TBPN Diet – April 8, 2026

Episode: Meta Tokenmaxxing, Intel Joins Terafab, Frontier AI vs. China
Hosts: John Coogan & Jordi Hays (with guest co-hosts Tyler & Chris)
Duration: ~30 minutes

Episode Overview

This episode of Diet TBPN dives into a whirlwind of major tech news:

• Meta’s internal obsession with “token maxing” and the resulting questions about Anthropic’s runaway revenue,
• Intel’s new partnership with Musk's Terafab to challenge semiconductor supply chains,
• The realities of corporate retreats (gone wild),
• Frontier AI arms races, including US tech giants banding together against model copying in China.

Energetic, irreverent, and packed with inside info, the hosts bring incisive analysis and first-hand Silicon Valley color to every discussion.

1. Meta’s “Tokenmaxxing” – Status Games in the Age of AI (00:00–12:10)

Main Story
Meta employees are competing to burn the most AI tokens, chasing “legend” status on a leaderboard called Claudonomics. This has led to rumors that Meta's internal token use represents a vast chunk of Anthropic’s skyrocketing revenue — sparking debate over incentives, waste, and AI-native company culture.

Key Discussion Points

• Tokenmaxxing ≈ Eyeball-Optimizing (00:40):

  • Tyler compares token leaderboard culture to the “eyeball” metrics of early web days: optimizing a metric so hard, its value becomes questionable.
  • Quote (Tyler, 00:55): “When a measure becomes a target, it ceases to be a good measure. Who knows what’s going on internally but Zuck’s pushing everybody to be AI-native. This guy loves spending money.”

• Goodhart’s Law & Meta’s Incentives (01:38):

  • Chris recounts Meta insiders saying the token leaderboard causes unhealthy gaming: “You do not want to be the guy at the bottom … You’re explaining your efficiency while someone else is just burning tokens in a loop.”

• Crunching the Real Numbers (03:00):

  • Headlines suggest Meta uses 60.2 trillion tokens in 30 days—claimed as 1/3 of Anthropic’s ARR.
  • Chris debunks this using cost-per-million-token rates: actual monthly spend likely closer to $55–$136M, not the rumored ~$1B.
  • Quote (Chris, 06:13): “If there are 30,000 engineers, it’s like $4,500 a month per engineer. That’s not absurd if you’re incentivizing usage.”

• Gaming the System (07:19):

  • Reports of Meta engineers writing bots to “burn tokens as fast as they can” because of the leaderboard.
  • Quote (Tyler, 07:19): “It’s similar to how Meta once measured engineering output by lines of code. Useful only if managers dig in, but lazy ones just look at the number.”

• Why Meta’s Bullish on Tokens (08:00):

  • John ties this spend to Meta’s strategic push: a multi-billion-dollar internal AI platform (MSL) is justified simply by saving on these token bills.
  • Quote (John, 09:03): “If they amortize training cost over internal use—they don’t need a viral AI app. Pure vertical integration pencils out.”

• Schizo Theory: Data Distillation (10:19):

  • Tyler suggests the tokenmaxxing is cover for a mass “distillation” run: rewriting all internal data for future model training.
  • John speculates on the legal/moral grey area: “Ship of Theseus world—if Meta pays Anthropic a billion to rewrite all internal data, do they own the tokens for next-gen models? Out-of-bounds per most enterprise contracts, but fascinating fuzziness here.”

Memorable Moments & Quotes:

“Ranking engineers by token spend is like ranking my marketing team by who spent the most money. Maybe we didn’t hit KPIs, but Joe spent $200,000 on a blimp that only flies over his house—he gets promoted to VP.”
— Christina at Linear via Tyler (07:19)

Takeaway:
Big tech is burning tokens—and cash—at unprecedented scale, but optimizing for easily-gamified metrics carries real risks. Meanwhile, Meta’s vertical integration play may be more about internal efficiency than external AI dominance.

2. Intel, Terafab, and the New AI Chip Wars (12:11–19:09)

Main Story
Intel just joined Elon Musk’s Terafab project alongside SpaceX, XAI, and Tesla—an audacious plan to build a trillion-watt/year fab in the US, breaking the industry’s dependency on TSMC and geopolitical risk.

Key Discussion Points

• Terafab’s Mission (12:33):

  • “Refactoring Silicon Valley fab technology” for ultra-high-performance AI/robotics chips, with on-site manufacturing/packaging in Austin.
  • John: “This is a step change. Chips for robo-taxis, Optimus, and even satellites— Musk wants compute everywhere.”

• Significance for Intel (13:35):

  • Once dominant, Intel has struggled with capacity and innovation. US government now owns 8.4% of Intel.
  • John: “Intel finally has a true demand-side story. TSMC isn’t scaling fast enough. Intel could break the bottleneck.”

• America’s Chip Security (15:26):

  • Tyler: “It’s good to get big buyers to commit to Intel and scale up alternatives to TSMC.”

• Data Centers in Space (17:23):

  • Deploying compute via Starlink satellites: tech’s not impossible, but economics/thermal issues aren’t solved at scale.
  • Chris: “There are GPUs in space today—H100’s even—but widespread space compute is an economic (not technical) frontier.”

Memorable Quote:

“There ARE computers in space! They do heat up because they’re routing packets, but humans have been moving heat around in orbit for decades.”
— John (17:29)

3. IPO Ticker Battles and Silicon Valley Games (19:09–20:22)

Quick Aside:
Musk plans to use “SPCX” as SpaceX’s ticker, reportedly acquired from ETF entrepreneur Matt Tuttle. Brief inside-baseball banter about ticker squatting and startup priorities.

4. Corporate Retreat Gone Wrong: Plex in Honduras (20:22–28:04)

Main Story
Plex, a remote tech company, flew 120 employees to Honduras for a “Survivor”-themed retreat—a case study in logistics fails and corporate culture gone wild.

Key Discussion Points

• Disaster Strikes (20:43):

  • Within days, the hotel GM and chef quit. On arrival: guard towers, machine guns, and unfiltered water warnings.
  • CEO became violently ill with E. Coli; arrivals greeted by an IV nailed to a bedpost.

• “Survivor” Hazing and Chaos (27:00):

  • Former Navy SEAL hired to run team-building drills: “Not a super fit group… people passing out, hit by fire ants, alligators, porcupines falling from the ceiling.”
  • Quote (John, 27:00): “It was fire festival for corporate retreats.”
  • Quote (Tyler, 28:04): “A FIRE festival of corporate retreats.”

Takeaway:
A cautionary tale: forced fun plus exotic locations can quickly become a nightmare.

5. AI Security, Model Copying & US-China Tech Rivalry (28:06–End)

Main Story
Anthropic previews “Mythos,” a model for bug-hunting in critical infrastructure, to selected partners (Amazon, Apple, Google, Microsoft, Linux Foundation), reflecting rising concerns over AI-driven cybersecurity and model copying.

Key Discussion Points

• B2B Go-to-Market Play (28:06):

  • Anthropic’s Mythos is about more than tech: it’s about adoption inside major orgs and preventing “AI apocalypse” scenarios.

• Big Tech Unites on AI Safety (29:00):

  • OpenAI, Anthropic, and Google share info to fight Chinese adversarial model “distillation”—the process of copying/miniaturizing cutting-edge US models.
  • Quote (John, 29:47): “The rare collaboration underscores the severity of the concern … Chinese competitors extracting results from US models to gain an edge and threaten national security.”

• Model Commoditization vs. Revenue Boom (31:48):

  • Even if copying accelerates, the demand for frontier models is insatiable—Anthropic now at $30B run rate (!).
  • John: “Commoditization is more of an AI safety problem… geopolitically, if other countries move faster there’s little incentive for US labs to slow down.”

Notable Quotes & Memorable Moments

• Meta’s token maxxing culture:

  “You don’t want to be at the bottom of the leaderboard … Someone else just set up an agent that counts one to a million, over and over.” — Tyler (02:25)

• Data distillation and legal grey zones:

  “If Meta pays Anthropic a billion to rewrite all internal data … do they own those tokens for training? It’s a Ship of Theseus world.” — John (10:41)

• Fire festival for retreats:

  “Former Navy SEAL to haze the team … it was 100 degrees, people passing out, fire ants, porcupines. Fire festival of corporate retreats.” — John & Tyler (27:57–28:04)

Key Timestamps

• [00:00–12:10] — Meta’s tokenmaxxing, incentives, engineering culture, and internal AI strategy
• [12:11–19:09] — Intel joins Terafab: the coming US AI fab boom, chip geopolitics, and space data centers
• [19:09–20:22] — IPO ticker drama and insider Silicon Valley games
• [20:22–28:04] — Plex’s corporate retreat in Honduras: disaster and hilarity
• [28:06–End] — Anthropic’s Mythos, AI cybersecurity arms race, and US-China model copying

Conclusion

Today’s Diet TBPN episode blends hard AI numbers, high-stakes geopolitics, and the sometimes-absurd human side of big tech. Whether discussing how “tokenmaxxing” might distort incentives at Meta or how global AI rivalry is reshaping alliances and infrastructure, the show delivers punchy analysis and industry gossip with a healthy dose of skepticism and good humor.

For more:
Subscribe to TBPN, follow the live stream, and check out their newsletter.