Morning Edition

The Futurist

Thursday, June 25, 2026
AI & Technology Markets & Crypto Ideas Worth Keeping
S&P 500$7,365.46▼ 1.44%
Dow Jones$51,666.84▼ 0.09%
Nasdaq$25,587.04▼ 2.21%
VIX$19.15▼ 1.69%
BITCOIN$62,617.00▲ 0.33%
RIPPLE$1.10▼ 0.49%
ETHEREUM$1,666.17▲ 0.95%
DOGECOIN$0.08▼ 0.49%
01

AI & Technology

A16Z

Mirendil Raises $200M Seed to Build Self-Improving AI

Former Google and Anthropic researchers just landed a $200M seed round — one of the largest ever — to build AI systems that contribute to their own development. The bet: self-accelerating AI is the next frontier, not just AI as a productivity tool.

Simon Willison

The LLM Portfolio Problem Is Getting Worse

Tom MacWright flagged a disturbing hiring pattern: job applicants linking to LLM-generated portfolios, which link to LLM-generated GitHub repos with LLM-generated commits. Hiring managers can no longer tell who they're actually talking to.

Simon Willison

Prompt Injection Reframed as Role Confusion

A new paper reframes prompt injection attacks as a fundamental role-confusion problem in LLM architecture — and published a readable blog version alongside the academic paper. The security implications for agentic AI are significant.

404 Media

Hackers Stole 45GB from Madison Square Garden

Attackers exfiltrated over 45GB of data from MSG, including talent and New York Knicks files. The breach anatomy is now coming into focus, pointing to a specific entry vector.

404 Media

The Cosmic Web Is Bigger Than Physics Predicted

Dark matter structures scaffolding the universe are far larger and more persistent than cosmological models assumed. The standard model of cosmic structure may need a serious revision.

Greg Isenberg

A 19-Year-Old Is Making $10K/Month With AI Mobile Apps

George Lampropoulos built WrestleAI to 100K+ downloads and nearly $200K revenue using a repeatable AI-app framework. The playbook is disturbingly simple: find a niche, build fast, distribute deliberately.

02

X / Twitter Signal

Karpathy called Anthropic's Slack integration a genuinely new paradigm — AI that's "inline" with all other human activity across an org, rather than a separate tool you context-switch to. When Karpathy calls something a paradigm shift, it's worth taking seriously.

Mollick flagged new research showing that many people who claim they never use AI are actually using it — just secretly. Shadow AI adoption is real, it's widespread, and it makes every survey about AI workplace usage functionally useless.

a16z announced leading Mirendil's $200M seed round with a pointed framing: "Frontier AI work has been locked inside a few big labs." This isn't just a funding announcement — it's a thesis that the self-improving AI race is about to open up beyond OpenAI and Google.

Brent crude officially erased all gains since the Iran War began in late February. That's a macro signal worth watching — energy markets are pricing out the geopolitical risk premium faster than anyone expected, which has ripple effects across inflation models and Fed positioning.

03

The Thread

The biggest story today isn't the MSG hack or the cosmic web — it's the $200M seed round for Mirendil and what it signals. When a16z writes one of the largest seed checks in venture history for a team building self-improving AI, the subtext is clear: the next phase of the race isn't about who has the best model today, it's about who builds the system that improves itself fastest tomorrow.
"Frontier AI work has been locked inside a few big labs. Mirendil is building a system that can meaningfully contribute to its own development."
That framing lands differently alongside Karpathy's observation about Anthropic's Slack integration. The two threads connect: AI moving inline with human workflows (Slack) and AI contributing to its own development (Mirendil) are both pointing at the same destination — systems where the boundary between AI-as-tool and AI-as-collaborator dissolves. The Mollick shadow-AI data makes it stranger still: people are already quietly crossing that line at scale, they just won't admit it. Meanwhile, Tom MacWright's LLM-portfolio problem is the unglamorous friction point nobody wants to solve. A hiring market flooded with AI-generated signals of competence — portfolios, commits, cover letters — isn't just annoying for recruiters. It's eroding the epistemic infrastructure that lets meritocracy function at all. The Mirendil round and the fake GitHub repo are the same story told from opposite ends.