issue #005 · June 12, 2026
Bezos's $12B physical-AGI bet, Apache Burr, and Claude Fable 5's fine print
85% of enterprises want agentic AI in 3 years. 76% admit they can't ship it.
| AI Week Radar |
#5 Fri, Jun 12 |
Prometheus raised twelve billion dollars before shipping a product, while MIT's latest survey suggests three-quarters of enterprises can't operationalize the agents they're already buying. The gap between capital and capability has rarely looked wider.
— Jarek
Featured
MEGARAISE
Jeff Bezos's Prometheus raises $12B to build an 'artificial general engineer' — a $12B round at a $41B valuation, pre-product, aimed at autonomous systems for engineering and drug design. The pitch is embodied AI for industrial automation; the skepticism writes itself when the cap table arrives before the demo. — Marina Temkin
Articles & Videos
DEEP DIVE
Rethinking organizational design in the age of agentic AI — "85% of organizations say they want to be agentic within the next three years, 76% say their current operations and infrastructure can't support that change." The bottleneck isn't model quality — it's workflows, ownership, and the unglamorous middleware nobody wants to refactor. — MIT Technology Review Insights
CAPITAL FLOWS
SpaceX prices shares at $135 in the largest IPO ever — not an AI story on its face, but the largest IPO on record reshapes where capital pools next — and Starlink is increasingly part of the training/inference logistics conversation. Worth a glance if you track compute and connectivity infrastructure. — Tim Fernholz
Releases
POLICY
Anthropic Claude Fable 5 — Mythos but Safe, with Controversial Terms — a new Mythos-class model lands alongside updated usage terms that some builders are calling restrictive. Watch whether enterprise deployments hit friction, and whether competitors quietly position around the carve-outs. — latent.space
Tools & Papers
Apache Burr: build reliable AI agents and applications — an open-source framework for stateful agents with persistence, execution tracking, and debugging baked in. Competes directly with LangGraph and Temporal — worth a serious look if your multi-step agent is currently held together by Redis and prayer. — anhldbk
Releases
Apple will let you build workflows using AI in its new Shortcuts app — natural-language workflow generation arrives in Shortcuts. Friction goes down for casual automation; the real test is whether Apple's on-device models handle multi-step chains as well as Zapier or Raycast already do. — Sarah Perez
Articles & Videos
Rehumanizing global health care with agentic AI — the pitch: agents for triage, scheduling, and documentation to claw back clinician hours. The reality: EHR integrations and regulatory approval remain the unsexy gating factors that determine whether any of this ships. — MIT Technology Review Insights
How to refactor code with Claude Code — a walkthrough of using Claude Code on existing codebases. Light on benchmarks against manual review, but a useful template if you're still figuring out the prompting cadence for non-trivial refactors. — Eivind Kjosbakken
Last but not least
Persona Atlas: mapping how famous minds think — a Hugging Face hackathon project that tries to model the reasoning patterns of notable figures via LLMs and retrieval. Mostly prompt engineering and RAG dressed up — but a fun reference if you're shipping persona-based agents. — huggingface.co
See you next week — try to ship something between the funding rounds.
Read on web · Manage preferences · Unsubscribe in one click
© AI Week Radar · 8cells Jarosław Kaczmarski · ul. Ketlinga 33a, 96-313 Jaktorów, Poland