We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/mpnikhil/lenny-rag-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
Keith Coleman & Jay Baxter.json•34.6 KiB
{
"episode": {
"guest": "Keith Coleman (Product Lead, Community Notes) and Jay Baxter (Founding ML Engineer and Researcher, Community Notes)",
"expertise_tags": [
"Product Management",
"Machine Learning",
"Misinformation Detection",
"Community-Driven Systems",
"Algorithm Design",
"Organizational Structure",
"Team Building"
],
"summary": "Keith Coleman and Jay Baxter discuss the origin, design, and impact of Community Notes, X's crowdsourced fact-checking system. They explain how the bridging-based algorithm identifies notes helpful to people who normally disagree, making them trustworthy across political divides. The conversation covers the product's surprising emergence from an internal contest, its survival through four different leaders (Jack Dorsey, Kayvon Rostamizadeh, Parag Agrawal, Elon Musk, and Linda Yaccarino), and how a small, autonomous team with clear principles achieved internet-scale impact. They share lessons on team structure, the importance of proving product value incrementally, and why radical transparency and user agency were essential to the product's success.",
"key_frameworks": [
"Bridging-based agreement algorithm",
"Thermal team structure",
"Voice of the people principle",
"Radical transparency model",
"Pseudonymous contribution system",
"Iterative proof-of-concept validation",
"Small autonomous team with single decision-maker"
]
},
"topics": [
{
"id": "topic_1",
"title": "What is Community Notes and how does it work",
"summary": "Overview of Community Notes as a system allowing X users to propose context-adding notes on potentially misleading posts. Notes are rated by other users and shown when rated helpful by people who normally disagree, indicating neutrality and accuracy.",
"timestamp_start": "00:00:00",
"timestamp_end": "00:06:57",
"line_start": 1,
"line_end": 71
},
{
"id": "topic_2",
"title": "The bridging-based agreement algorithm",
"summary": "Deep dive into the core algorithm that makes Community Notes work: finding agreement from people who have disagreed in the past. Explains why this approach prevents bias amplification and manipulation compared to simpler voting systems.",
"timestamp_start": "00:07:22",
"timestamp_end": "00:09:23",
"line_start": 73,
"line_end": 95
},
{
"id": "topic_3",
"title": "Design principles: openness to all participants",
"summary": "Discussion of the decision to allow any user (with verified phone) to become a Community Notes contributor, rather than restricting to journalists or experts. Covers random selection process and commitment to including all of humanity in the system.",
"timestamp_start": "00:09:40",
"timestamp_end": "00:13:20",
"line_start": 97,
"line_end": 126
},
{
"id": "topic_4",
"title": "Scale and impact metrics",
"summary": "Statistics on Community Notes growth: 950,000 contributors, hundreds of notes per day, 95,000 notes seen 30 billion times in 2024 (double prior year), automatic matching of notes across identical images/videos.",
"timestamp_start": "00:13:39",
"timestamp_end": "00:16:11",
"line_start": 148,
"line_end": 165
},
{
"id": "topic_5",
"title": "Handling polarized and extreme viewpoints",
"summary": "How the algorithm handles super polarized topics and users with extreme views. Notes on super polarized topics may not show if no bridging agreement exists, but the system benefits from including all viewpoints in the rating pool.",
"timestamp_start": "00:21:31",
"timestamp_end": "00:24:39",
"line_start": 232,
"line_end": 249
},
{
"id": "topic_6",
"title": "Keith's journey to starting Community Notes",
"summary": "Keith's career path from startup founder to Twitter in 2016, observing the misinformation problem during the Trump-Clinton election, his role in Twitter's turnaround, and his decision to leave PM management to focus on this single problem.",
"timestamp_start": "00:30:19",
"timestamp_end": "00:33:46",
"line_start": 310,
"line_end": 320
},
{
"id": "topic_7",
"title": "The Thermal team structure and autonomy",
"summary": "Explanation of Twitter's Thermal program that gave the Community Notes team autonomy, isolation from bureaucracy, 100% focus, clear decision-making (one sponsor, one driver), and freedom from standard practices like OKRs. How this structure was critical to the product's success.",
"timestamp_start": "00:40:37",
"timestamp_end": "00:45:15",
"line_start": 367,
"line_end": 404
},
{
"id": "topic_8",
"title": "Small team composition and hiring",
"summary": "Started with five people: one ML engineer, frontend, backend, designer, and researcher. Emphasis on self-selection—people chose to join rather than being assigned. Keith and team interviewed every hire to ensure alignment with mission.",
"timestamp_start": "00:45:19",
"timestamp_end": "00:57:10",
"line_start": 409,
"line_end": 487
},
{
"id": "topic_9",
"title": "Algorithm evolution from PageRank to bridging-based approach",
"summary": "Jay's journey from initial PageRank-based anti-manipulation algorithm to realizing bias was the real problem. Internal competition/bake-off among ML engineers to find the right approach, inspired by Chris Bale's research on cross-partisan agreement.",
"timestamp_start": "00:47:34",
"timestamp_end": "00:50:34",
"line_start": 416,
"line_end": 437
},
{
"id": "topic_10",
"title": "Lightweight project management with Google Docs",
"summary": "Team coordination through a single, ever-growing Google Docs file rather than Jira or Asana. Daily meetings to discuss what's most important, dynamic goal-setting at natural milestones rather than quarterly OKRs.",
"timestamp_start": "00:50:47",
"timestamp_end": "00:54:00",
"line_start": 442,
"line_end": 466
},
{
"id": "topic_11",
"title": "Launch strategy: low expectations and incremental proof",
"summary": "Methodology of proving the concept at every step: showing mockups to diverse groups, MTurk testing, small pilot with 1,000 people, then gradual expansion. Focus on demonstrating quality rather than scale from day one.",
"timestamp_start": "00:05:53",
"timestamp_end": "01:09:19",
"line_start": 541,
"line_end": 557
},
{
"id": "topic_12",
"title": "The name change from Birdwatch to Community Notes",
"summary": "Product was initially called Birdwatch but Elon suggested changing to Community Notes (which was Keith's original name in the first Figma mockup). Discussion of why Community Notes is more intuitive and descriptive.",
"timestamp_start": "00:34:55",
"timestamp_end": "00:36:57",
"line_start": 331,
"line_end": 347
},
{
"id": "topic_13",
"title": "Core principles: voice of the people and radical transparency",
"summary": "Discussion of two foundational principles: (1) notes show based on community consensus, not company decisions—no button to change note status; (2) complete transparency—algorithm, code, and data are open source so anyone can verify and audit.",
"timestamp_start": "01:11:16",
"timestamp_end": "01:15:49",
"line_start": 574,
"line_end": 590
},
{
"id": "topic_14",
"title": "Technical challenges of open-sourcing the algorithm",
"summary": "Architectural decisions required to make the algorithm runnable by external users. Matrix factorization approach needs 500GB RAM and takes ~1 day to run locally, but makes the system verifiable by people like Vitalik Buterin.",
"timestamp_start": "01:15:50",
"timestamp_end": "01:17:11",
"line_start": 593,
"line_end": 614
},
{
"id": "topic_15",
"title": "Israel-Hamas conflict as stress test",
"summary": "October 2023 conflict created unprecedented misinformation deluge. Community Notes successfully handled it with 500 notes in first 3 days covering out-of-context images, fake video game footage, false claims. Speed improvements and image matching feature launched just in time.",
"timestamp_start": "01:18:26",
"timestamp_end": "01:22:46",
"line_start": 625,
"line_end": 638
},
{
"id": "topic_16",
"title": "Impact metrics: engagement and resharing drops",
"summary": "A/B tests show 30-40% drop in likes and reposts when notes appear. External research finds 50-60% reduction in overall reposts via difference-in-differences analysis. Authors 80% more likely to delete posts after being noted.",
"timestamp_start": "00:26:36",
"timestamp_end": "00:28:45",
"line_start": 268,
"line_end": 284
},
{
"id": "topic_17",
"title": "Shift to pseudonymous and anonymous contributors",
"summary": "Discovered that anonymity/pseudonymity actually improved the system: people more willing to write notes on controversial topics without fear of harassment, and more willing to cross partisan boundaries when anonymous than under real names.",
"timestamp_start": "01:26:26",
"timestamp_end": "01:28:48",
"line_start": 661,
"line_end": 695
},
{
"id": "topic_18",
"title": "Surviving leadership changes across four CEOs",
"summary": "Community Notes survived transitions from Jack Dorsey to Kayvon, then Parag, then Elon, then Linda Yaccarino. Key to survival: product's nature produces universally helpful output, team executed brilliantly, focused on outcomes and proving value at each step.",
"timestamp_start": "01:32:54",
"timestamp_end": "01:35:51",
"line_start": 736,
"line_end": 749
},
{
"id": "topic_19",
"title": "Low ego leadership and service orientation",
"summary": "Keith's approach focuses on solving the problem rather than personal credit. Doesn't care about name or control—willing to change from Birdwatch to Community Notes. Motivated by seeing the product help people and serve the community.",
"timestamp_start": "01:36:19",
"timestamp_end": "01:37:44",
"line_start": 754,
"line_end": 762
},
{
"id": "topic_20",
"title": "Future directions: more, better, faster notes with AI",
"summary": "Roadmap includes core product improvements (Community Notes bat signal feature), algorithm enhancements, and AI/LLM assistance. Supernotes project shows promise of external collaborators improving the algorithm. Vision for product built by the people, not just contributed to by them.",
"timestamp_start": "01:38:03",
"timestamp_end": "01:42:08",
"line_start": 772,
"line_end": 788
}
],
"insights": [
{
"id": "i1",
"text": "The key insight is that notes rated helpful by people who normally disagree indicate the note is probably accurate, neutral, and informative—this bridging-based agreement is what makes the system work at scale.",
"context": "Jay explaining why simple majority voting would fail but cross-partisan agreement indicates quality",
"topic_id": "topic_2",
"line_start": 73,
"line_end": 78
},
{
"id": "i2",
"text": "External studies show that when people see a post with a Community Note versus without, their agreement with the core claims actually changes—this proves fact-checking can change minds despite conventional wisdom.",
"context": "Jay citing independent research validating the product's effectiveness",
"topic_id": "topic_2",
"line_start": 237,
"line_end": 239
},
{
"id": "i3",
"text": "For a product built on crowdsourcing, you need to include all of humanity—not curate who participates. Having the full demographic diversity gives you the data to understand what's actually helpful to humanity.",
"context": "Keith explaining why open participation is essential, not a bug",
"topic_id": "topic_3",
"line_start": 244,
"line_end": 249
},
{
"id": "i4",
"text": "The worst possible mistake is showing a bad note because that undermines trust in the entire system. Trust is why people like the product, so it's better to miss good notes (showing 8% of proposed) than risk showing bad ones.",
"context": "Keith explaining the philosophy of high quality standards",
"topic_id": "topic_4",
"line_start": 226,
"line_end": 227
},
{
"id": "i5",
"text": "A myth that prevents leaders from taking bold action: the idea that managing more people or having a larger scope equals more impact. Building Community Notes had way bigger impact than running a large PM team managing OKRs.",
"context": "Keith reflecting on his career decision to go from leading 50+ people to small team",
"topic_id": "topic_6",
"line_start": 356,
"line_end": 362
},
{
"id": "i6",
"text": "The Thermal structure's key attributes: one clear driver/founder, one clear senior decision-maker outside the team, 100% team focus, ability to set own goals/cadence, and avoidance of standard practices like OKRs when they don't fit the work.",
"context": "Keith explaining the organizational structure that enabled success",
"topic_id": "topic_7",
"line_start": 385,
"line_end": 401
},
{
"id": "i7",
"text": "When everyone on a team is 100% focused, the iteration speed increases dramatically. A one-week delay for help completely changes project momentum. This focus multiplier is more valuable than adding more people.",
"context": "Keith explaining why small dedicated teams execute faster than matrix organizations",
"topic_id": "topic_7",
"line_start": 398,
"line_end": 404
},
{
"id": "i8",
"text": "People self-selecting into a project is more important than hiring talent. People who opt-in are bought into the goal, the mission, and the team—this intrinsic alignment makes the difference between success and failure.",
"context": "Keith reflecting on the importance of self-selection at team formation",
"topic_id": "topic_8",
"line_start": 484,
"line_end": 486
},
{
"id": "i9",
"text": "You can make radical organizational changes at large scale if you get people to opt-in. Elon's 'fork in the road' email asking people to self-select to stay created a company of people who wanted to be there.",
"context": "Keith observing that self-selection works even at company scale, not just for small teams",
"topic_id": "topic_8",
"line_start": 488,
"line_end": 494
},
{
"id": "i10",
"text": "When forced to operate with a very small team, deleting code becomes more important than writing it. Engineers often add maintenance burden through small wins without realizing the long-term cost.",
"context": "Jay explaining how lean teams force architectural discipline",
"topic_id": "topic_8",
"line_start": 523,
"line_end": 525
},
{
"id": "i11",
"text": "The algorithm emerged from an internal competition/bake-off approach. The team didn't know bridging-based agreement was the answer until they tested it against data from the pilot—proof that real data beats theory.",
"context": "Jay explaining how they discovered the right algorithmic approach",
"topic_id": "topic_9",
"line_start": 416,
"line_end": 437
},
{
"id": "i12",
"text": "Heavy task management systems (Jira, Asana) create their own overhead. A Google Doc that people naturally prune is lighter weight and keeps focus on actual priorities instead of backlog grooming.",
"context": "Keith and Jay explaining why they rejected standard PM tools",
"topic_id": "topic_10",
"line_start": 455,
"line_end": 458
},
{
"id": "i13",
"text": "Prove the concept at every step before expanding: mockups → usertesting → MTurk test → 1,000 person pilot → wider rollout. Each stage builds confidence that the next stage will work.",
"context": "Keith describing the systematic de-risking approach",
"topic_id": "topic_11",
"line_start": 541,
"line_end": 557
},
{
"id": "i14",
"text": "Set expectations low on launch so you manage surprise. Keith considered using a dumpster fire GIF to signal this was experimental, which shows thinking about user expectations and trust.",
"context": "Keith reflecting on launch positioning strategy",
"topic_id": "topic_11",
"line_start": 548,
"line_end": 554
},
{
"id": "i15",
"text": "The hardest principle to enforce: no button to change note status. If the company can override community votes, the system loses credibility. If a note is bad, the system design must be fixed, not the note.",
"context": "Keith explaining the voice-of-the-people principle and its controversial nature",
"topic_id": "topic_13",
"line_start": 575,
"line_end": 578
},
{
"id": "i16",
"text": "Complete transparency (open code, open data) is not a feature—it's essential to trust. Black box approaches to misinformation failed because people didn't trust them. Being completely open lets people verify you're not cheating.",
"context": "Keith explaining why radical transparency is fundamental",
"topic_id": "topic_13",
"line_start": 581,
"line_end": 587
},
{
"id": "i17",
"text": "Open-sourcing comes with real architectural costs. To let people run the algorithm themselves on downloaded data, you can't use typical distributed system patterns. You accept constraints to maintain principles.",
"context": "Jay explaining the engineering trade-offs of transparency",
"topic_id": "topic_14",
"line_start": 589,
"line_end": 596
},
{
"id": "i18",
"text": "People are more willing to cross partisan boundaries when anonymous than when using their real name. This is well-researched but counterintuitive—anonymity increases honesty about what you actually think.",
"context": "Keith explaining why pseudonymous contributions improved the system",
"topic_id": "topic_17",
"line_start": 668,
"line_end": 669
},
{
"id": "i19",
"text": "A product that produces information helpful to people across divides will survive leadership transitions because leaders of any ideology will see it works. The product's nature protects it from politics.",
"context": "Keith explaining why Community Notes survived four different leaders",
"topic_id": "topic_18",
"line_start": 737,
"line_end": 738
},
{
"id": "i20",
"text": "The team didn't get distracted during the acquisition chaos. Shipping every week and intense focus on the goal—making notes work—earned trust from leadership to continue supporting it.",
"context": "Keith explaining how execution excellence built credibility",
"topic_id": "topic_18",
"line_start": 742,
"line_end": 743
},
{
"id": "i21",
"text": "Cost savings was never mentioned as a motivation for Community Notes, even though it replaced fact-checking teams. The focus was always on solving the actual problem: helping people stay informed at scale.",
"context": "Keith correcting an external assumption about the project's motivation",
"topic_id": "topic_18",
"line_start": 748,
"line_end": 749
},
{
"id": "i22",
"text": "Low ego in leadership enables flexibility and focus. Caring about naming, credit, or control distracts from what actually solves the problem. Service orientation to the mission drives better decisions.",
"context": "Keith explaining his philosophy on ego and leadership",
"topic_id": "topic_19",
"line_start": 755,
"line_end": 756
},
{
"id": "i23",
"text": "Even on the most controversial topics, there's surprising agreement across political divides. While the world feels polarized, an 80% set of people probably agrees on quite a lot. This is reason for optimism.",
"context": "Keith reflecting on what Community Notes revealed about society",
"topic_id": "topic_19",
"line_start": 830,
"line_end": 833
},
{
"id": "i24",
"text": "Users will find uses for your product you never expected. Community Notes contributors debated Messi vs Ronaldo and added spam flags beyond misinformation. The product is truly driven by the people, not the designers.",
"context": "Jay noting unexpected user behaviors emerged organically",
"topic_id": "topic_20",
"line_start": 797,
"line_end": 800
},
{
"id": "i25",
"text": "The dream is that the scoring algorithm itself is written by the public, not just ratings contributed by them. Supernotes (external researchers building LLM-assisted note generation) shows this future is possible.",
"context": "Keith explaining the vision for community-built algorithms",
"topic_id": "topic_20",
"line_start": 787,
"line_end": 788
}
],
"examples": [
{
"id": "ex1",
"explicit_text": "At my previous company, we did X...",
"inferred_identity": "Twitter",
"confidence": 0.95,
"tags": [
"Twitter",
"social media",
"misinformation",
"product management",
"turnaround"
],
"lesson": "Shows the importance of identifying real-world problems within your company that you can work on",
"topic_id": "topic_6",
"line_start": 310,
"line_end": 320
},
{
"id": "ex2",
"explicit_text": "When the White House received a Community Note on a tweet, they deleted the tweet and reissued an updated statement",
"inferred_identity": "U.S. White House official account",
"confidence": 0.99,
"tags": [
"White House",
"government",
"impact",
"influential accounts",
"fact-correction",
"policy change"
],
"lesson": "Demonstrates the real-world impact: even the most powerful institutions changed their behavior based on crowdsourced notes",
"topic_id": "topic_6",
"line_start": 254,
"line_end": 257
},
{
"id": "ex3",
"explicit_text": "A post claimed 'A Palestinian boy shares his bread with a dog' but the image was actually a cat. A Community Note simply said 'That's a cat' with a Wikipedia link",
"inferred_identity": "Anonymous X user",
"confidence": 0.85,
"tags": [
"community moderation",
"humor",
"factual errors",
"user-driven",
"lightweight corrections"
],
"lesson": "Shows that Community Notes handles even trivial errors and proves the system is truly user-driven, not curated by experts",
"topic_id": "topic_1",
"line_start": 112,
"line_end": 113
},
{
"id": "ex4",
"explicit_text": "In the first three days of the Israel-Hamas conflict in October 2023, Community Notes created 500 notes covering out-of-context imagery, fake video game footage, and false claims",
"inferred_identity": "October 2023 Israel-Hamas conflict event",
"confidence": 0.99,
"tags": [
"Israel-Hamas conflict",
"misinformation crisis",
"scale test",
"real-time response",
"out-of-context media"
],
"lesson": "Demonstrated that Community Notes could handle the largest misinformation deluge ever, proving the system works under extreme stress",
"topic_id": "topic_15",
"line_start": 625,
"line_end": 630
},
{
"id": "ex5",
"explicit_text": "During the 2016 Trump-Clinton election, Lenny observed Twitter was where people were debating things that mattered and where information was being formed, but questionable information was also floating around",
"inferred_identity": "2016 U.S. Presidential Election / Twitter during election",
"confidence": 0.98,
"tags": [
"election",
"misinformation",
"political polarization",
"social media influence",
"problem identification"
],
"lesson": "Identifies how major political events on social media create information quality problems that need solving at scale",
"topic_id": "topic_6",
"line_start": 310,
"line_end": 315
},
{
"id": "ex6",
"explicit_text": "Vitalik Buterin published a blog post where he explored the Community Notes algorithm to verify it really does what it claims by running the open-source code",
"inferred_identity": "Vitalik Buterin, Ethereum founder",
"confidence": 0.95,
"tags": [
"Ethereum",
"crypto",
"transparency verification",
"external auditing",
"trusted validator"
],
"lesson": "Demonstrates that radical transparency enables credible third-party verification by respected figures in tech",
"topic_id": "topic_14",
"line_start": 614,
"line_end": 614
},
{
"id": "ex7",
"explicit_text": "Meta recently adopted Community Notes as their main fact-checking tool instead of having tens of thousands of fact checkers",
"inferred_identity": "Meta (Facebook parent company)",
"confidence": 0.99,
"tags": [
"Meta",
"Facebook",
"social media",
"fact-checking",
"adoption",
"industry standard"
],
"lesson": "Shows how a product's success drives adoption by other major tech companies, validating the approach across platforms",
"topic_id": "topic_1",
"line_start": 104,
"line_end": 104
},
{
"id": "ex8",
"explicit_text": "Someone took a screenshot of the Figma prototype of Community Notes and tweeted it, an NBC reporter wrote about it, and Elon Musk replied 'Definitely worth trying, IMO'",
"inferred_identity": "Elon Musk (before acquiring Twitter)",
"confidence": 0.98,
"tags": [
"Elon Musk",
"prototype feedback",
"influential support",
"pre-acquisition endorsement",
"serendipity"
],
"lesson": "Shows how early external validation from influential figures can come organically and presage future involvement",
"topic_id": "topic_12",
"line_start": 560,
"line_end": 563
},
{
"id": "ex9",
"explicit_text": "When Twitter staff proposed that journalists or select groups be the first Community Notes participants, Keith specifically rejected this to keep it open to everyone and maintain fairness",
"inferred_identity": "Twitter internal proposal/staff",
"confidence": 0.85,
"tags": [
"Twitter",
"editorial decisions",
"governance",
"design principle",
"inclusivity"
],
"lesson": "Demonstrates the discipline to reject easier, seemingly more credible paths (expert-first) to maintain core principles (openness to all)",
"topic_id": "topic_3",
"line_start": 100,
"line_end": 101
},
{
"id": "ex10",
"explicit_text": "After a note appears, authors are 80% more likely to delete their posts. High-quality notes are so effective that the original post gets removed, preventing other people from seeing the correction",
"inferred_identity": "X/Twitter users posting misinformation",
"confidence": 0.9,
"tags": [
"user behavior",
"misinformation authors",
"deletion rate",
"unintended consequence",
"post removal"
],
"lesson": "Reveals the unintended consequence of success: effective notes cause post deletion, removing the opportunity for others to learn from the correction",
"topic_id": "topic_16",
"line_start": 287,
"line_end": 287
},
{
"id": "ex11",
"explicit_text": "At the Spaces feature scale-up project, Elon asked the team to handle a million concurrent users, and a tiny team accomplished it in 2-3 weeks (what would have taken a year at Twitter 1.0)",
"inferred_identity": "X/Twitter Spaces feature team",
"confidence": 0.9,
"tags": [
"Spaces feature",
"infrastructure",
"rapid scaling",
"small team execution",
"ambitious deadline"
],
"lesson": "Demonstrates how lean, motivated teams with clear goals and Elon's decision-making authority can achieve in weeks what bureaucratic organizations take years to accomplish",
"topic_id": "topic_8",
"line_start": 500,
"line_end": 503
},
{
"id": "ex12",
"explicit_text": "Keith was recruited by Twitter when the company made an acquisition offer for his startup in 2016, during the Trump-Clinton election",
"inferred_identity": "Keith Coleman's unnamed startup (acquired by Twitter)",
"confidence": 0.9,
"tags": [
"startup founder",
"acquisition",
"career transition",
"startup to big tech",
"2016"
],
"lesson": "Shows how product leaders with startup experience can bring entrepreneurial mindset to large organizations",
"topic_id": "topic_6",
"line_start": 310,
"line_end": 311
},
{
"id": "ex13",
"explicit_text": "Jay had been working on personalizing push notification frequency, which drove significant DAU gains and put him on track for promotion with low risk",
"inferred_identity": "Twitter / X notification system",
"confidence": 0.85,
"tags": [
"notifications",
"engagement",
"DAU growth",
"personalization",
"low-risk project"
],
"lesson": "Illustrates the career risk calculation: safer projects with clear promotion paths versus higher-risk moonshot projects like Community Notes",
"topic_id": "topic_8",
"line_start": 482,
"line_end": 482
},
{
"id": "ex14",
"explicit_text": "An image post claimed something was happening in San Francisco but it was actually from a different city, taken years ago—this is the type of misleading out-of-context imagery that Community Notes frequently handles",
"inferred_identity": "Generic X/Twitter misinformation post",
"confidence": 0.7,
"tags": [
"out-of-context images",
"false location claims",
"visual misinformation",
"common pattern"
],
"lesson": "Shows the ubiquity of out-of-context imagery as a misinformation vector that Community Notes must address at scale",
"topic_id": "topic_1",
"line_start": 83,
"line_end": 86
},
{
"id": "ex15",
"explicit_text": "During the Israel-Hamas conflict, some posts featured fake battle footage created in the video game Arma 3 that looked realistic",
"inferred_identity": "October 2023 conflict misinformation",
"confidence": 0.85,
"tags": [
"Israel-Hamas",
"fake video",
"deepfakes",
"Arma 3",
"game footage",
"visual deception"
],
"lesson": "Shows sophisticated misinformation tactics using game engines to create believable fake footage that requires specialized knowledge to debunk",
"topic_id": "topic_15",
"line_start": 626,
"line_end": 629
},
{
"id": "ex16",
"explicit_text": "A note appeared on Elon's tweet itself, demonstrating that even the founder/CEO is not exempt from community oversight",
"inferred_identity": "Elon Musk's X/Twitter account",
"confidence": 0.85,
"tags": [
"Elon Musk",
"founder accountability",
"no exemptions",
"transparency",
"oversight"
],
"lesson": "Demonstrates that Community Notes' principle of equal application extends even to the most powerful account holder",
"topic_id": "topic_3",
"line_start": 14,
"line_end": 14
},
{
"id": "ex17",
"explicit_text": "Keith came back from paternity leave and told his boss Kayvon he wanted to stop managing the PM team and instead work on solving the misinformation problem",
"inferred_identity": "Kayvon Rostamizadeh, Twitter executive",
"confidence": 0.95,
"tags": [
"Kayvon Rostamizadeh",
"manager",
"mentorship",
"autonomy grant",
"project creation"
],
"lesson": "Shows a great manager's willingness to let top talent pursue their passion even if it means losing them from their current role",
"topic_id": "topic_6",
"line_start": 317,
"line_end": 318
},
{
"id": "ex18",
"explicit_text": "Jack Dorsey was CEO when Community Notes (then Birdwatch) was launched and had supported the exploratory phase",
"inferred_identity": "Jack Dorsey, Twitter CEO",
"confidence": 0.98,
"tags": [
"Jack Dorsey",
"Twitter",
"leadership support",
"product sponsorship",
"turnaround era"
],
"lesson": "Shows that even when leadership changes, projects with strong principle-driven execution and proven value can survive",
"topic_id": "topic_6",
"line_start": 326,
"line_end": 326
},
{
"id": "ex19",
"explicit_text": "Researchers outside Community Notes conducted A/B tests showing that posts with notes had 30-40% fewer likes and reposts, and difference-in-differences analysis showed 50-60% reduction in total reposts",
"inferred_identity": "UC Berkeley and other independent researchers",
"confidence": 0.9,
"tags": [
"academic research",
"validation",
"independent verification",
"A/B testing",
"misinformation reduction"
],
"lesson": "External peer-reviewed research provides credibility that internal metrics cannot, validating the product's impact on the most important metric: reducing misinformation spread",
"topic_id": "topic_16",
"line_start": 269,
"line_end": 272
},
{
"id": "ex20",
"explicit_text": "Lenny personally signed up as a Community Notes contributor and is rating notes but hasn't yet qualified to write them",
"inferred_identity": "Lenny Rachitsky, podcast host",
"confidence": 0.98,
"tags": [
"Lenny Rachitsky",
"contributor",
"rating",
"qualification barrier",
"personal experience"
],
"lesson": "Even prominent individuals must earn the ability to write notes, showing the meritocratic and democratic nature of the system",
"topic_id": "topic_3",
"line_start": 134,
"line_end": 140
}
]
}