CoRecursive: Coding Stories

Episode: From Hacker News to TikTok – How Algorithms Learned to Hook Us
Host: Adam Gordon Bell
Guests: Liam (Co-host), Corey (Developer, special guest)
Date: March 2, 2026

Episode Overview

In this episode, Adam takes listeners on a story-driven exploration of how recommendation algorithms have evolved from simple community voting systems on early internet platforms to highly personalized, real-time addictive mechanics fueling modern social media like TikTok and Instagram Reels. Adam, joined by Liam and Corey, unpacks how these algorithms influence our content diets, habits, and even mental health—illustrated through Corey’s struggle with AI-generated cat videos clogging his feed. The conversation moves from early aggregator sites like Hacker News and Reddit through Facebook’s social graph, YouTube’s collaborative filtering, and finally to the cutting-edge, real-time personalization powering TikTok, reflecting on the impacts for both adults and the next generation.

Key Discussion Points & Insights

1. Corey's Cat Video Problem: Setting the Stage

00:07 - 01:46

• Corey recounts how his feed on Instagram Reels became saturated by AI-generated cat videos after he began watching and sharing them.
  • "We're cat people... my feed just turned into nothing but AI slop cats." – Corey [00:16]
• He finds himself scrolling for hours, stuck in a loop engineered by an algorithm that learned his preferences.
• Adam and Liam ponder: If a developer like Corey can't beat the algorithm, can anyone?

2. The Origins – Simple Algorithms, Shared Spaces

01:59 – 06:23

• Adam traces the earliest recommendation systems:
  • Blogs: User selection and subscription.
  • Slashdot, Digg, Reddit, Hacker News: Content surfaced via simple upvote mechanisms and time decay—compared to parallel games of Flappy Bird.
    • “All the people on the Internet are like pressing the space bar, raising [posts] up or down, and the gravity is pulling them down over time.” – Adam [03:21]
• Benefits: Created a unified community front page—shared context, no machine learning or personalization, just “votes and time.”
• Limitation: No personalization, simple consensus-finding within small communities.

3. The Rise of Controversy – Engagement Beyond Consensus

06:23 – 09:49

• Introduction of "Sort by Controversial" on Reddit:
  • Surfaces posts with both lots of upvotes and downvotes, often leading to detailed debates but also inflaming tempers.
  • Inspired by Scott Alexander's short story, "Sort by controversial," where algorithms end up generating the most divisive content.
    • “You can find these stories... that just cause such visceral reactions that people can't help but interact with it.” – Adam [08:25]
• Broader Insight: Algorithms that optimize for “interaction” or “controversy” can inadvertently amplify divisive content by simply measuring what gets the most attention, not what’s good for the community.

4. Facebook and the Power of Social Comparison

09:53 – 14:33

• Transition to Social Feeds:
  • Early Facebook's News Feed leveraged natural human tendencies toward gossip, comparison, and envy.
    • “The news feed automated [gossip]... It’s like, ‘here's all the people you're following, and here's the things that they're saying.” – Adam [11:16]
  • Simple algorithm (“EdgeRank”): Prioritized closeness, post type, and recency.
• Critical Point: The algorithm’s power stemmed from our social wiring, not its technical sophistication.
  • “The thing that made [Facebook's feed] so powerful isn’t this great insight that the algorithm has. It's actually... in us.” – Adam [13:13]
• Contrast: Still not explaining personalized, solitary scrolling experiences like Corey's.

5. Engagement at All Costs – The Facebook ‘MSI’ Pivot

14:33 – 21:54

• Haugen Leaks: 2017 Facebook data revealed a shift:
  • To combat declining engagement, Facebook began optimizing for 'Meaningful Social Interactions' (MSI)—comments, shares, reactions.
    • “Set a metric and you optimize it, and the number goes up, but there are side effects you can't see on the dashboard.” – Adam [16:47]
  • Unintended consequence: Content generating outrage (not delight) gets amplified (“sort by controversial” at scale).
    • “By optimizing for engagement, they had built this system that just finds the most divisive and angry things, that finds across 2 billion people...” – Adam [17:36]
  • Company Dilemma: Algorithmic divisiveness drives growth, but tries to tune for less anger “would hurt numbers.”
    • “You’ve actually found the best growth metric... I don’t know how we turn it down without hurting our numbers.” – Netflix engineer (paraphrased by Adam) [18:47]
• Takeaway: Engagement metrics, even well-intentioned, can warp platform dynamics, fueling toxicity for the sake of retention and profit.

6. Unlocking Addictiveness – From YouTube to TikTok

21:54 – 29:58

• YouTube’s Collaborative Filtering:
  • Focus on maximizing watch time, not clicks; builds a “model” of user interests based on video history, using complex but conceptually clear math.
    • “They take the last 50 videos you watched... and then they find other people near that [interest] spot and... show you the next [video].” – Adam [24:38]
• Limitation: Still relatively slow to adapt—weeks rather than minutes.
• TikTok’s Real Innovation:
  • Signal Density: With short-form video, platforms get exponentially more signals from each user in less time, allowing rapid personalization.
  • Real-time Adaptation: TikTok's leaked algorithm (“Monolith”) updates its model of the user every second for the last 30 minutes—triggers immediate content adjustments.
    • “Every time you're interacting, they're real time updating that model of you of what you might want to watch next.” – Adam [28:47]
  • Demo Metaphor: YouTube as a chef learning your taste from one big burger; TikTok as a sushi train chef adjusting every few bites.
    • “A sushi train... the chef can see that you really like the chicken teriyaki sushi... and start piling on more.” – Liam [31:23]

7. Case Study – TikTok’s Algorithm in Action (Wall Street Journal Test)

31:44 – 33:53

• Experiment: WSJ researchers created bot personas to see how fast TikTok could ‘learn’ their interests (e.g., sadness, politics).
  • “In just 3 minutes, 15 videos in... by the time that 30 minute was in place, Kentucky96 had seen 224 videos and 93 were about people who were depressed or thinking of suicide.” – Adam [31:54]
• Consequence: The algorithm can rapidly funnel users into niche or even unhealthy content bubbles—with no intent, just relentless reinforcement.

8. The Feedback Loop and Impact on Kids

33:53 – 39:37

• Parenting Dilemma: Corey worries about his son’s multitasking and inability to focus, toggling between Roblox and Youtube Shorts.
  • “I'm worried that I'm raising kids who don't understand how to actually focus on something for longer than four or five seconds.” – Corey [34:18]
• Does the algorithm “lock in” interests?
  • Adam discovers resetting recommendations in Instagram/Reels is possible, offering a partial “escape hatch.”
    • “You can clear your recommended content across Explorer reels and feeds. So that's your key. It's just under Content preferences Reset suggested content.” – Adam [35:27]
• Cheesecake Analogy: Adam likens social media to ultra-palatable food—addictive because it amplifies what we already crave.
  • “Social media is like cheesecake... it's addictive because it's things that we like and enjoy.” – Adam [36:20]
• On solutions: Resets and moderation help, but fighting craving is a human challenge, not entirely technical.

9. Regulation and Responsibility

40:01 – End

• Emerging Legal/Political Consequences:
  • Mark Zuckerberg testifies in court about Instagram’s addictiveness; internal docs reveal growth often won over user well-being (e.g., delaying interventions that would benefit kids’ sleep).
• Philosophical Takeaway:
  • “It’s easy to hear things like that... and think there's your villain. But I don't think it's the right lesson... every platform is tapping into something innate in us. That cheesecake is always going to get made because we want it.” – Adam [40:37]
• Closing Note:
  • Regulation (like for alcohol or cigarettes) may be necessary, as algorithms are only as powerful as the human cravings they serve.

Notable Quotes & Memorable Moments

• “All the people on the Internet are like pressing the space bar, raising [posts] up or down, and the gravity is pulling them down over time.” – Adam [03:21]
• “You can find these stories... that just cause such visceral reactions that people can't help but interact with it.” – Adam [08:25]
• “The thing that made [Facebook's feed] so powerful isn’t this great insight that the algorithm has. It's actually... in us.” – Adam [13:13]
• “Set a metric and you optimize it, and the number goes up, but there are side effects you can't see on the dashboard.” – Adam [16:47]
• “Every time you're interacting, they're real time updating that model of you of what you might want to watch next.” – Adam [28:47]
• “A sushi train... the chef can see that you really like the chicken teriyaki sushi... and start piling on more.” – Liam [31:23]
• “Social media is like cheesecake... it's addictive because it's things that we like and enjoy.” – Adam [36:20]
• “That cheesecake is always going to get made because we want it.” – Adam [40:37]

Important Segment Timestamps

• [00:07] – Corey’s cat video dilemma, setting the problem.
• [03:13] – Flappy Bird and gravity as metaphor for simple algorithms.
• [08:24] – Reddit’s “sort by controversial” amplifies divisiveness.
• [13:13] – Facebook’s success tied to human nature, not technical wizardry.
• [16:44] – Facebook’s engagement optimization backfires.
• [24:55] – YouTube’s vector-based collaborative filtering.
• [28:47] – TikTok’s real-time adaptation and signal density.
• [31:23] – Sushi train analogy for TikTok’s algorithm.
• [31:54] – Wall Street Journal TikTok bot experiment, rapid discovery of user interests.
• [34:18] – Corey’s concern about kids’ attention spans.
• [35:27] – How to reset Instagram’s recommendation algorithm.
• [36:20] – The “cheesecake” nature of addictive content.
• [40:37] – Regulation, responsibility, and the innate human cravings algorithms exploit.

Final Thoughts

Adam and guests deftly trace how internet algorithms shifted from facilitating community interests to hijacking our time and attention with precision-targeted, instantaneous feedback loops. The technology is both brilliant and unnerving—deeply intertwined with human nature and our vulnerabilities. The conclusion: Understanding the machinery is only a first step; personal boundaries, parental vigilance, and perhaps even regulation are needed guardrails, because the algorithms, like the allure of cheesecake, will always be there, waiting to be indulged.