Radar.

Most ideas start with a convenience argument. It's easier. It's faster. People are already doing X, we just make it simpler. That framing gets you through a lot of early conversations, but it rarely survives contact with a skeptical investor or a CFO asking why they'd change a process that isn't broken.

Working with the founders of Heard.reviews, a voice-first verified review capture platform, the first thing we had to do was throw the convenience argument away.

The real insight wasn't that leaving a review by voice is easier than typing one (though it is). It was that the UK's Digital Markets, Competition and Consumers Act 2024 had quietly made fake reviews illegal, with CMA enforcement beginning in earnest from July 2025 and penalties running up to 10% of global turnover. Overnight, "we help you get more reviews" became "we help you stay on the right side of the law."

That shift changes everything. It changes who you talk to first (compliance leads, not marketing managers). It changes how you price (value-based, not cost-comparative). And it changes how defensible the business is, because regulatory forcing functions don't go away.

The other thing worth noting: voice as a verification layer, not just an interface. Text is trivially easy to fake at scale. Voice, especially with fingerprinting capability, raises the bar meaningfully. In a market where an estimated 30% of online reviews are considered fake or inauthentic, that's not a feature. It's the point.

The lesson I keep coming back to: the strongest positioning usually comes from asking not what does this make easier, but what problem does this make unavoidable to solve. Regulation, when it arrives, tends to answer that question very cleanly.

Heard.reviews is at pre-seed stage and raising now.

The security equilibrium we've relied on for 20 years is ending.

Anthropic published something significant yesterday. Their new model, Claude Mythos Preview, can autonomously find and exploit zero-day vulnerabilities — in every major operating system, every major browser — without human intervention after a single prompt.

This isn't a gradual improvement. A month ago, their previous model turned known vulnerabilities into working exploits twice out of several hundred attempts. Mythos Preview did it 181 times.

What makes this different from the usual AI capability announcement is what Anthropic did with it. They're not releasing the model publicly. Instead they've launched Project Glasswing — a coordinated effort to use it defensively first, hardening critical software before models with similar capabilities become broadly available. From others, if not from them.

That last part is the point. Anthropic can choose to move carefully. The next lab might not.

The paper is honest about the transition ahead. For two decades, the security industry has operated in a rough equilibrium — new attacks, but of broadly the same shape. What's coming breaks that. The work that previously required expert professionals taking days or weeks can now be done autonomously, overnight, at the cost of a few hundred dollars of API calls.

The question isn't whether this changes things. It's whether defenders will be ready before attackers are.

Take-Two just fired its AI team.
Everyone's asking whether this means big games companies are going cold on AI. Wrong question.

Take-Two didn't fail at AI. They failed at owning AI. The team was largely inherited from the Zynga acquisition, built for mobile and social games, not for the handcrafted open worlds Rockstar obsesses over. Central tooling teams at holding companies rarely land. Studios protect their pipelines. The mismatch was structural from day one.

Meanwhile, Zelnick told investors the company has "hundreds of pilots and implementations" running. The capability isn't going awaythe internal build is.

This is actually the more interesting trend: large publishers quietly concluding that AI is infrastructure, not a headcount story. You use it through vendors, embed it in workflows, and don't build a team to wave the flag.

The flag-wavers just got laid off.

AI brings a whole world of possibilities for creative content. But it begs the question: when does it cross the line, and what can we do to protect IP rights holders?

300 hours in before the relaunch. Bethesda's bet on a full rework rather than incremental patches is unusual — and the kind of long-term product thinking most studios can't afford.