TBPN Episode Summary: “2025 in Review, Cursor Acquires Graphite, TikTok's $50B Profit”

Hosts: John Coogan & Jordi Hays
Date: December 19, 2025
Special Guests: Michael Truell & Merrill Lutsky (Cursor & Graphite), Pranav Myana (space data centers), Anna Goldie (Recursive Intelligence), Edward Mehr (Machina Labs)

Episode Overview

This festive year-end episode of TBPN reflects on major technology, business, and cultural stories of 2025. Broadcasting live from the TBPN “ultradome,” John and Jordi celebrate the holidays with costume antics, recap the show’s growth, highlight industry-defining moments, and welcome a series of top-tier guests. Central topics include Cursor’s acquisition of Graphite, ByteDance/TikTok’s explosive profits, the evolving AI frontier, and the future of hardware, manufacturing, and compute, both on Earth and in space.

Show Highlights & Year-in-Review

00:00–15:00

• The hosts kick things off in full holiday spirit, reflecting on the show’s format, audience, and the camaraderie that built TBPN:

  “We interviewed 912 unique guests and ... did 225 livestreams this year... almost hit a thousand guests.” – John Coogan [04:46]

• Notable guests and partners are thanked; the show grew with support from sponsors like Ramp, Adquick, and Public:

  “They took a huge gamble on us when the show was really, really small.” – John Coogan [05:21]

• TBPN’s distinct style—fast-paced, meme-rich, heavy on community engagement—is celebrated (“We democratized the trading card this year...”).

• Recap of some signature 2025 media events:

  • Streaming from YC Demo Day at Gary Tan’s invite
  • Figma IPO coverage at the NYSE; focusing on story/cultural impact over price action

Notable Segments & Quotes

Cursor Acquires Graphite: Vision for Devtools

[24:26–47:29]
Guests: Michael Truell (Cursor), Merrill Lutsky (Graphite)

• Why the Acquisition?

  • Both companies shared a vision for radically automating software development, but from “complementary angles”—Cursor focusing on the “single-player” (individual developer) and Graphite on the “multi-player” (team/code review) experience.
  • The New York offices “literally across the street” made integration feel natural.

  “We both believe that the way people build software ... is going to change radically. A lot of coding as we know today will be automated.” – Michael Truell [26:04]

  “Cursor has so dramatically changed ... how much code engineers can generate. The bottleneck has now shifted to what we call the outer loop—reviewing, validating, merging changes.” – Merrill Lutsky [28:35]

• 2025 as the Year of AI Software Agents

  • Agents became “useful in a professional setting.” Shift from AI answering quick questions to handing off whole tasks.
  • The biggest value-add: integrating code writing and code review so developers aren’t switching contexts or tools.

  “You shouldn't have to jump to a different tool for your editor, for code review, ... all this should just be one nicely integrated surface.” – Merrill Lutsky [29:24]

• Integration & Roadmap

  • Immediate plans: shared surfaces between coding and PR/review, surfacing agent logs in code review.
  • Ambitious long-term build to be revealed late 2026.

• Enterprise vs. SMB Adoption of AI Devtools

  • Mid-market demand has “compounded consistently,” but upmarket/enterprise adoption is outpacing expectations.
  • 64% of the Fortune 500 “pay [Cursor] in some way.”

    “It’s shocking how many companies are software companies. Even Starbucks, PwC, Hilton, ... are deep customers.” – Michael Truell [33:41]

• Bottlenecks & Change Management

  • Biggest barriers are not technological but behavioral and process-oriented—training teams to change their workflows.
  • Resistance is mainly in “super legacy languages” and conservative procurement processes.

• Advice to Aspiring Developers

  “It’s actually a really exciting time ... especially for people who are new, because it's quick for them to pick up new habits.” – Michael Truell [36:17]

• Research Frontiers: What’s Next for AI Coding Tools?

  • More ‘games’ for models to play = better task automation
  • Looking forward to a split between in-the-loop (pair programming) and async agent usage.

  “We’d like to make it possible for anyone to build anything they’d like on a computer. ... Automate coding.” – Michael Truell [43:27]

• Dashboard vs. Pipes (Graphite’s Positioning)

  “We want to be the homescreen for developers ... everyone comes in and checks, ‘what do I have to do to unblock my team?’” – Merrill Lutsky [45:28]

TikTok/ByteDance Hits $50B Profit

[17:09–22:26]

• ByteDance on track for ~$50B in profit in 2025; now nearly as large/profitable as Meta.

• Diversified Chinese business (Douyin, Toutiao, CapCut, etc.), not just TikTok.

• US TikTok sale: Oracle, Silver Lake, MGX to own 45% of US entity, ByteDance down to 20%.

  “ByteDance ... is now basically the same size as Meta, which is insane.” – John Coogan [17:40]

• US sale means the new venture will focus on retraining content moderation algorithms exclusively on US data.

AI Image/Video Model Wars: Meta’s “Mango”

[22:26–24:22]

• Meta is developing new AI models (“Mango” and “Avocado”) specializing in image and video.
• The hosts debate whether these can rival Instagram-embedded startups like Midjourney, given Meta’s massive data and compute.

AI Speed vs. Intelligence, OpenAI “Code Red” & the New Arms Race

[64:09–69:05]

• After Google’s Gemini upstaged ChatGPT in benchmarks, OpenAI reportedly called “Code Red”—not for the first time.

• Hosts argue it's more about optics/internally rallying focus than true existential crisis.

  “If you declare a ‘Great Lock In,’ everyone’s excited... but if you declare ‘Code Red,’ and you’re already at the top, it makes you seem shaky.” – John Coogan [65:41]

• 2026 prediction: AI user experience hinges on speed, not just smarts.

  “Customers cannot tell the difference between 120IQ, 130IQ chatbot... What can they tell? Speed.” – Jordi [67:37]

Space Data Centers & Energy Constraints

[76:14–88:23]
Guest: Pranav Myana

• Why space data centers? Compute demand will soon outstrip land, water, and power availability on Earth.

• Cooling a massive constraint in space—but solvable with radiators, engineering.

• Timeline debate: Pranav and (via Musk tweet) suggests feasible in under three years, at least for inference workloads.

  “If you’re betting against space data centers, you’re betting against compute to grow.” – Pranav Myana [77:07]

• The “future is optical orbital data centers—photonics with 10–20x lower heat constraint than electronics.”

Recursive Intelligence: AI-Driven Chip Design

[94:12–107:44]
Guest: Anna Goldie, co-founder/CEO

• Background: Anna’s roots at MIT, Stanford, Google Brain, Anthropic.

• Mission: “AI for chip design and chip design for AI.” Recursive’s three phases:

  1. Accelerate physical chip design (years → months)
  2. End-to-end AI-designed custom chips (from model to fab-ready)
  3. Eventually enable a Cambrian explosion of custom chips for many workloads

• TSMC could have 10–100x more design customers as chip design becomes more accessible.

  “There’s a huge space of chips that could exist and maybe should.” – Anna Goldie [99:41]

• Hardware–software co-design (like Google’s TPUs matched to Gemini models) is a secret edge over just scaling up models.

Machina Labs: Distributed, Flexible Metal Manufacturing

[117:42–126:18]
Guest: Ed Mehr, co-founder/CEO

• Machina’s “robocraftsman” containers can be shipped and set up anywhere to manufacture metal goods on-demand (e.g., drone skins, automotive customizations).

• Vision: Distributed manufacturing for defense and allied production—counterpoint to centralized, slow plants.

• Partnership with Toyota for custom car panels; moving to scale (next facility = 50 units).

  “If you want throughput, you replicate horizontally, 50–100 units ... all defense, all heavy equipment, this is a good choice.” – Ed Mehr [122:08]

Fun & Memorable Moments

• The “Trading Card” Meme
  Tech recruiting is now like “college football signing day,” with trading cards for elite engineering interns:

  “We democratized the trading card this year and I’m glad we did.” – John Coogan [17:14]

• Meta AI Boardroom, Zuckerberg as “Gangster”
  Boardroom drama and Meta’s AI hiring spending (“$65 million avg. salary for 10 engineers” – possibly stock inclusive):

  “He’s an animal.” – Michael [48:26]

• Onion Futures Act and AI Crime
  Anthropic model “eager” to trade illegal onion futures; tangent into the real-life “onion ring” finance scandal of 1955.

• Robot Solves Rubik’s Cube in 0.1 seconds
  Proof even childhood talents are no match for automation:

  “If your job was doing Rubik’s cubes, you’re cooked.” – John Coogan [129:33]

• Hosts Celebrate Team, Audience, and Each Other
  Multiple moments of gratitude and “one last gong” for the community to close out 2025:

  “It’s an honor to build this show with the team and with all of you in the audience.” – Jordi [140:19]

Timeline of Key Segments

| Time | Segment | |------|---------| | 00:00–07:50 | Hosts’ intro, holidays, show stats, year in review | | 07:50–17:09 | Community engagement, trading card meme | | 17:09–22:26 | TikTok/ByteDance’s $50B profit, US sale | | 22:26–24:22 | AI model wars, Meta’s “Mango” model | | 24:26–47:29 | Cursor acquires Graphite: full-stack AI devtools vision (Michael, Merrill) | | 47:29–54:11 | App store market share, Meta distribution | | 54:11–69:05 | OpenAI “Code Red”; AI in 2026 is all about speed | | 69:05–76:11 | Government–AI partnerships, Stargate compute | | 76:14–88:23 | Space data centers, photonics (Pranav) | | 94:12–107:44 | Recursive Intelligence, AI for custom chips (Anna) | | 117:42–126:18 | Machina Labs, distributed manufacturing (Ed) | | 129:33–139:19 | “Robot job killer,” penny funeral, watches, numbers go up | | 139:19–140:52 | “Shrek yourself,” hosts’ heartfelt year-end thanks, final gong |

Closing: Reflections & Takeaways

TBPN’s final broadcast of 2025 is a whirlwind of gratitude, tech optimism, inside jokes, and live interviews with founders pushing the frontier on devtools, AI hardware, and infrastructure. The show’s playful, irreverent style is evident throughout, balanced by deep discussions about the shifting landscape of tech, AI adoption, and cultural moments that shaped the year.

“Thank you so much for all the support. ... It’s an honor to build this show with the team and with all of you in the audience.” – Jordi [140:19]

Host sign-off:

“One last gong for 2025. What a year. Merry Christmas, Happy New Year. We will see you in 2026.” – John Coogan [140:45]