engineers | Erik Bernhardsson | Jan 9, 2025 | — |
| 1B | “a billion” | Karpathy: Cognitive core models will likely shrink to one billion parameters | Andrej Karpathy | Dec 26, 2024 | — |
| 30M | “thirty million” | Torres: HubSpot Had $30M Revenue and 200 Employees at Performable Acquisition | Elias Torres | Dec 19, 2024 | Partly supported |
| 1B | “a billion” | Torres: HubSpot Scaled from $30M to $130M Revenue in Three Years | Elias Torres | Dec 19, 2024 | Partly supported |
| 30M | “thirty million” | Torres: HubSpot Scaled from $30M to $130M Revenue in Three Years | Elias Torres | Dec 19, 2024 | Partly supported |
| 1B | “a billion” | Torres: Drift reached $100M revenue before $1B cash sale | Elias Torres | Dec 19, 2024 | Partly supported |
| 100M | “a hundred million” | Torres: Drift reached $100M revenue before $1B cash sale | Elias Torres | Dec 19, 2024 | Partly supported |
| 31B | “a thirty billion” | Torres: Drift reached $100M revenue before $1B cash sale | Elias Torres | Dec 19, 2024 | Partly supported |
| 1M | “a million” | Torres: Customer success reps scrutinize LLM emails far more than their own | Elias Torres | Dec 19, 2024 | — |
| 3× | “three times” | Mitchell: Power, water, and heat will halt 2x-3x compute scaling cycles | Ty Mitchell | Dec 12, 2024 | — |
| 100M | “a hundred million” | Simons: AI Software Composers Could Eventually Reach 100 Million | Eric Simons | Dec 6, 2024 | — |
| 3M | “three million” | Simons: StackBlitz has 3 million monthly active developers | Eric Simons | Dec 6, 2024 | — |
| $10M | “ten million dollars” | GPT-4 Class Enterprise Models Now Cost $10M to $20M to Train | Aidan Gomez | Nov 21, 2024 | — |
| $20M | “twenty million dollars” | GPT-4 Class Enterprise Models Now Cost $10M to $20M to Train | Aidan Gomez | Nov 21, 2024 | — |
| $7B | “seven billion dollars” | Cohere Strategically Lags Frontier Labs by Six Months to Avoid $7B Burn | Aidan Gomez | Nov 21, 2024 | — |
| 100% | “hundred percent” | AI Will Eventually Run Experiments, but Scaling Will Take Many Years | Aidan Gomez | Nov 21, 2024 | — |
| 2× | “two x” | Huang: AI compute will follow hyper Moore's Law over the next decade | Jensen Huang | Nov 7, 2024 | Open |
| 1M | “a million” | Gil: GPT-4 Equivalent Token Costs Dropped 240x in 18 Months | Elad Gil | Nov 7, 2024 | Supported |
| 240× | “240 X” | Gil: GPT-4 Equivalent Token Costs Dropped 240x in 18 Months | Elad Gil | Nov 7, 2024 | Supported |
| 1M | “a million” | Jensen Huang: Computing marginal costs fell 1,000,000x over the last decade | Jensen Huang | Nov 7, 2024 | Contradicted |
| 55% | “55%” | Guo: Trump at 55% election odds is underpriced | Sarah Guo | Oct 31, 2024 | — |
| 4.1% | “4.1%” | Mansour: Kalshi pays 4.1% interest on active trade collateral | Tarek Mansour | Oct 31, 2024 | Partly supported |
| 6× | “six x” | Mansour: Kalshi grew sign-ups and volume 6x day-over-day after election launch | Tarek Mansour | Oct 31, 2024 | — |
| $10M | “ten million dollars” | Mansour: Kalshi expects $100M trades if institutional onboarding completes | Tarek Mansour | Oct 31, 2024 | — |
| $20M | “twenty million dollars” | Mansour: Kalshi expects $100M trades if institutional onboarding completes | Tarek Mansour | Oct 31, 2024 | — |
| 100M | “a hundred million” | Mansour: Kalshi expects $100M trades if institutional onboarding completes | Tarek Mansour | Oct 31, 2024 | — |
| 55% | “55%” | Mansour: Prediction market odds reflect win probabilities, not poll margins | Tarek Mansour | Oct 31, 2024 | — |
| 5% | “five percent” | Mansour: Prediction market odds reflect win probabilities, not poll margins | Tarek Mansour | Oct 31, 2024 | — |
| 1% | “one percent” | Mansour: Prediction market odds move more than underlying polling shifts | Tarek Mansour | Oct 31, 2024 | — |
| 50% | “50%” | Mansour: Half of Kalshi's markets originate from user submissions | Tarek Mansour | Oct 31, 2024 | — |
| 2% | “two percent” | Mansour: Probability of Elon Musk stepping down from Tesla is 15-20% | Tarek Mansour | Oct 31, 2024 | — |
| 15% | “15%” | Mansour: Probability of Elon Musk stepping down from Tesla is 15-20% | Tarek Mansour | Oct 31, 2024 | — |
| 1M | “a million” | Dolgov: Waymo completes over 100,000 trips and one million miles weekly | Dmitri Dolgov | Oct 24, 2024 | Supported |
| 100% | “hundred percent” | Dolgov: Swiss Re study showed 100% bodily injury claim reduction for Waymo | Dmitri Dolgov | Oct 24, 2024 | Supported |
| 4× | “four X” | Dolgov: Swiss Re study showed 100% bodily injury claim reduction for Waymo | Dmitri Dolgov | Oct 24, 2024 | Supported |
| 3T | “three trillion” | Waymo focuses on robotaxis but plans future trucking and delivery deployments | Dmitri Dolgov | Oct 24, 2024 | — |
| 50% | “50%” | Goyal: 50% of enterprise AI production use cases involve RAG | Ankur Goyal | Oct 8, 2024 | — |
| 3% | “three percent” | Khan: Under 3% of Reported Deals Are Investigated by FTC and DOJ | Lina Khan | Oct 3, 2024 | Supported |
| 100% | “hundred percent” | Tech companies pivoted to AI because their existing product roadmaps sucked | Matt MacInnis | Sep 25, 2024 | — |
| 200× | “200 X” | Gil: LLM API costs dropped roughly 200x in 18 to 24 months | Elad Gil | Sep 12, 2024 | Supported |
| 700M | “a six hundred million” | Guo: AMD's $600M Silo AI acquisition was essentially a massive acqui-hire | Sarah Guo | Sep 12, 2024 | Supported |
| 2B | “two billion” | Guo: AMD bought ZT Systems as a billion-dollar data center acqui-hire | Sarah Guo | Sep 12, 2024 | — |
| 1K | “a thousand” | Guo: AMD bought ZT Systems as a billion-dollar data center acqui-hire | Sarah Guo | Sep 12, 2024 | — |
| 3% | “three percent” | Karpathy: RoPE is the only major transformer architecture change in five years | Andrej Karpathy | Sep 5, 2024 | — |
| 1B | “a billion” | Karpathy: One billion human thought trajectories would roughly achieve AGI | Andrej Karpathy | Sep 5, 2024 | — |
| $1M | “a million dollars” | Steinberger: AI inference spending will follow an extreme power-law distribution | Eric Steinberger | Aug 30, 2024 | — |
| 4B | “four billion” | Steinberger: AI inference spending will follow an extreme power-law distribution | Eric Steinberger | Aug 30, 2024 | — |
| 99% | “99%” | Steinberger: 99% of users wouldn't notice Gemini 4 replacing 1.5 Pro | Eric Steinberger | Aug 30, 2024 | — |
| 5% | “five percent” | Steinberger: AI coding tool adoption will be a bimodal step-function | Eric Steinberger | Aug 30, 2024 | — |
| 90% | “90%” | Steinberger: AI coding tool adoption will be a bimodal step-function | Eric Steinberger | Aug 30, 2024 | — |
| 500M | “five hundred million” | AWS grew revenue from $500M in 2010 to $90B in 2023 | Elad Gil | Aug 29, 2024 | Supported |
| 90B | “ninety billion” | AWS grew revenue from $500M in 2010 to $90B in 2023 | Elad Gil | Aug 29, 2024 | Supported |
| 100B | “a hundred billion” | Garman estimates 85% of enterprise workloads still run on-premise | Matt Garman | Aug 29, 2024 | — |
| 85% | “85%” | Garman estimates 85% of enterprise workloads still run on-premise | Matt Garman | Aug 29, 2024 | — |
| 10× | “ten-x” | Garman estimates 85% of enterprise workloads still run on-premise | Matt Garman | Aug 29, 2024 | — |
| $27M | “twenty seven million dollars” | Garman's first startup burned $27 million in 18 months before folding | Matt Garman | Aug 29, 2024 | — |
| 10× | “10 X” | Davis: Democratizing AI Compute Will Increase AlphaFold-Level Breakthroughs 10x to 100x | Jared Quincy Davis | Aug 22, 2024 | — |
| 100× | “hundred X” | Davis: Democratizing AI Compute Will Increase AlphaFold-Level Breakthroughs 10x to 100x | Jared Quincy Davis | Aug 22, 2024 | — |
| 20% | “20%” | Davis: AI teams hold 10% to 20% of GPUs idle as healing buffer | Jared Quincy Davis | Aug 22, 2024 | — |
| 80% | “eighty percent” | Davis: Large-scale GPU pre-training utilization is sub-80% and often below 50% | Jared Quincy Davis | Aug 22, 2024 | — |
| 50% | “50%” | Davis: Large-scale GPU pre-training utilization is sub-80% and often below 50% | Jared Quincy Davis | Aug 22, 2024 | — |
| 1K | “a thousand” | Davis: AI cloud customers are forced into unwanted 3-year GPU contracts | Jared Quincy Davis | Aug 22, 2024 | — |
| 20M | “twenty million” | Davis: Peak Ethereum mining equaled 10 to 20 million V100 GPUs | Jared Quincy Davis | Aug 22, 2024 | Supported |
| 20% | “20%” | Davis: NVIDIA H100 GPU utilization is 25% or lower | Jared Quincy Davis | Aug 22, 2024 | — |
| 25% | “25%” | Davis: NVIDIA H100 GPU utilization is 25% or lower | Jared Quincy Davis | Aug 22, 2024 | — |
| 3.7% | “3.7%” | Davis: Compound AI approach increased prime factorization accuracy from 3.7% to 36.6% | Jared Quincy Davis | Aug 22, 2024 | Supported |
| 36.6% | “36.6%” | Davis: Compound AI approach increased prime factorization accuracy from 3.7% to 36.6% | Jared Quincy Davis | Aug 22, 2024 | Supported |
| 10× | “10 x” | Davis: Compound AI approach increased prime factorization accuracy from 3.7% to 36.6% | Jared Quincy Davis | Aug 22, 2024 | Supported |
| 3% | “three percent” | Davis: Compound AI approach yielded a 3% MMLU performance bump | Jared Quincy Davis | Aug 22, 2024 | Supported |
| 1% | “one percent” | Davis: Compound AI approach yielded a 3% MMLU performance bump | Jared Quincy Davis | Aug 22, 2024 | Supported |
| 2.8% | “2.8%” | Davis: Compound AI approach yielded a 3% MMLU performance bump | Jared Quincy Davis | Aug 22, 2024 | Supported |
| 1% | “one percent” | Glyman: Preventing spending is 100x better than earning 1% cash back | Eric Glyman | Aug 15, 2024 | — |
| 100× | “hundred times” | Glyman: Preventing spending is 100x better than earning 1% cash back | Eric Glyman | Aug 15, 2024 | — |
| 90% | “90%” | Atiyeh: Ramp uses GPT-4o Mini for 90% of its production AI tasks | Karim Atiyeh | Aug 15, 2024 | — |
| 9% | “nine percent” | Glyman: Internal data shows only 4-9% of finance roles are actually strategic | Eric Glyman | Aug 15, 2024 | — |
| $10B | “ten billion dollars” | Glyman: A $10B revenue company uses hundreds of finance tools | Eric Glyman | Aug 15, 2024 | — |
| 80% | “80%” | Franceschi: Brex hit an 80% AI suggestion acceptance rate in accounting | Pedro Franceschi | Aug 8, 2024 | — |
| 100B | “a hundred billion” | Franceschi: Brex recently closed a $100B market cap public company customer | Pedro Franceschi | Aug 8, 2024 | — |
| 50% | “50%” | Vinyals predicts pre-training compute will drop to ~50% as RL expands | Oriol Vinyals | Aug 1, 2024 | — |
| 20% | “20%” | Vinyals: General AI models are roughly 20% good at everything | Oriol Vinyals | Aug 1, 2024 | — |
| 1M | “a million” | Liu: Today's AI models hold a million times more untapped economic value | Howie Liu | Jul 25, 2024 | — |
| 10M | “ten million” | Guo: Healthcare AI startups hit $5M to $10M ARR in year one | Sarah Guo | Jul 18, 2024 | Supported |
| $500B | “half trillion dollars” | Gil: AI-transformable services headcount is $5T compared to $500B software spend | Elad Gil | Jul 18, 2024 | — |
| 5T | “five trillion” | Gil: AI-transformable services headcount is $5T compared to $500B software spend | Elad Gil | Jul 18, 2024 | — |
| 5× | “five X” | Gil: Major engineering cost shifts from AI coding agents remain years away | Elad Gil | Jul 18, 2024 | — |
| 34% | “34%” | Knoop: ARC-AGI benchmark performance only moved from 20% to 34% in four years | Mike Knoop | Jun 11, 2024 | Supported |
| 20% | “20%” | Knoop: ARC-AGI benchmark performance only moved from 20% to 34% in four years | Mike Knoop | Jun 11, 2024 | Supported |
| 200B | “two hundred billion” | Knoop: ARC solution will likely need under 10k code lines, not massive LLMs | Mike Knoop | Jun 11, 2024 | Open |
| 50M | “fifty million” | Knoop: Over 50M AI tasks have run on Zapier | Mike Knoop | Jun 11, 2024 | — |
| 2× | “two X” | Ma: Sophia Optimizer Improves LLM Pre-Training Efficiency by 2x | Tengyu Ma | Jun 6, 2024 | Supported |
| 100B | “a hundred billion” | Ma: Meta Achieved 1.6x Training Efficiency Gain Using Sophia Optimizer | Tengyu Ma | Jun 6, 2024 | — |
| 100M | “a hundred million” | Ma: 100M-Token Enterprises Cannot Afford Long-Context Inference Costs | Tengyu Ma | Jun 6, 2024 | — |
| 1T | “one trillion” | Ma: Voyage Fine-Tuned Models on 2T Code and 1T Legal Tokens | Tengyu Ma | Jun 6, 2024 | — |
| 10B | “ten billion” | Ma: Latency Limits Embedding Models to 10 Billion Parameters | Tengyu Ma | Jun 6, 2024 | — |
| 20% | “20%” | Ma: Domain-Specific Fine-Tuning Yields 15-20% Retrieval Gains in Code | Tengyu Ma | Jun 6, 2024 | Supported |
| 3× | “three X” | Ma: Voyage AI Produces Embeddings 3-4x Smaller Than Competitors | Tengyu Ma | Jun 6, 2024 | Supported |
| 4× | “four X” | Ma: Voyage AI Produces Embeddings 3-4x Smaller Than Competitors | Tengyu Ma | Jun 6, 2024 | Supported |
| 20% | “20%” | Ma: Proprietary Data Fine-Tuning Adds 10-20% Retrieval Accuracy | Tengyu Ma | Jun 6, 2024 | — |
| 70% | “70%” | Tan: About 70% of current Y Combinator batch companies are AI-centric | Garry Tan | May 23, 2024 | Supported |
| $6M | “six million dollars” | Tan: Last YC Batch Grew Aggregate ARR from $6M to $30M | Garry Tan | May 23, 2024 | — |
| $1M | “a million dollars” | Tan: Median YC startup post-batch fundraising has reached $1.5M–$2M | Garry Tan | May 23, 2024 | Supported |
| 1M | “a million” | Tan: Median YC startup post-batch fundraising has reached $1.5M–$2M | Garry Tan | May 23, 2024 | Supported |
| 10,000× | “10,000 times” | Alexandr Wang: High-quality frontier data is 10,000x more valuable than enterprise data. | Alexandr Wang | May 22, 2024 | — |
| 1B | “a billion” | Generative AI will dramatically increase time and money spent on music | Mikey Shulman | May 16, 2024 | — |
| $200B | “two hundred billion dollars” | Sarah Guo estimates hyperscalers will spend nearly $200 billion on AI compute | Sarah Guo | May 9, 2024 | Supported |
| 10M | “ten million” | Elad Gil predicts AI context windows will exceed 10 million tokens soon | Elad Gil | May 9, 2024 | Held up |
NO PRIORSLEDGER All shows Try the podcast clipper →