Summary of Anthropic’s Mythos is Back, OpenAI Releases GPT 5.6, Apple’s Price Increases

Overview of Big Technology Podcast — Anthropic’s Mythos is Back, OpenAI Releases GPT 5.6, Apple’s Price Increases

This Friday edition of Big Technology Podcast focused on three major tech-business stories: Anthropic’s restricted return of its “Mythos” model, OpenAI’s limited preview of GPT 5.6, and Apple’s sharp price hikes across Macs and iPads amid rising memory costs. The conversation centered on the bigger strategic question: whether frontier AI companies can sustain their business model if governments control access, customers move to cheaper models, and the market shifts toward model routing and open source alternatives. The episode closed with a tribute to tech writer and investor Om Malik, who passed away this week.

Anthropic’s Mythos Returns Under Government Oversight

What happened

• The U.S. government lifted its block on Anthropic’s powerful “Mythos” model, allowing it to be released to more than 100 U.S. institutions, including major companies and government agencies.
• The transcript suggests “Fable,” a consumer-facing version, may follow, though timing is unclear.
• The hosts framed this as a major sign that the government is now effectively part of the approval process for frontier AI releases.

Key discussion points

• The hosts questioned whether Anthropic and the government have actually created meaningful safeguards or just a more controlled release process.
• They noted that Anthropic itself appears to have helped push the idea of government review for frontier models.
• This raised broader concerns about:
  • government “picking winners and losers”
  • reduced access for smaller firms and the public
  • possible pressure on companies to move toward open source or non-U.S. alternatives

Main takeaway

The Mythos news is less about one model launch and more about a new regulatory precedent: frontier AI may increasingly be released only to a limited set of trusted users first.

OpenAI’s GPT 5.6: Stronger, Cheaper, but Not Yet Public

What OpenAI announced

• OpenAI previewed GPT 5.6 in a limited release.
• The model family includes:
  • Soul — flagship model
  • Terra — balanced model
  • Luna — fast, affordable model
• The hosts spent time mocking the naming, especially “Terra” and “Luna,” given their association with the failed Terra/Luna crypto collapse.

Capabilities and pricing

• GPT 5.6 reportedly performed similarly to Mythos on a cybersecurity benchmark.
• It was described as:
  • more token-efficient than Mythos
  • about half the cost
• Like Mythos, it is only available to a small set of trusted partners for now.

Strategic implications

• The hosts argued that restricted rollout changes the economics of frontier AI:
  • customers may delay adoption
  • companies may route work to cheaper models
  • open source may gain traction
• They also raised the possibility that controlled access may reduce distillation risk, since fewer people can try to copy or reverse-engineer the model.

Main takeaway

OpenAI’s release strategy shows that “best model wins” is no longer the whole story; access, pricing, and customer trust are becoming just as important.

The Frontier AI Business Model Is Under Pressure

The core debate

A major thread throughout the episode was whether frontier-model companies can sustain their growth story if:

• customers increasingly use cheaper models
• model routing becomes standard
• the best models are gated behind limited access
• open source models keep improving

Points raised

• The hosts said the early AI growth story relied heavily on customers using the most expensive frontier models.
• Now, many companies are shifting to:
  • cheaper “flash” models
  • routing different tasks to different models
  • open source for cost efficiency
• That creates a problem for companies whose valuation and IPO narrative depend on massive usage of premium models.

Competitive concern

• The discussion also touched on geopolitical risk:
  • if U.S. labs restrict access too much, China or open source ecosystems could gain ground
  • but if access is too open, safety and misuse concerns grow
• This tension was framed as one of the central unresolved issues in AI policy.

Main takeaway

The AI market may be moving from “one premium model for everything” to a more fragmented, cost-optimized ecosystem, which could weaken the frontier labs’ moat.

OpenAI’s IPO Timing and the Changing Market Narrative

IPO reporting

• The New York Times reported that OpenAI may delay its IPO until next year.
• The hosts suggested this may be a smart move because the market narrative has shifted.

Why the narrative changed

• A few months ago, frontier models were the dominant story and usage growth looked explosive.
• Now, executives and customers are talking more about:
  • model routing
  • cheaper models
  • efficiency gains
  • avoiding high AI bills
• The hosts argued that this weakens the “insane growth curve” story that an IPO would need.

Main takeaway

OpenAI may want to wait for a more mature product story and cleaner financials before going public.

AI Customers Are Cutting Bills and Auditing Overcharges

Reporting discussed

The episode covered reporting that AI customers are actively lowering their spend on Anthropic and OpenAI:

• some are switching to cheaper models from the same providers
• others are turning to third-party or open source options
• some are using model routing to reserve premium models for only the hardest tasks

Example

• A hospital software provider reportedly shifted to an OpenAI model that was 1/20th the cost of the premium alternative, generating major annual savings.

Billing issues

• Another report described AI billing inaccuracies:
  • an auditing startup reviewed $34 million in bills
  • it found about $1.7 million in mistaken overcharges
• The hosts noted that AI billing is often opaque, especially with timeouts and token consumption that customers may not fully understand.

Main takeaway

AI spend is becoming a real management issue, not just a line item, and customers are increasingly scrutinizing every token.

Apple’s Price Hikes: Memory Costs or Greed?

What Apple changed

• Apple raised prices on Macs and iPads by roughly 15% to 25%.
• Examples cited:
  • MacBook Air: up $200
  • MacBook Pro: up $300
  • iPad Air: up $150
  • iPad Pro: up $200

The debate

• Apple said soaring memory and storage costs forced the increase.
• The hosts questioned whether the hikes were justified, especially because:
  • Apple’s margins are already strong
  • the memory content in these devices is relatively small
  • the price increases are large relative to the actual component-cost change
• One host argued the move looked like greed, not just supply-and-demand.

Counterpoint

• The discussion also acknowledged that memory makers like Micron have endured years of low margins and industry consolidation.
• In that sense, Apple and hyperscalers may be paying the price after years of squeezing suppliers.

Main takeaway

Apple’s move is technically explainable by memory-market pressure, but the size of the hikes made the hosts view it as excessive and consumer-hostile.

Tribute to Om Malik

Closing remembrance

• The episode ended with a tribute to Om Malik, who died on June 24 at age 59.
• The hosts remembered him as:
  • a foundational tech blogger
  • generous, kind, and blunt in a memorable way
  • an influential voice in modern tech media
• They highlighted his warmth, humility, and impact on the industry.

Main takeaway

The show closed on a reflective note, honoring Om Malik’s role in shaping tech journalism and community.

Bottom Line

This episode argued that the AI industry is entering a more complicated phase:

• government-controlled access may become the norm for frontier models
• pricing pressure and model routing are changing customer behavior
• IPO stories may need rewriting
• and even big tech hardware companies like Apple are passing through input-cost shocks in ways that raise questions about fairness and strategy

The broader theme: the AI boom is still moving fast, but the easy assumptions behind it are starting to crack.