Receipts or Regrets | Episode 33

Scotty predicted robots in homes by 2025. While Tesla's Optimus and Figure didn't deliver, Chinese manufacturers came through with $20K humanoid robots available for purchase. Called a 'slop receipt' - right prediction, wrong company.

• •Chinese companies delivered $20K home robots while US companies (Tesla, Figure) failed to ship
• •Elon fired Optimus head 6 months ago, completely restarting the architecture
• •Timeline now pushed to 2028 for proper Western robotics products
• •Demonstrates importance of tracking who executes, not just the trend

Scotty's proposal for Australia to leverage $20 trillion in raw materials (critical for AI infrastructure) by inviting tech giants to build data centers in exchange for equity and free compute for citizens. Government has no AI strategy three years into the hype cycle.

• •Australia has $20T in raw materials needed for AI infrastructure
• •Proposal: Tender access to materials in exchange for data centers, citizen equity, and free compute
• •Government banning social media but has no AI czar, adviser, or national strategy
• •Escalating the call - need to 'shout it from rooftops' as opportunity window closes

Scotty called out Guzman & Gomez's absurd valuation at IPO ($5M per store, 1000x multiple). Stock has crashed 55% from $45 peak to $20, validating the 'no AI in the avocado' thesis.

• •IPO valued at $5M per 100 sqm store - required 'AI in the avocado' to justify
• •Stock down 55% from February peak of $45 to current $20
• •Market cap dropped from $4B to $2B - classic overvaluation correction
• •Demonstrates importance of fundamental analysis over hype in tech-adjacent businesses

Scotty's AI wearable pendant company Limitless got acquired by Meta. The device recorded his argument with his dog Hank, which AI summarized as workplace conflict. Now Zuckerberg has all the wearable data, and the product is canceled in countries with strict privacy laws.

• •Limitless changed terms/privacy policy, then announced Meta acquisition one hour later
• •Product canceled in UK, Brazil, and other countries that don't accept Meta's privacy policy
• •Device was a year late, had terrible battery life, and no 'Find My' feature
• •AI misidentified dog argument as workplace conflict with 'coworker Hank'
• •Regret: Early wearable purchase now feeds data to Meta's surveillance apparatus

Brett Adcock's claim that Figure AI would be worth 200x Apple ($600-800 trillion) is looking increasingly absurd. Three years in with no product, just capital burns and robot raves. Both hosts regret their earlier fanboying.

• •Figure AI has no product after 3 years, just expensive parties (Deadmau5 robot rave)
• •200x Apple = $600-800 trillion target vs Apple's $416B actual revenue
• •Figure AI revenue appears minimal compared to ambitious targets
• •Waiting for 'talent hires working directly for Brett' before reassessing
• •Receipt turned to regret: Capital efficiency matters more than hype

Both predicted Tesla would win the robotaxi race long-term despite Waymo's current lead. Tesla can produce robotaxis at 1/10th the cost. Waymo's 'hat' strategy allows retrofitting any car for autonomy, potentially partnering with OEMs.

• •Tesla's economics: can produce robotaxi for ~10% of Waymo's cost
• •Waymo's retrofit strategy with the 'hat' enables partnerships with existing car manufacturers
• •Room for two players: Elon's polarization creates 30% of market that won't use Tesla
• •Waymo can capture European market, left-leaning customers, and B2B OEM partnerships
• •Receipt maintained: Tesla wins on economics, but Waymo has viable alternative strategy

Strong critique of AI startups reporting annual recurring revenue in their first year by multiplying monthly revenue by 12. High churn rates and Y Combinator circular economy inflate numbers. True ARR should only count annual prepaid subscriptions.

• •Monthly subscription x12 ≠ ARR without proving 12+ month retention
• •Consumer AI companies have very high monthly churn rates that aren't disclosed
• •Y Combinator ecosystem creates circular economy - startups buying from each other
• •Proposed rule: First-year ARR should only count annual prepaid subscriptions
• •Need to report monthly churn rate alongside any ARR claims for transparency

Matty's prediction that software will shift from pre-coded pathways to AI-generated interfaces on-the-fly is proving correct. LLMs write code dynamically to create interfaces and business logic that weren't predefined, already happening in BuildPass and other tools.

• •Just-in-time software: AI writes code on-the-fly instead of pre-defining all pathways
• •Already being implemented in BuildPass and other production systems
• •Dynamic interfaces driven by AI rather than predefined business logic workflows
• •Parallel to construction: 'Should we just build this ourselves?' now applies to software
• •Teams can fix internal bottlenecks directly rather than buying point solutions

Scotty's passionate rant about software developers appropriating construction terminology. Engineers, architects, and especially 'builders' should be reserved for people with physical liability. Microsoft's Build conference calling coders 'the best builders in the world' alongside Elon Musk was the final straw.

• •Software roles have taken: engineers, architects, and now 'builders' from construction
• •Real builders face jail time if workers die on-site - different stakes than coding
• •Microsoft Build conference: Satya called software developers 'best builders' next to Elon (who builds rockets)
• •Suggested alternative: 'cursor jockeys' for developers using AI coding tools
• •Receipt: Leave construction terminology sacred - physical liability vs digital work