PLUS: AI reads a 2,000-year-old scroll and the AI cost-cutting era begins
Happy reading
OpenAI has just announced GPT-5.6, its most advanced model yet—but there's a catch. Instead of a standard public rollout, access is being tightly controlled through a government-vetted preview program. Washington is now acting as gatekeeper to the world's most powerful AI systems.
This marks a significant shift in how frontier models reach the market. What does it mean when federal oversight becomes the default for cutting-edge AI? And how will this two-tiered access system shape the future of innovation and competition?
In today's Next in AI:
OpenAI's GPT-5.6 launches under government control
AI successfully decodes 2,000-year-old Vesuvius scroll
The AI industry pivots to efficiency and cost-cutting
Anthropic's Mythos 5 joins the government-gated release model
OpenAI's Next Generation
Next in AI: OpenAI has unveiled its next-generation flagship model series, GPT-5.6, but is launching it with a limited preview for partners vetted by the US government.
Explained:
The new series includes three tiers: Sol (the flagship), Terra (a balanced model), and Luna (a fast, affordable option), all showing major improvements in coding, biology, and cybersecurity tasks.
In a significant policy shift, the initial release is a government-gated preview, with the US administration vetting the small group of trusted partners allowed early access.
OpenAI also detailed a new pricing structure, with Sol priced at $5 per 1M input tokens and $30 for output, making it their most powerful and premium model available to partners.
Why It Matters: This phased, government-vetted release establishes a new precedent for how frontier AI models are introduced, balancing rapid innovation with national security interests. For developers and businesses, this signals a future where access to the most capable AI may involve more oversight and a slower, more deliberate rollout process.
AI Reads Ancient History
Next in AI: In a historic breakthrough, researchers used advanced AI to digitally unroll and read a 2,000-year-old papyrus scroll that was carbonized during the eruption of Mount Vesuvius in 79 AD.
Explained:
The process, called virtual unwrapping, uses CT scans to create a 3D model of the charred scroll and AI algorithms to detect ancient ink hidden on the fragile layers.
The text reveals a previously unknown philosophical work that discusses Stoic concepts of impulse and practical wisdom, offering a new glimpse into ancient thought.
This achievement from the Vesuvius Challenge successfully deciphers an entire scroll that experts had previously deemed completely unreadable.
Why It Matters: This fusion of AI and archaeology provides a powerful new method for accessing lost knowledge from antiquity. The technique sets a precedent for deciphering other damaged historical artifacts previously thought to be lost forever.
AI's Cost-Cutting Era
Next in AI: The AI industry is entering a new phase of maturity as the "tokenmaxxing" era—chasing performance at any cost—gives way to a sharp focus on efficiency. Companies now demand a clear return on investment, forcing AI labs and users to get smarter about spending.
Explained:
Enterprise AI bills have ballooned, with companies like Uber reportedly blowing through their budget in just four months and prompting an industry-wide spending crackdown.
In response, leading labs are adapting, with OpenAI launching new spending controls that allow administrators to set usage limits and monitor costs across their organizations.
A new wave of competition is emerging from cheaper, open-weight models and intelligent model routing systems that automatically select the most cost-effective AI for a given task.
Why It Matters: This trend marks the AI industry’s transition from a high-growth, experimental phase to one focused on sustainable, real-world business value. For developers and businesses, this cost-conscious era will drive innovation in efficiency, expand model choices, and prioritize practical applications over raw power.
Washington's Hand on AI
Next in AI: The U.S. government is loosening its grip on Anthropic's powerful Mythos 5 AI, releasing it to a vetted list of over 100 U.S. organizations. This move signals a new era of direct federal oversight for the most capable AI models.
Explained:
The reversal comes just two weeks after the government ordered access to Mythos 5 and its sibling Fable 5 suspended over national security risks related to potential misuse.
Access is now limited to a list of trusted partners, including Fortune 500 companies and institutions that operate and defend critical U.S. infrastructure.
This decision mirrors a pattern of controlled releases, as OpenAI also limited its new GPT-5.6 model to a small group of government-approved partners on the same day.
Why It Matters: Washington is now acting as a gatekeeper for frontier AI, directly deciding which companies get a competitive edge with the most powerful tools. This new regulatory framework creates a two-tiered system, shaping the global AI race by controlling access to cutting-edge technology.
AI Pulse
OpenAI weighs delaying its IPO until 2027, citing market volatility as a potential obstacle to its goal of a $1 trillion valuation.
Chatbots show a significant left-leaning bias according to a new Washington Post study, with OpenAI’s models providing liberal-only arguments 80% of the time.
AI's influence on U.S. elections is growing, with a new project tracking hundreds of millions in political spending from the industry.