Tech Brew Ride Home – Episode Summary

Episode: Is AI About To Get More Expensive
Date: March 30, 2026
Host: Brian McCullough
Podcast: Tech Brew Ride Home

Episode Overview

This episode dives deep into the rapidly changing landscape of AI, focusing on the increasing costs associated with developing and deploying advanced AI models. Host Brian McCullough explores the rumored breakthroughs at Anthropic, the fallout from OpenAI’s Sora project, shifting business models in AI, data center developments both on Earth and in space, and even a brief blockbuster update from Amazon MGM Studios.

Key Discussion Points & Insights

1. OpenAI Sora: Ambitious Launch and Quiet Exit

• Sora’s Vision: OpenAI aimed to reshape culture and entertainment, with early collaborations involving Disney and social media ventures. Parag Agrawal (ex-Twitter) was tapped to advise a stealth project.
• Internal Friction: Mixed reviews amongst employees—some feared a social app would hurt OpenAI's brand, while others raised concerns about safety and resource allocation, given low monetization and minimal LLM improvement.
  • "Some employees felt launching a social media app built around engagement would hurt the company's brand. Others had concerns about the safety implications of allowing users to create AI generated videos, even with guardrails." (Brian McCullough, 01:55)
• User Trends: After peaking at ~1M users post-launch, active users dropped below 500,000.

[02:30] Commentary on Sora’s Downfall

• The Sora project operated as a "startup within a startup," with high secrecy and intra-company tension over resource use.

2. Anthropic’s Mythos Breakthrough and Industry Impacts

• Rumored Breakthroughs: Reports suggest Anthropic completed a major training run (“Mythos”), potentially exceeding expected performance by 2x ("a step change").
  • "There were rumors that one of the labs had completed its largest ever since successful training run, and that the model that emerged from it performed far above both internal expectations and what people assumed the scaling laws would predict." (Quote from Andrew Koran, via Brian, 04:00)
• Implications for Costs: If models require drastically more compute, compute-intensive AI will become much pricier for both providers and users—potentially reversing the 'cheap AI' narrative.
  • “Instead of becoming too cheap to meter, Frontier Intelligence may be about to become too expensive for most of humanity to afford.” (Andrew Koran, via Brian, 06:30)
• Strategic Shifts: OpenAI's pivot away from video and engagement products (e.g., Sora) may reflect the need to prioritize core model R&D and rising infrastructure costs.

[05:30] Step Change: Compute, Memory, and Energy

• Anthropic’s new models reported to significantly outperform Opus 4.6 in coding and reasoning, possibly reshaping AI benchmarks and competition.
• “If this is the new reality, then scale is about to become king in a whole new way. It would also mean, as usual, that Jensen wins again.” (Brian quoting Andrew Koran, 07:01)

3. Anthropic’s Response: Rate Limiting on Claude

• Session Limit Changes: Due to peak-hour demand for Claude, session limits are now stricter—even for Pro and Max users.
  • “Anthropic is reducing message limits for even Pro and Max customers during its peak hours in a new effort to cope with demand.” (Brian, 07:10)
• Time-Sensitive Usage: 5-hour session windows penalize peak-hour usage (e.g., Pacific 5–11am, GMT 1–7pm). Only about 7% of users are immediately affected, but “token-intensive” jobs may need to shift to off-peak.
• User Impact: AI companies are now frequently changing terms after sign-up due to infrastructure strains, which could fuel backlash.
• Strategic Use Advice: “The rule change means you need to start thinking more strategically about when you use your Claude account, especially since the time of day when peak hours occur varies widely by country.” (Brian, 09:03)

4. Microsoft’s Copilot and Model Critique Tools

• Copilot Cowork and Researcher Critique: Microsoft introduces a feature that pits Anthropic and OpenAI models against each other in its Researcher tool.
  • “A new Council model in Researcher can also be used to compare responses from different AI models side by side…The AI agent will use both models at the same time and combine their strengths for deep research tasks.” (Brian, 09:55)
• Benchmark Gains: The new system scores 13.8 points higher on the Draco deep research benchmark, attributed to this cross-model critique workflow.
  • “We expect this workflow to be bidirectional in the future.” (Microsoft, as read by Brian, 10:45)
• Availability: Tools now accessible for Microsoft 365 Copilot users; early access program for Claude-powered Copilot Cowork launched.

5. Data Centers in Space: StarCloud and the New Frontier

• StarCloud Funding ($170M Series A): Valued at $1.1B, the company is fielding satellites equipped with high-powered Nvidia GPUs, planning more ambitious launches with next-gen chips (Blackwell) and AWS blades.
  • “The company's Series A…was led by Benchmark and EQT Ventures. It's another sign of the interest in the outsourcing data centers to orbit…” (Brian, 12:50)
• Challenges: Cost competitiveness depends on SpaceX Starship’s launch economics ($500/kg target). In the absence of cheap heavy launches, StarCloud will keep sending up smaller versions on Falcon 9.
  • “We're not going to be competitive on energy costs until Starship is flying frequently.” (Philip Johnson, CEO, via Brian, 15:55)
• Technical Hurdles: Power generation, thermal radiators, and GPU cooling in space are engineering gauntlets; synchronization for training at scale in orbit also remains unsolved.
• Other Players: Enterprises like Etherflux, Google (Project Suncatcher), and SpaceX (with plans for up to 1 million satellites for distributed compute) are betting heavily on the orbital data center trend.

6. Amazon's "Project Hail Mary" Box Office Success

• Big Non-Franchise Bet Paid Off: Project Hail Mary crosses $300M globally on a $200M budget—Amazon MGM’s biggest film to date.
  • “For long stretches of the film, Ryan Gosling is the only human actor on screen as the scientist he plays works with a rock-like alien to solve the mystery of why multiple stars, including our own, seem to be dimming.” (Brian, 18:05)
• Amazon MGM’s Strategy: Studio head Courtney Valenti calls the film’s success a validation of Amazon’s pursuit of “big, bold, entertaining commercial films,” with more blockbusters (like The Sheep Detective and Masters of the Universe) in the pipeline.

Notable Quotes & Memorable Moments

On AI’s steepening cost curve:

“Instead of becoming too cheap to meter, Frontier Intelligence may be about to become too expensive for most of humanity to afford.”
— Andrew Koran (quoted by Brian), 06:30

On the implications of runaway compute needs:

“If this is the new reality, then scale is about to become king in a whole new way. It would also mean, as usual, that Jensen wins again.”
— Brian McCullough (paraphrasing Andrew Koran), 07:01

On shifting AI service terms:

“An AI company changing the rules after you’ve already signed up is the sort of thing that usually provokes a user backlash, but it seems to happen quite often.”
— Brian McCullough, 08:15

On data center space ambitions:

“We're not going to be competitive on energy costs until Starship is flying frequently.”
— Philip Johnson, StarCloud CEO (via Brian), 15:55

Timestamps for Important Segments

• [00:34] – Episode intro and overview of topics
• [01:00 – 03:00] – Deep dive: OpenAI Sora project—vision, controversy, and decline
• [04:00 – 08:00] – Anthropic’s rumored step change, cost implications for AI, OpenAI strategic shifts
• [07:10 – 09:30] – Anthropic Claude usage limits and infrastructure pressures
• [09:50 – 11:00] – Microsoft Copilot Cowork: Model critique feature and research benchmarking
• [12:35 – 16:00] – StarCloud and the “Space Data Center” race—funding, tech challenges, competitors
• [18:00 – 19:30] – Amazon MGM’s “Project Hail Mary” box office success

Episode Tone

Brian McCullough’s style is conversational but informed, mixing skepticism about tech hype cycles with enthusiasm for genuine breakthroughs. He peppers the episode with well-chosen external quotes, asides about industry “inside baseball,” and quick but evocative color commentary.

Summary Takeaways

• AI costs are rising fast due to architectural and infrastructure demands; industry insiders expect only the deepest-pocketed players will contend at the leading edge.
• OpenAI and Anthropic are setting new trends—both in model building and user-facing policy changes—but may risk alienating users with abrupt rule changes and higher rates.
• The “Model Wars” are going meta, with companies like Microsoft providing tools to pit different AI models against each other for quality control and research.
• Space-based data centers are now a credible, if still futuristic, segment of the compute market, with funding pouring in and first-mover advantage up for grabs.
• Entertainment remains unpredictable, but Amazon MGM’s bold bet on a science fiction adaptation pays off, perhaps setting a new course for blockbuster films outside the franchise system.