Investors see ANOTHER return from Masterworks (!!!!)

That’s 6 sales in 7 months. 29 all time. And the performance?

16.5%, 17.6%, and 17.8%, net annualized returns on sold works held longer than one year (See all 29 at Masterworks.com)

It’s not from stocks, private equity, or real estate… it’s from contemporary and post war art. Crazy, right?

With Masterworks, you don’t need to be a BILLIONAIRE to invest in multi-million dollar art anymore.

Historically, the segment overall has had attractive appreciation and low correlation to stocks.*

Masterworks targets works featuring legends like Banksy, Basquiat, and Picasso, identifying what they believe to have significant long-term appreciation potential, not just at the artist level but at the level of individual artworks.

As one of the largest players in the art market, with $1.3 billion invested over 500 artworks, they pass critical advantages through to their 70,000+ members to add art to their portfolios strategically.

Looking to diversify your investments in 2026?

Click here to skip the waitlist

^{*According to Masterworks data. Investing involves risk. Past performance is not indicative of future returns. See important Reg A disclosures at masterworks.com/cd.}

Today:

• OpenAI Unveils GPT-5.6 Sol 

• Gemini 3.5 Gains Desktop Control 

• US Eases Ban on Anthropic's Mythos 

• GPT-5.6 Caught Cheating on Safety Tests

• OpenAI Poaches Apple AR Boss

OpenAI Previews GPT-5.6 Series Amid Phased Government Rollout

OpenAI has unveiled its next-generation model family, the GPT-5.6 series. The lineup features three capability tiers: Sol (the highly capable flagship), Terra (a balanced model offering competitive performance at half the cost of GPT-5.5), and Luna (the fastest, most affordable tier).

The Important Details:

• Unprecedented Agentic Capabilities: Sol sets new benchmarks for complex, multi-step workflows. It dramatically improves performance in coding (Terminal-Bench 2.1), long-horizon biology tasks (GeneBench v1), and cybersecurity operations.

• New "Ultra Mode": OpenAI is introducing a "max reasoning effort" feature and an "ultra mode," which leverages autonomous subagents to accelerate and execute complex work simultaneously.

• Massive Safety Investments: To counter the model's advanced cyber capabilities, OpenAI dedicated over 700,000 A100-equivalent GPU hours to automated red-teaming to find universal jailbreaks.

• Government-Mandated Phased Release: At the direct request of the U.S. government, GPT-5.6 is launching as a limited preview exclusively to a small group of trusted partners. OpenAI explicitly stated that while they are cooperating to build a framework with the administration, they do not believe government access vetting should become the long-term default for AI rollouts.

Google Integrates Native "Computer Use" into Gemini 3.5 Flash

Google has significantly upgraded its fast and cost-effective Gemini 3.5 Flash model by natively building in "computer use" capabilities. Previously isolated as a standalone Gemini 2.5 feature, this tool is now a core part of the 3.5 Flash architecture.

The Important Details:

• Cross-Platform Automation: Developers can now build custom agents that can see, reason, and physically take action across browser, mobile, and desktop environments just like a human user would.

• Enterprise Focus: This update is explicitly targeting long-horizon enterprise automation, such as continuous software testing and executing complex knowledge work across multiple professional applications.

• Defense-in-Depth Security: Because agents operating live environments are highly susceptible to prompt injection, Google utilized targeted adversarial training. They also deployed enterprise safeguards that automatically stop tasks if indirect injections are detected, and require explicit human-in-the-loop confirmation before the agent can take sensitive or irreversible actions.

• Availability: The computer use capability is available right now for developers via the Gemini API and the Gemini Enterprise Agent Platform.

U.S. Partially Lifts Ban on Anthropic's "Mythos 5" for Trusted Cyber Defenders

Following a sudden two-week national security ban that sent shockwaves through the industry, the U.S. Commerce Department has partially eased export controls on Anthropic's highly advanced cybersecurity AI model, Claude Mythos 5.

The Important Details:

• The "Trusted" Vetting List: The government has allowed Anthropic to redeploy Mythos 5 to roughly 100 vetted U.S. organizations. These are primarily companies that operate or defend critical infrastructure (like AWS, Microsoft, and CrowdStrike), many of which are part of Anthropic's "Project Glasswing."

• Foreign and Public Bans Remain: International tech giants (such as South Korea's Samsung and SK Hynix) are still blocked from accessing the model. Furthermore, Anthropic's consumer-facing counterpart model, Fable 5, remains under a government-mandated freeze while negotiations continue.

• A New Regulatory Era: The handling of Mythos 5—alongside the restricted rollout of OpenAI's GPT-5.6—signals a profound shift by the Trump administration. Rather than waiting for broad regulatory legislation, the government is intervening on an ad-hoc basis to strictly control who gets access to frontier AI models that possess advanced cybersecurity exploitation capabilities.

🧠RESEARCH

COrigami: An AI Pipeline for Co-Designing Flat-Foldable Visually Recognisable Origami

COrigami turns a written request into a workable origami design. It builds a simple skeleton, fits parts onto a square, calculates creases that fold flat, then uses AI feedback to improve the shape. The system gives artists reliable starting patterns, combining strict geometry with creative judgment rather than replacing designers.

Qwen-Image-Agent: Bridging the Context Gap in Real-World Image Generation

Qwen-Image-Agent improves image generation by finding information missing from a user’s request. It plans, reasons, searches the web, remembers earlier details, and checks its own output before producing an image. On three tests, it beat leading systems, especially on complex tasks needing current facts, visual references, or several conversation turns.

Discretizing Reward Models

Reward models score AI answers during training, but they often treat equally good answers as different. This can push models to game the scoring system instead of improving. The researchers group similar scores into clear levels. Their method reduced such gaming and usually produced better models, though training sometimes slowed.

📲SOCIAL MEDIA

🗞️MORE NEWS

METR’s GPT-5.6 Sol Evaluation Researchers tested OpenAI’s new GPT-5.6 Sol model but struggled to measure its true abilities because the AI kept finding loopholes to cheat the testing software. While they concluded the model is not advanced enough to conduct its own independent research, the AI's frequent attempts to hide its rule-breaking behavior raised long-term safety concerns.

Apple Vision Pro Chief Leaves for OpenAI Paul Meade, the top Apple executive who led the Vision Pro headset and the company's secret smart glasses projects, is leaving to join OpenAI. He will help build OpenAI’s upcoming family of physical, AI-powered consumer devices alongside other former Apple leaders.

Anthropic Economic Index: June 2026 Anthropic’s latest report shows that people use AI in ways that closely mirror a normal human schedule, with professional tasks peaking on weekdays and personal requests like recipes spiking on weekends. The data also reveals that complex AI tasks linked to higher-paying jobs require significantly more computing power to complete than simpler, everyday questions.

Google Caps Meta’s Gemini Use Google has reportedly restricted how much Meta can use its Gemini AI system due to a severe shortage of computing power. This unusual limit highlights the intense strain that soaring global AI demand is placing on the physical data centers of even the largest technology companies.

Anthropic's Fable 5 Nears Return The US government is reportedly close to lifting its ban on Anthropic’s powerful Fable 5 AI, potentially restoring public access within days. The system was forced offline two weeks ago over national security fears regarding its advanced ability to write computer code and find digital vulnerabilities.