AI Built Better Quantum Circuits Than the Experts – What It Means | AI Explained Simply

An AI worked on a task for 53 hours with almost no human help – and found a blueprint for quantum computers that needs 17 times fewer computing steps than the best one published so far, with 12% more memory. I explain, without jargon, what Anthropic published in October 2026: what a quantum circuit is, why fewer computing steps decide when quantum computers become useful, how Claude worked with tools and an automatic checker, and what came out of it. Plus, honestly: the limits, the downside for today's encryption, a vision of where this could go, and what you can learn from it for your own work with AI. 0:00 Intro 0:35 What is a quantum circuit? 1:07 Why fewer steps matter 1:43 How Claude worked 2:42 Results – and their limits 3:30 Uses – and the encryption downside 4:06 Vision: where this could go 5:43 What you can learn 6:09 Sources Sources (accessed Oct 26, 2026): • Anthropic, "Designing more efficient quantum circuits with Claude", Oct 16, 2026: https://www.anthropic.com/research/de... • Gu, Kemeny, Yelin, Lukin: "Discovering state-of-the-art quantum circuits with large language models", arXiv 2610.22367, Oct 15, 2026 (preprint): https://arxiv.org/abs/2610.22367 • Bar chart figures: Table 1 of the preprint (metric: Toffoli gates × qubits) Disclosures: • These are designs and resource estimates. Nothing has run on real quantum hardware; the preprint has not been peer reviewed yet. • The 17× applies to elliptic-curve point addition (with 12% more qubits); for the error-correction decoder it is about 12×. • The search path, curve shape, circuit drawings and the FeMoco molecule are schematic illustrations; end points and figures come from the sources. • That the results could bring attacks on today's encryption closer is my own conclusion; Anthropic makes no statement on this. • The vision (AI designs, a real quantum computer runs it) is my own assessment, not a prediction. • OneLog is a product of my company. 🔔 If this video helps you: subscribe to the channel. 💬 What would you use a quantum computer for first? Tell me in the comments. More videos: • How AI learned language (7,000 romance novels):    • How AI Learned Language – From 7,000 Roman...   • Why AI is so convincingly wrong:    • Why AI Is So Convincingly Wrong – and How ...   • One prompt, three results (chat or agent?):    • One Prompt, Three Results: Why AI Builds a...   How your team uses AI safely and with understanding: Kompetenzanker, my learning platform for AI literacy. The AI Foundation has 6 chapters, is jargon-free, EU-hosted and free for teams. https://kompetenzanker.de/gratis?lang... Synthetic voice (ElevenLabs), generated from my own voice. Content by Marcus. No avatar. Animation: own production. #AI #QuantumComputing #AIExplained #Claude #AIinPractice