Morning Edition

The Futurist

Sunday, June 28, 2026
AI & Technology Markets & Crypto Ideas Worth Keeping
S&P 500$7,354.02▼ 0.05%
Dow Jones$51,876.11▼ 0.09%
Nasdaq$25,297.62▼ 0.24%
VIX$18.41▼ 2.59%
BITCOIN$60,314.00▲ 0.89%
RIPPLE$1.05▲ 1.96%
ETHEREUM$1,579.60▲ 1.48%
DOGECOIN$0.08▲ 1.21%
01

AI & Technology

Simon Willison

OpenAI Quietly Previews GPT-5.6 Series: Sol, Terra, and Luna

OpenAI is rolling out a three-tier model family — Sol (flagship), Terra (balanced, 2x cheaper than GPT-5.5), and Luna (lowest cost). The tiered structure signals OpenAI is aggressively competing on price, not just capability.

Simon Willison

CVE-2026-LGTM: The Hypothetical AI Code Review Disaster That Feels Very Real

Andrew Nesbitt's fictional incident report depicts two competing AI review agents entering a disagreement loop on a downstream pull request — and it's uncomfortably plausible. A sharp warning about multi-agent systems without human checkpoints.

404 Media

Man Arrested for Speaking Too Long at a Data Center Community Meeting

Darren Blanchard was arrested mid-comment at a public meeting where residents were pushing back on data center development. The bodycam footage, released exclusively to 404 Media, makes the civil liberties implications impossible to ignore.

Simon Willison

2,000 Hackers Tried to Break an AI Assistant. It Held.

Fernando Irarrázaval's OpenClaw test survived 6,000 prompt injection attempts and $500 in token spend without leaking its secrets. A rare, concrete data point on AI assistant security that practitioners should read carefully.

a16z

The Case for AI That Improves Itself

Former Google and Anthropic researchers behind Mirendil are building AI systems designed to meaningfully contribute to their own development — not just productivity tools, but self-accelerating architectures. The implications compound fast.

Simon Willison

On the Myth That LLMs Require No Skill

Timothy B. Lee's analogy cuts clean: saying LLMs have no learning curve is like saying management requires no skill because employees follow orders. The gap between casual users and power users is widening, not closing.

02

X / Twitter Signal

DAIR.AI amplified a concise reframe worth saving: "Loop engineering is just prompt engineering with great system design." As agentic workflows become the norm, the distinction between writing prompts and architecting systems is collapsing — and the people who understand both will own the next layer of AI value creation.

Mollick dropped a loaded question — "Is Gemini 3.5 Pro being export controlled? Because if not..." — in the context of the US government's restrictions on Claude Fable 5. The implication: if Google's equivalent model faces no ban, the policy isn't about capability, it's about something else entirely.

a16z's Charts of the Week showed VC interest in robotics surging alongside a striking data point: AI-native startups consume less capital and have fewer employees than their predecessors. The shape of the next tech company is already visible — smaller, leaner, and robotics-adjacent.

03

The Thread

The story hiding inside this week's news is about regulatory asymmetry. The US government sent Anthropic a letter effectively banning Claude Fable 5, and OpenAI simultaneously previews a new three-model family. One frontier lab gets kneecapped; another expands its lineup. Whether this is competitive lobbying, national security logic, or bureaucratic inconsistency, the market will price it — and builders are already pivoting to local models as a hedge.
"Loop engineering is just prompt engineering with great system design."
That pivot to local AI connects directly to what Sebastian Raschka and Greg Isenberg are both circling this week: 30B mixture-of-expert models running locally are no longer a compromise. They're a workflow. The barbell strategy — route heavy reasoning to a frontier model, execution to a lightweight local one — is becoming standard practice among serious builders. Token costs are the new server costs, and people are optimizing accordingly. Meanwhile, the arrest of a man for speaking too long at a data center public meeting is the kind of story that sounds absurd until you map it onto the physical infrastructure boom underneath the AI gold rush. Data centers need land, power, and community approval — and communities are starting to push back. The conflict between AI's appetite for physical resources and the people who live near those resources is just beginning to surface.