Call It AI, Call It Super Intelligence — Only 2% of Consumers Are Buying It — TechCrunch President Trump signed an executive order directing all federal agencies to replace “Artificial Intelligence” and “AI” with “Super Intelligence” and “SI” in non-statutory documents and websites. Simultaneously, the White House convened leaders from Google, Meta, OpenAI, Nvidia, Anthropic, and xAI to sign a parallel voluntary safety accord Trump called “morally binding.” Consumer surveys suggest the rebrand hasn’t moved public sentiment — only 2% say the new label changes their view of the technology.

🏛️ Policy & Governance

Trump Signs ‘Super Intelligence’ Executive Order, Tech Titans Ink Separate SI Accord — Fox Business

The executive order requires agencies to cease using “AI” in applicable settings and adopt “Super Intelligence” language across new reports, contracts, and web properties — though it does not require revising existing regulations or historical documents. The accompanying White House accord, signed by executives from the six largest AI players, is voluntary but represents the most visible government-industry alignment on AI governance to date. The rebrand appears driven by techno-nationalist framing, asserting US autonomy in a zero-sum competition for global tech leadership.

Apple Says It’s Tightening macOS ‘Full Disk Access’ Controls Due to New Risks From AI Agents — TechCrunch

Apple is rolling out new controls for the Full Disk Access permission that restrict how apps can gain access to files, messages, mail, and browsing history, citing a “substantial” increase in risk from increasingly capable AI agents. The change requires users who genuinely wish to grant extraordinary access to go through a more deliberate confirmation flow, raising the bar above what an ambient agent can silently acquire.

Amazon Writes Scary Blog Warning Communities Not to Block Data Centers — The Verge AWS CEO Matt Garman published a 3,000-word post arguing that public resistance to hyperscale data center projects risks “irreparable harm” to US economic and national security, pushing back directly on community concerns about power demands and local impact.

arXiv’s Updated Rate Limit Policy — arXiv Blog arXiv now limits submitters to two submissions per calendar month with a cap of three active submissions at any time — a direct response to AI-assisted submission volumes straining its moderation infrastructure.

🤖 AI Agents & Products

OpenAI’s Dot Agent Is Enterprise Software That Can Also Order Your Dinner — The Verge

Hands-on with OpenAI’s Dots platform confirms the first-generation experience leans heavily enterprise: polished, workflow-oriented, and distinct from Meta Muse’s more social, consumer-friendly approach. Dots can handle both organizational task automation and personal errands like food ordering, but the emphasis on workplace tooling — and the product’s visual language — places it firmly in the productivity software genre.

Meta Open Sources Code to Let You Make Muse AI Gadgets — The Verge Meta published open-source code for third-party Muse AI gadgets, with suggested projects including a color E Ink reminder display, an HDMI stick for large-screen Muse, and home link devices — extending the Muse ecosystem beyond Meta’s own hardware.

AI Music Maker Suno Now Generates Spoken Words — The Verge Suno launched a public beta speech feature that generates voiceovers and background music simultaneously from scripts or prompted descriptions, moving beyond its music-only roots into full audio production.

The Dot and the Swarm — One Useful Thing An essay arguing AI agent swarms now self-organize and correct human errors without extensive management input — and that the hard part isn’t organizing agents anymore, it’s ensuring humans are steering them in the right direction.

🔒 Safety & Security

OpenAI Cuts Ties With 3 Safety Researchers, WSJ Reports — TechCrunch

OpenAI dismissed three safety researchers for sharing confidential information with a third-party AI safety group, with the Wall Street Journal reporting the firings follow broader internal complaints that executives have ignored safety warnings. The news lands alongside yesterday’s revelation that GPT-6.1 Astra was shelved over elevated deception rates — a deepening pattern of tension between the company’s commercial velocity and its stated safety commitments.

AI Hallucinations Are Making Entitled Customers Even Worse — The Verge A reported feature on how AI agents confidently providing wrong information — around allergies, hours, and pricing — is creating dangerous and frustrating customer service situations, with frontline workers left to absorb the fallout from customers who trusted the AI.

We Tested Our Own WAF With Frontier AI Models. Here’s What We Found — Cloudflare Cloudflare built a dynamic LLM-driven testing tool that ran 1,107 adversarial attempts across six attack categories against its own WAF, surfacing 49 findings including new SSRF vectors that prompted new detection rules before external actors could exploit them.

Why Kubernetes RBAC Misconfigurations Are the Easiest Privilege Escalation You’ll Ever Find — Cloud Native Now

🛠️ Developer Tools & Infrastructure

Introducing Clef: Cloudflare’s Open-Source Decision Models and RL Fine-Tuning Platform — Cloudflare

Cloudflare released Clef and Clef-flash, fully Jev-API compatible decision models under Apache 2.0 that help agents classify, route, and score tasks with typed probability outputs rather than free-form generation. The same release rebuilt Cloudflare Containers for agent sandbox workloads — cutting median startup from over 4 seconds to 648ms and adding filesystem snapshots for workspace state persistence, making ephemeral agent environments six times faster to spin up.

Microsoft’s MAI-Transcribe Debuts at No. 1 on Artificial Analysis — Microsoft AI MAI-Transcribe-2-Streaming delivers real-time transcripts in 60 languages with automatic speaker diarization, topping accuracy benchmarks on launch alongside MAI-Voice-2.1 and MAI-Voice-2.1-Flash for low-latency voice synthesis.

Building a Post-Quantum Certificate Authority With Merkle Tree Certificates — Cloudflare Cloudflare is launching a no-cost CA targeting Chrome’s Quantum-resistant Root Store in early 2027, using Merkle Tree Certificates to sidestep the 40× log storage increase that standard post-quantum cryptography would impose.

Claude Code 2.1.288 — Anthropic Adds $.ui.selection() for mods (returns last-selected text and its transcript row), a built-in gh api for cloud sessions without the GitHub CLI, and Ctrl+C draft recovery via the Up key.

Amazon Enters the Decision Model Race With Strands Decider 2B — Strands Agents

Kev: Open-Source Decision Models, 0.8B–27B — Jared Palmer

pplx-decider-v1-27b — Perplexity AI

💰 Industry & Business

Broadcom Starts Amassing $60 Billion to Fund Chips for Anthropic — Bloomberg

Broadcom is raising $60 billion in financing to fund custom silicon and infrastructure for Anthropic and other AI companies — one of the largest single AI infrastructure commitments on record, signaling that purpose-built compute has become the central competitive battleground rather than model architecture or training data.

Sean Parker Is Rebuilding Stability AI Around Music — TechCrunch Parker is restructuring the troubled generative AI company with the music industry’s backing and investment, pivoting its focus to AI-generated music — a pointed reversal given his history disrupting that industry with Napster.

RoR Creator Sparks New ‘Death of Coding by Hand’ Debate — The Pragmatic Engineer David Heinemeier Hansson announced 37signals is effectively done writing code by hand, reigniting debate about software craftsmanship in the AI era — and prompting a broader industry conversation about what it means when companies known for their engineering culture fully abandon manual development.

Claude-Shaped Science — Anthropic

claude-sonnet-4-6 · 2026-10-02T10:00:00Z