| 80% | “80%” | Taskaya: 80% of promotional video content will be AI-generated within 12 months | Batuhan Taskaya | Sep 8, 2025 | — |
| $1M | “a million dollars” | Taskaya: Training a state-of-the-art image model costs under $1M | Batuhan Taskaya | Sep 8, 2025 | Held up |
| 2M | “two million” | Landgraf: Gitpod grew to over two million developers | Johannes (Johannes Landgraf) | Sep 1, 2025 | — |
| 75% | “75%” | Landgraf: 75% of internal Ona pull requests are co-authored by AI | Johannes (Johannes Landgraf) | Sep 1, 2025 | — |
| 90% | “90%” | Landgraf: Ona has a 90% gross margin on enterprise offering | Johannes (Johannes Landgraf) | Sep 1, 2025 | — |
| 3× | “three times” | Weichel claims he is 3x more productive on his phone with Ona | Chris (Christian Weichel) | Sep 1, 2025 | — |
| 1M | “a million” | Morcos: Soft inductive biases become harmful past 1M data points in vision | Ari Morcos | Aug 29, 2025 | Supported |
| 10× | “10 X” | Morcos: Power-law scaling yields diminishing returns for every 10x data increase | Ari Morcos | Aug 29, 2025 | — |
| 10% | “10%” | Morcos: Datology matches DCLM performance 12x faster with under 10% tokens | Ari Morcos | Aug 29, 2025 | Open |
| 12× | “12 x” | Morcos: Datology matches DCLM performance 12x faster with under 10% tokens | Ari Morcos | Aug 29, 2025 | Open |
| 10× | “10 X” | Morcos: Data curation still has at least 100x in performance gains ahead | Ari Morcos | Aug 29, 2025 | — |
| 100× | “hundred X” | Morcos: Data curation still has at least 100x in performance gains ahead | Ari Morcos | Aug 29, 2025 | — |
| $1M | “a million dollars” | Morcos: Training a specialized frontier model will cost under $1M very soon | Ari Morcos | Aug 29, 2025 | Held up |
| 10× | “10 times” | Morcos: Proper training curricula could reduce model training costs by 10x | Ari Morcos | Aug 29, 2025 | — |
| 1T | “one trillion” | Morcos: Arcee 4.5B beat Gemma before reaching one trillion tokens | Ari Morcos | Aug 29, 2025 | Not publicly verifiable |
| 5M | “five million” | Huber: Flawless 60k-token reasoning is more valuable than 5M-token context | Jeff Huber | Aug 19, 2025 | — |
| 1K | “a thousand” | Huber: LLMs will largely replace purpose-built re-rankers | Jeff Huber | Aug 19, 2025 | — |
| 90% | “90%” | Huber: Regex handles 90% of code queries; embeddings add marginal improvement | Jeff Huber | Aug 19, 2025 | — |
| 85% | “85%” | Huber: Regex handles 90% of code queries; embeddings add marginal improvement | Jeff Huber | Aug 19, 2025 | — |
| 15% | “15%” | Huber: Regex handles 90% of code queries; embeddings add marginal improvement | Jeff Huber | Aug 19, 2025 | — |
| 10% | “10%” | Huber: Regex handles 90% of code queries; embeddings add marginal improvement | Jeff Huber | Aug 19, 2025 | — |
| 5% | “five percent” | Huber: Regex handles 90% of code queries; embeddings add marginal improvement | Jeff Huber | Aug 19, 2025 | — |
| 1M | “a million” | Huber: A couple hundred high-quality labeled examples offer massive ML returns | Jeff Huber | Aug 19, 2025 | — |
| 1.5B | “half a billion” | Agrawal: Lambda Labs generates well over $500M in ARR | Mitesh Agrawal | Aug 18, 2025 | Supported |
| 93% | “93%” | Sohmers: Positron AI hardware achieves 93% of theoretical memory bandwidth | Thomas Sohmers | Aug 18, 2025 | Open |
| 70% | “70%” | Sohmers: Positron hardware achieves 70% higher performance than NVIDIA at lower power | Thomas Sohmers | Aug 18, 2025 | Open |
| 51M | “fifty one million” | Agrawal: Positron AI raised a $51M Series A for next-gen silicon | Mitesh Agrawal | Aug 18, 2025 | — |
| 400M | “a three hundred million” | Brockman: OpenAI's Dota AI used only 300 million parameters | Greg Brockman | Aug 15, 2025 | Contradicted |
| 13T | “13 trillion” | Brockman: Arc Institute trained 40B DNA model on 13T base pairs | Greg Brockman | Aug 15, 2025 | Partly supported |
| 80% | “80%” | Brockman: OpenAI's 80% o3 price cut yielded neutral or positive revenue | Greg Brockman | Aug 15, 2025 | — |
| 10× | “10 x” | Brockman: AI developer productivity gains will increase demand for engineers | Greg Brockman | Aug 15, 2025 | — |
| 100× | “hundred x” | Brockman: AI developer productivity gains will increase demand for engineers | Greg Brockman | Aug 15, 2025 | — |
| 6M | “a five million” | Fanelli: Small venture funds cannot back AI inference startups with $5M rounds | Alessio Fanelli | Aug 6, 2025 | — |
| 75% | “75%” | 75% of DeepSeek-R1 usage at a major inference provider is for distillation | Stephanie Palazzolo | Aug 6, 2025 | — |
| 90% | “90%” | Palazzolo: Meta should focus on app integration over frontier models | Stephanie Palazzolo | Aug 6, 2025 | — |
| 12B | “twelve billion” | The Information: OpenAI hit $12B ARR as burn rose to $8B | Stephanie Palazzolo | Aug 6, 2025 | Partly supported |
| 1B | “one billion” | The Information: OpenAI hit $12B ARR as burn rose to $8B | Stephanie Palazzolo | Aug 6, 2025 | Partly supported |
| 8B | “eight billion” | The Information: OpenAI hit $12B ARR as burn rose to $8B | Stephanie Palazzolo | Aug 6, 2025 | Partly supported |
| 3% | “three percent” | Dax Reed: Optimizing marginal LLM efficiency yields plateauing returns. | Dax Reed | Aug 5, 2025 | — |
| 80% | “80%” | Dax Reed notes developers avoid context limits via frequent session restarts. | Dax Reed | Aug 5, 2025 | — |
| 10× | “10 X” | Inception generalist model matches Claude Haiku quality at 5-10x speed | Stefano Ermon | Aug 4, 2025 | Partly supported |
| 10× | “10 X” | Ermon: Power constraints will drive diffusion models to replace frontier LLMs | Stefano Ermon | Aug 4, 2025 | — |
| 95% | “95%” | Ganatra: 95% of Composio integrations are built and maintained by agents | Soham Ganatra | Aug 4, 2025 | — |
| 100× | “hundred X” | Lambert: Hybrid reasoners may be phased out except for niche uses | Nathan Lambert | Jul 31, 2025 | — |
| 1K | “a thousand” | Lambert: RLVR is harder to over-optimize on math than code | Nathan Lambert | Jul 31, 2025 | — |
| 50% | “50%” | Epic controls over 50% and Oracle Cerner holds 27% of EHR market | Brendan Fortuna | Jul 29, 2025 | Supported |
| 27% | “27%” | Epic controls over 50% and Oracle Cerner holds 27% of EHR market | Brendan Fortuna | Jul 29, 2025 | Supported |
| 5% | “five percent” | Fortuna: Ambient Scribing Is Only Five Percent of AI's Healthcare Value | Brendan Fortuna | Jul 29, 2025 | — |
| 40% | “40%” | RFT boosted o3-mini to 57% F1 on medical coding versus clinicians' 40% | Brendan Fortuna | Jul 29, 2025 | Partly supported |
| 57% | “57%” | RFT boosted o3-mini to 57% F1 on medical coding versus clinicians' 40% | Brendan Fortuna | Jul 29, 2025 | Partly supported |
| 97% | “97%” | Fanelli: ExaFunction customer cut compute costs 97% on single GPU | Alessio Fanelli | Jul 28, 2025 | — |
| 70% | “70%” | Swix: Copilot report estimates 60-70% of AI-generated code is checked in | Shawn Wang | Jul 28, 2025 | Contradicted |
| 5% | “five percent” | Mohan: Codeium reached 10k users and 5% daily growth in late 2022 | Varun Mohan | Jul 28, 2025 | — |
| 170B | “one hundred seventy billion” | Mohan: Training optimizer state requires 14x model parameter size in memory | Varun Mohan | Jul 28, 2025 | Supported |
| 14× | “14 times” | Mohan: Training optimizer state requires 14x model parameter size in memory | Varun Mohan | Jul 28, 2025 | Supported |
| 1M | “a million” | Hou: Codeium surpassed 1.5M downloads across 40 IDEs | Kevin Hou | Jul 28, 2025 | Supported |
| 100× | “100 X” | Hou: Codeium inference costs 1/100th of competitors by avoiding third-party APIs | Kevin Hou | Jul 28, 2025 | — |
| 20% | “20%” | Scott Wu: Software engineers spend 80-90% of time on implementation | Scott Wu | Jul 28, 2025 | — |
| 90% | “90%” | Scott Wu: Software engineers spend 80-90% of time on implementation | Scott Wu | Jul 28, 2025 | — |
| 10× | “10 X” | Scott Wu: AI will make engineers 5-10x more effective | Scott Wu | Jul 28, 2025 | — |
| 10% | “10%” | Mohan: Squeezing the last 10% from AI benchmarks is counterproductive | Varun Mohan | Jul 28, 2025 | — |
| 80% | “80%” | Mohan: Over 80% of software developers are on Windows | Varun Mohan | Jul 28, 2025 | Contradicted |
| 10M | “ten million” | Ramachandran: Codeium grew from zero to $10M ARR in under a year | Anshul Ramachandran | Jul 28, 2025 | — |
| 4.5B | “4.5 billion” | Hou: Windsurf generated 4.5 billion lines of code in three months | Kevin Hou | Jul 28, 2025 | — |
| 99% | “99%” | Hou: 99% of AI editor rules file contents will be automatically inferred | Kevin Hou | Jul 28, 2025 | — |
| 90% | “90%” | Hou: 90% of code written by Windsurf users is generated by Cascade | Kevin Hou | Jul 28, 2025 | — |
| 30% | “30%” | Hou: 90% of code written by Windsurf users is generated by Cascade | Kevin Hou | Jul 28, 2025 | — |
| 64× | “64 X” | Wu: Autonomous coding agent capability currently doubles every 70 days | Scott Wu | Jul 28, 2025 | — |
| 64× | “64 X” | Wu predicts AI coding agents will advance 16x to 64x in 12 months | Scott Wu | Jul 28, 2025 | — |
| 90% | “90%” | Windsurf aims to shift coding workflows from 80% agent to 99% agent | Kevin Hou | Jul 28, 2025 | — |
| 20% | “20%” | Windsurf aims to shift coding workflows from 80% agent to 99% agent | Kevin Hou | Jul 28, 2025 | — |
| 99% | “99%” | Windsurf aims to shift coding workflows from 80% agent to 99% agent | Kevin Hou | Jul 28, 2025 | — |
| 1% | “one percent” | Windsurf aims to shift coding workflows from 80% agent to 99% agent | Kevin Hou | Jul 28, 2025 | — |
| 1M | “a million” | Wu: Global software engineers grew from under 1M in 2000 to over 30M in 2025 | Scott Wu | Jul 28, 2025 | Partly supported |
| 30M | “thirty million” | Wu: Global software engineers grew from under 1M in 2000 to over 30M in 2025 | Scott Wu | Jul 28, 2025 | Partly supported |
| 1M | “one million” | The Lean theorem proving language has only about one million training tokens | Dr. Jasper Zhang | Jul 24, 2025 | Contradicted |
| 6× | “six X” | McCloy: Clerk Achieved 6x AI Traffic Growth and 9x Lift in Conversions | Robert McCloy | Jul 23, 2025 | — |
| 9× | “nine X” | McCloy: Clerk Achieved 6x AI Traffic Growth and 9x Lift in Conversions | Robert McCloy | Jul 23, 2025 | — |
| $1M | “a million dollars” | Kamradt: Mike Knoop Put Up $1M for ARC Prize Bounty | Greg Kamradt | Jul 18, 2025 | Supported |
| 1M | “a million” | Kamradt: Random Brute Force Agent Fails ARC-AGI-3 Locksmith Game | Greg Kamradt | Jul 18, 2025 | Supported |
| 90% | “90%” | Hsu: 80% to 90% of Tech for Superhuman AI Tutors Exists | Andrew Hsu | Jul 11, 2025 | — |
| 50M | “fifty million” | Speak has surpassed $50 million in annual recurring revenue | Andrew Hsu | Jul 11, 2025 | Supported |
| 90% | “90%” | Hsu: Speak has 90% of product team in SF and only hires there | Andrew Hsu | Jul 11, 2025 | — |
| 100× | “hundred X” | Hsu: AI Gives Content and Engineering 100x Leverage While Still Requiring Review | Andrew Hsu | Jul 11, 2025 | — |
| 90% | “90%” | Olivia Moore: 90% of TikTok and Reels feeds are AI-generated video | Olivia Moore | Jul 9, 2025 | — |
| 80% | “80%” | Anthropic finds multi-agent architecture outperforms single-agent baseline by 80% | Dylan Davis | Jul 5, 2025 | Partly supported |
| 4× | “four X” | Davis: Multi-agent research consumes 15x baseline tokens versus 4x for single-agent | Dylan Davis | Jul 5, 2025 | Supported |
| 15× | “15 X” | Davis: Multi-agent research consumes 15x baseline tokens versus 4x for single-agent | Dylan Davis | Jul 5, 2025 | Supported |
| 100× | “hundred X” | Swix: AI Inference Costs for Fixed Intelligence Fall 100x Annually | Shawn Wang | Jul 5, 2025 | Partly supported |
| 500M | “five hundred million” | Swyx: Stanford RL students founded pre-product startup with $500M valuation | Shawn Wang | Jul 2, 2025 | Supported |
| 90% | “90%” | Morris: New embedding inversion model exactly recovers 90% of source text | Jack Morris | Jul 2, 2025 | Supported |
| 1M | “a million” | Zach Lloyd: Warp relies on over one million lines of Rust code | Zach Lloyd | Jun 25, 2025 | — |
| 15% | “15%” | Zach Lloyd: Warp's revenue is growing 5% to 15% week-over-week | Zach Lloyd | Jun 25, 2025 | — |
| 1K | “a thousand” | Lloyd: Warp expands Pro and Turbo request limits and adds credit top-ups | Zach Lloyd | Jun 25, 2025 | — |
| 5× | “five X” | Intelligent compute routing delivers 5x performance TCO benefit | Chris Lattner | Jun 13, 2025 | — |
| 5% | “five percent” | Ameisen: Naive model pruning fails because superposition distributes critical representations | Emmanuel Ameisen | Jun 6, 2025 | — |
| 50% | “50%” | Abraham: CloudChef robot outperforms expert chefs on 40-50% of commercial cuisines | Nikhil Abraham | May 31, 2025 | — |
| 30% | “30%” | Abraham: Average restaurant operates at 130% staff turnover | Nikhil Abraham | May 31, 2025 | Partly supported |
| 40% | “40%” | Abraham: CloudChef's $12-an-hour robot costs 40% of loaded human labor | Nikhil Abraham | May 31, 2025 | Supported |
| 100% | “100%” | CloudChef Operates with 100% Decision Autonomy and 90% Action Autonomy | Nikhil Abraham | May 31, 2025 | — |
| 90% | “90%” | CloudChef Operates with 100% Decision Autonomy and 90% Action Autonomy | Nikhil Abraham | May 31, 2025 | — |
| 15% | “15%” | Enterprises demand full AI delegation, not 15% to 20% speed gains | Eno Reyes | May 29, 2025 | — |
| 20% | “20%” | Enterprises demand full AI delegation, not 15% to 20% speed gains | Eno Reyes | May 29, 2025 | — |
| 100% | “hundred percent” | Optimal future AI coding interfaces will not evolve from traditional IDEs | Matan Grinberg | May 29, 2025 | — |
| 3% | “three percent” | Reyes: High-quality enterprise codebases experience only 3% to 4% code churn | Eno Reyes | May 29, 2025 | — |
| 4% | “four percent” | Reyes: High-quality enterprise codebases experience only 3% to 4% code churn | Eno Reyes | May 29, 2025 | — |
| 20% | “20%” | Reyes: High-quality enterprise codebases experience only 3% to 4% code churn | Eno Reyes | May 29, 2025 | — |
| 100% | “hundred percent” | Brown: Prompting alone cannot reliably force LLMs to use thinking tokens | Will Brown | May 23, 2025 | — |
| 80% | “80%” | Alberti: 80% to 90% of DeepWiki users view pre-indexed popular repos | Silas Alberti | May 21, 2025 | — |
| 90% | “90%” | Alberti: 80% to 90% of DeepWiki users view pre-indexed popular repos | Silas Alberti | May 21, 2025 | — |
| 1K | “a thousand” | Alberti: Over 1,000 GitHub projects added DeepWiki badges to their repositories | Silas Alberti | May 21, 2025 | Supported |
| 85% | “85%” | Fanelli: Series A Startups Have 80% to 85% AI-Generated Code | Alessio Fanelli | May 7, 2025 | — |
| 1K | “a thousand” | Wu: Internal operational tools are a major use case for Claude Code | Kat Wu | May 7, 2025 | — |
| 3× | “three times” | Cherny: Claude Code spawns parallel sub-agents to investigate complex coding tasks | Boris Cherny | May 7, 2025 | Supported |
| 5× | “five times” | Cherny: Claude Code spawns parallel sub-agents to investigate complex coding tasks | Boris Cherny | May 7, 2025 | Supported |
| 2× | “two X” | Cherny: Claude Code delivers up to 10x productivity gains for Anthropic engineers | Boris Cherny | May 7, 2025 | — |
| 10× | “10 X” | Cherny: Claude Code delivers up to 10x productivity gains for Anthropic engineers | Boris Cherny | May 7, 2025 | — |
| 2.5M | “a half a million” | Sobo: Zed engineered an editor from scratch to 500,000 lines of Rust | Nathan Sobo | May 7, 2025 | Supported |
| 600M | “six hundred million” | NVIDIA's 600M Parameter Parakeet Model Tops Speech Transcription Leaderboards | Kwindla Hultman Kramer | May 6, 2025 | Supported |
| 99% | “99%” | 99% of Current Monetizable Voice AI Use Cases Are Telephony | Kwindla Hultman Kramer | May 6, 2025 | — |
| 50% | “50%” | Up to 75% of Future UX Interfaces Will Be Voice-Driven | Kwindla Hultman Kramer | May 6, 2025 | — |
| 60% | “60%” | Up to 75% of Future UX Interfaces Will Be Voice-Driven | Kwindla Hultman Kramer | May 6, 2025 | — |
| 75% | “75%” | Up to 75% of Future UX Interfaces Will Be Voice-Driven | Kwindla Hultman Kramer | May 6, 2025 | — |
| 100% | “hundred percent” | Up to 75% of Future UX Interfaces Will Be Voice-Driven | Kwindla Hultman Kramer | May 6, 2025 | — |
| 1M | “a million” | Factorio requires one million resources to beat compared to Minecraft's 200 | Jack Hopkins | Apr 27, 2025 | Supported |
| 1B | “a billion” | Dual reward signals prevent AI agent behavioral collapse in Factorio | Jack Hopkins | Apr 27, 2025 | — |
| 1T | “a trillion” | Dual reward signals prevent AI agent behavioral collapse in Factorio | Jack Hopkins | Apr 27, 2025 | — |
| 1K | “A thousand” | Providing agents with RAG factory blueprints yielded zero benchmark score improvement | Jack Hopkins | Apr 27, 2025 | Not publicly verifiable |
| 100× | “hundred times” | Untrained AI models exhibit a 100x competency gap versus human players | Jack Hopkins | Apr 27, 2025 | Contradicted |
| 15M | “fifteen million” | Mlejnsky: E2B ran around 15 million sandboxes in March 2025 | Vasek Mlejnsky | Apr 24, 2025 | — |
| 1.5M | “half a million” | Mlejnsky: E2B sees 250k JavaScript and ~500k Python SDK monthly downloads | Vasek Mlejnsky | Apr 24, 2025 | Not publicly verifiable |
| 20M | “twenty million” | Mlejnsky: LangChain gets 20 million monthly downloads and remains popular | Vasek Mlejnsky | Apr 24, 2025 | Supported |
| 10× | “10 times” | Mlejnsky: DevTool Startups in Existing Categories Do Not Need to Be in SF | Vasek Mlejnsky | Apr 24, 2025 | — |
| 100× | “hundred times” | Mlejnsky: DevTool Startups in Existing Categories Do Not Need to Be in SF | Vasek Mlejnsky | Apr 24, 2025 | — |
| 3× | “three times” | Mlejnsky: Patrick Collison Said Stripe Did the 'Collison Installation' Only Three Times | Vasek Mlejnsky | Apr 24, 2025 | — |
| 1M | “a million” | Oleve's Unstuck AI hit 1M users in under 9 weeks | Sid Bendre | Apr 23, 2025 | — |
| 5M | “five million” | Oleve reports $6M ARR, 5M users, and sustained profitability | Sid Bendre | Apr 23, 2025 | — |
| $6M | “six million dollars” | Oleve reports $6M ARR, 5M users, and sustained profitability | Sid Bendre | Apr 23, 2025 | — |
| 50M | “fifty million” | Unstuck AI's launch campaign reached 250 million views in one month | Sid Bendre | Apr 23, 2025 | — |
| 1M | “one million” | OpenAI launches GPT-4.1 model lineup featuring 1M-token context window | Michelle Pokrass | Apr 15, 2025 | Supported |
| 9% | “nine percent” | GPT-4.1 reduces extraneous edit rate to 2%, down from GPT-4o's 9% | Michelle Pokrass | Apr 15, 2025 | Supported |
| 2% | “two percent” | GPT-4.1 reduces extraneous edit rate to 2%, down from GPT-4o's 9% | Michelle Pokrass | Apr 15, 2025 | Supported |
| 50% | “50%” | OpenAI increases prompt caching discount from 50% to 75% on GPT-4.1 | Michelle Pokrass | Apr 15, 2025 | Supported |
| 75% | “75%” | OpenAI increases prompt caching discount from 50% to 75% on GPT-4.1 | Michelle Pokrass | Apr 15, 2025 | Supported |
| 5% | “five percent” | Conrad: Incremental GPUs always drive model performance and revenue, unlike CPUs | Evan Conrad | Apr 11, 2025 | — |
| $100M | “hundred million dollars” | Conrad: Software margins on GPU clusters drive customers to build in-house | Evan Conrad | Apr 11, 2025 | — |
| $50M | “fifty million dollars” | Conrad: Software margins on GPU clusters drive customers to build in-house | Evan Conrad | Apr 11, 2025 | — |
| 10% | “10%” | Conrad: Software margins on GPU clusters drive customers to build in-house | Evan Conrad | Apr 11, 2025 | — |
| 77% | “77%” | Swyx: Microsoft and OpenAI account for 77% of CoreWeave revenue | Michael Swix (Swyx) | Apr 11, 2025 | Partly supported |
| 5B | “five billion” | Swyx: At $5B+ training runs, designing custom chips makes economic sense | Michael Swix (Swyx) | Apr 11, 2025 | — |
| 50B | “fifty billion” | Swyx: At $5B+ training runs, designing custom chips makes economic sense | Michael Swix (Swyx) | Apr 11, 2025 | — |
| 100% | “hundred percent” | Conrad: Spot GPU cluster utilization nears 100% through dynamic price clearing | Evan Conrad | Apr 11, 2025 | — |
| 10% | “10%” | Hershey: Chain-of-thought between tool calls only improves agent progress 10% | David Hershey | Apr 5, 2025 | — |
| 80% | “80%” | Developers will eventually spend 80% of their time controlling agents outside IDEs | Guy Gur-Ari | Apr 2, 2025 | — |
| 20% | “20%” | Developers will eventually spend 80% of their time controlling agents outside IDEs | Guy Gur-Ari | Apr 2, 2025 | — |
| 1.3M | “1.3 million” | Shah: Agent.ai has 1.3M users and 1,000 published agents | Dharmesh Shah | Mar 28, 2025 | — |
| 1K | “a thousand” | Shah: Agent.ai has 1.3M users and 1,000 published agents | Dharmesh Shah | Mar 28, 2025 | — |
| 3M | “three million” | Shah: I built a personal vector store indexing 3 million of my emails | Dharmesh Shah | Mar 28, 2025 | — |
| 52M | “two-fifty million” | Agarwal: Synthetic Data Distillation Can Outperform Logits on Benchmarks | Rishabh Agarwal | Mar 23, 2025 | Supported |
| 80% | “80%” | Agarwal: Synthetic data distillation achieves 80% to 90% of target gains | Rishabh Agarwal | Mar 23, 2025 | — |
| 90% | “90%” | Agarwal: Synthetic data distillation achieves 80% to 90% of target gains | Rishabh Agarwal | Mar 23, 2025 | — |
| 99% | “99%” | Ben-Smith: Snipd indexes 99% of all podcasts in-house | Kevin Ben-Smith | Mar 14, 2025 | — |
| 90% | “90%” | Snipd builds on Python, GCP, and Flutter for cross-platform clients | Kevin Ben-Smith | Mar 14, 2025 | — |
| 1M | “a million” | Snipd has processed more than 1 million podcasts | Kevin Ben-Smith | Mar 14, 2025 | — |
| 1K | “a thousand” | Ben-Smith: Perplexity web search API costs roughly $5 per 1,000 queries | Kevin Ben-Smith | Mar 14, 2025 | Supported |
| 75% | “75%” | Husain: 80% of LLM-as-a-judge implementations are unhelpful | Hamel Husain | Mar 13, 2025 | — |
| 80% | “80%” | Husain: 80% of LLM-as-a-judge implementations are unhelpful | Hamel Husain | Mar 13, 2025 | — |
| 30% | “30%” | Alessio Fanelli: GPT-4o Search jumps to 90% accuracy on simple QA | Alessio Fanelli | Mar 11, 2025 | Partly supported |
| 90% | “90%” | Alessio Fanelli: GPT-4o Search jumps to 90% accuracy on simple QA | Alessio Fanelli | Mar 11, 2025 | Partly supported |
| 10× | “10 times” | Reinforcement Learning Fails Without Initial SFT to Seed Rewardable Behaviors | Misha Laskin | Mar 7, 2025 | — |
| 50× | “50 times” | Reinforcement Learning Fails Without Initial SFT to Seed Rewardable Behaviors | Misha Laskin | Mar 7, 2025 | — |
| 90% | “90%” | A 90% SWE-Bench Score Can Still Fall Flat in Customer Environments | Misha Laskin | Mar 7, 2025 | — |
| 1K | “a thousand” | Klein: Web scraping workflows should use a tiered waterfall architecture | Paul Klein | Feb 28, 2025 | — |
| 100% | “hundred percent” | Klein: Web scraping workflows should use a tiered waterfall architecture | Paul Klein | Feb 28, 2025 | — |
| 90% | “90%” | Klein: Browserbase delivers 90% of computer-use functionality at 10% of OS cost | Paul Klein | Feb 28, 2025 | — |
| 10% | “10%” | Klein: Browserbase delivers 90% of computer-use functionality at 10% of OS cost | Paul Klein | Feb 28, 2025 | — |
| 1B | “a billion” | Klein: Browserbase will be a billion-dollar company within five years | Paul Klein | Feb 28, 2025 | Open |
| 1M | “a million” | Google simplifies Gemini Flash pricing to flat 10 cents per million tokens | Logan Kilpatrick | Feb 28, 2025 | Supported |
| 2M | “two million” | Kilpatrick: Reasoning will solve multi-item retrieval in long context | Logan Kilpatrick | Feb 28, 2025 | — |
| 80% | “80%” | Kilpatrick: Multimodal Live API provides 80% of Project Astra experience | Logan Kilpatrick | Feb 28, 2025 | — |
| 90% | “90%” | Roucher: AI agents will reach a 90% GAIA score by 2026 | Aymeric (Emmerich) | Feb 13, 2025 | Held up |
| 10× | “10 X” | Bret Taylor's Rewrite Slashed Google Maps Bundle Size to 20KB | Bret Taylor | Feb 11, 2025 | Not publicly verifiable |
| 300M | “three hundred million” | Swix: Pydantic reached nearly 300 million downloads in December | Shawn Wang | Feb 6, 2025 | Supported |
| 5× | “five times” | Colvin: Rust-native data storage could yield 3x to 5x Pydantic speedup | Samuel Colvin | Feb 6, 2025 | Open |
| 20% | “20%” | Colvin: Major AI lab cut time-to-first-token 20% upgrading to Pydantic v2 | Samuel Colvin | Feb 6, 2025 | — |
| 90% | “90%” | Agarwal: 90% of production LLM use cases do not use automatic routing | Rohit Agarwal | Feb 5, 2025 | — |
| 57% | “57%” | Shawn Lewis: o1 agent achieves 57% single-pass, 64% with parallel rollouts | Shawn Lewis | Jan 28, 2025 | Supported |
| 64% | “64%” | Shawn Lewis: o1 agent achieves 57% single-pass, 64% with parallel rollouts | Shawn Lewis | Jan 28, 2025 | Supported |
| 1K | “a thousand” | Lewis: Ran approximately 1,000 evaluations while developing SWE-bench agent | Shawn Lewis | Jan 28, 2025 | — |
| 6% | “six percent” | Topping SWE-bench requires multi-trajectory sampling and high compute costs | Shawn Lewis | Jan 28, 2025 | — |
| $10M | “ten million dollars” | Beauchamp: Small trading funds hold an advantage exploiting niche market anomalies | William Beauchamp | Jan 26, 2025 | — |
| 10M | “ten million” | Beauchamp: Small trading funds hold an advantage exploiting niche market anomalies | William Beauchamp | Jan 26, 2025 | — |
| 1% | “one percent” | Beauchamp: Small trading funds hold an advantage exploiting niche market anomalies | William Beauchamp | Jan 26, 2025 | — |
| 100% | “hundred percent” | Beauchamp: Small trading funds hold an advantage exploiting niche market anomalies | William Beauchamp | Jan 26, 2025 | — |
| 5M | “five million” | Beauchamp: Quant trading firm made £5 million annually with 15-person team | William Beauchamp | Jan 26, 2025 | Not publicly verifiable |
| 10× | “10 X” | Beauchamp: AI is 10x better at non-judgmental conversation than informative tasks | William Beauchamp | Jan 26, 2025 | — |
| 6B | “six billion” | Beauchamp: Chai unlocked growth by letting consumers build bots with GPT-J | William Beauchamp | Jan 26, 2025 | — |
| 2M | “two million” | Beauchamp: Bootstrapped Chai to 100k DAUs with £2 million personal investment | William Beauchamp | Jan 26, 2025 | — |
| 15% | “15%” | Beauchamp: Only 10% to 15% of Chai users engaged with voice | William Beauchamp | Jan 26, 2025 | — |
| 50% | “50%” | Beauchamp: Randomly blending specialized LLMs delivers an effective 80/20 user experience | William Beauchamp | Jan 26, 2025 | — |
LATENT SPACELEDGER All shows Try the podcast clipper →