← All episodes

Tokenmaxxing, SF Street Name Auction, Corporate Retreat Gone Wrong, Citrini

| 35 products mentioned
TBPN TBPN host
Adita Bandi guest
Zack Shore guest
Watch on YouTube ai spending and budgeting model distillation and security semiconductor manufacturing infrastructure development community-driven projects ai safety and responsible disclosure geopolitical ai competition

TBPN covers major AI industry developments including Meta's token-maxing internal competition that reveals massive spending on Claude API usage, Intel's partnership with Elon Musk's Terrafab chip manufacturing project, and the rise of model distillation as Chinese competitors copy frontier AI models. The episode also features Riley Walls discussing the successful auction of a San Francisco alley street naming rights, demonstrating how founders can monetize unusual assets through community engagement and viral mechanics.

Key takeaways
  • Meta employees are competing on internal token leaderboards, but actual spending is likely $1,800–$4,500 per engineer per month rather than the viral $1 billion/month claim, suggesting token budgets will become a standard engineering cost line item.
  • Model distillation—extracting outputs from frontier models to train cheaper competitors—is becoming a national security concern, prompting OpenAI, Anthropic, and Google to collaborate through the Frontier Model Forum and implement tiered KYC (Know Your Customer) protocols at scale thresholds to prevent unauthorized copying.
  • Intel's partnership with SpaceX, Tesla, and XAI on Terrafab signals the beginning of demand-side commitments to domestic semiconductor manufacturing, addressing the long-standing problem that ASIC companies couldn't escape Taiwan despite geopolitical pressure.
  • Anomalies and overlooked assets (foreclosed alleys, payphones, government spending data) can be monetized at scale through framing, community participation, and viral mechanics—Riley Walls' alley project generated $135k+ through free art submissions combined with a naming rights auction.
  • Anthropic's Mythos model is being distributed to 50+ critical infrastructure companies (Amazon, Microsoft, Google, Apple) to find and patch software vulnerabilities before malicious actors exploit them, establishing a responsible disclosure go-to-market that builds goodwill and product adoption simultaneously.
  • The data center opposition incident in Indianapolis highlights the political risk of infrastructure projects; builders should reframe community benefits beyond job counts to address environmental and quality-of-life concerns before projects face organized resistance.

Recommendations (8)

ChatGPT
ChatGPT uses

"if you opened ChatGPT and asked it to summarize"

TBPN · ▶ 10:55

Claude
Claude uses

"Meta has a version of open claw called my claw"

TBPN · ▶ 14:34

Databricks
Databricks uses

"We were early customers of Databricks and Snowflake as an example that let us invest in those companies"

TBPN · ▶ 2:49:42

Snowflake
Snowflake uses

"We were early customers of Databricks and Snowflake as an example that let us invest in those companies"

TBPN · ▶ 2:49:44

ChatGPT
ChatGPT uses

"I look almost every day at the chart of ChatGPT users download share how it's weathering the storm"

TBPN · ▶ 2:50:13

New York Times

"I learned as much about that industry reading the New York Times on Mondays, which was kind of the digital edition"

TBPN · ▶ 2:19:33

Vanity Fair

"reading Vanity Fair and Premier Magazine and anything I could get my hands on"

TBPN · ▶ 2:19:40

"reading Vanity Fair and Premier Magazine and anything I could get my hands on"

TBPN · ▶ 2:19:40

Mentioned (27)

XKCD
XKCD "we got to pull up this comic from XKCD in the comments here" ▶ 7:05
Meta (Facebook)
Meta (Facebook) "Meta employees are apparently token maxing and competing on an internal leaderboard called Claudo..." ▶ 5:22
Anthropic
Anthropic "Anthropic just passed 30 billion in run rate revenue with probably the steepest revenue growth ch..." ▶ 5:47
Blackmagic
Blackmagic "they've done a lot of standard setting around cinema camera gear. They were a big proponent of th..." ▶ 1:21:03
Dropbox "someone like Dropbox could integrate with us and start rendering documents" ▶ 1:22:55
Ralph Lauren
Ralph Lauren "Ralph sent me and he said I could get whatever I want" ▶ 2:25:18
John Steinbeck "John Steinbeck which was my favorite writer and I was reading a biography of him at the time" ▶ 2:30:31
iPod
iPod "the iPod was starting to really gain traction and an iPod was a semiconductor product" ▶ 2:34:50
Evernote
Evernote "Evernote, I think, was one of the first" ▶ 2:43:58
Box "So, we did deals like Evernote and Box" ▶ 2:44:08
Snapchat
Snapchat "our most successful deal is one that broke all of the rules, which was Snapchat" ▶ 2:44:15
Onavo "this amazing data set called Onavo which actually gave you engagement data from users on phone an..." ▶ 2:50:24
Cerebras
Cerebras "that led us to believe the series being Cerebras that I think will be kind of a generational kind..." ▶ 2:52:17
Nvidia
Nvidia "when we saw what Jensen was building and the momentum that Nvidia was building in the data center" ▶ 2:52:32
TSMC
TSMC "You know I've personally been in Taiwan many times, visited TSMC" ▶ 2:52:51
Netflix "I mean for a long time we used Netflix" ▶ 3:01:49
Cursor
Cursor "How much of my spend per developer on cursor openai and anthropic is happening" ▶ 3:00:10
OpenAI
OpenAI "OpenAI is probably the most important company in the world today in the sense that it's the drive..." ▶ 2:49:58
Intel
Intel "Intel is joining Terafab project with SpaceX, XAI, and Tesla to help refactor silicon fab technology" ▶ 20:18
Plex
Plex "Technology company Plex took its 120 employees to Honduras for a week-long bonding experience" ▶ 34:38
Spotify
Spotify "we scraped a bunch of data about the US like how much money the US government spends and we'll ho..." ▶ 1:15:15
paint.com
paint.com "paint.com/auction" ▶ 1:34:34
Void "Netflix's new project Void. The AI removes objects from videos but it even corrects the physics a..." ▶ 1:19:14
Yahoo Mail
Yahoo Mail "Yahoo bought it for Yahoo mail so we can part of the whole attachment layer for document" ▶ 1:23:00
Codex
Codex "we don't need any of that. We directly sit on top of your product code" ▶ 1:27:16