Good morning. It's Wednesday, April 15, and we're covering Claude’s push into automated coding workflows, a growing AI divide between experts and the public, NVIDIA’s latest quantum leap, and more.
NVIDIA Open-Sources Models to Tackle Quantum Computing
NVIDIA claims its new Ising models can improve quantum error correction by up to 3x while speeding decoding 2.5x—addressing one of the field’s most stubborn technical hurdles. The open-source model family targets quantum processor calibration and error correction, positioning AI as a core control layer for hybrid quantum-classical systems. The models include a vision-language system for calibration and convolutional networks for decoding, with adoption already spanning institutions like Harvard, Fermilab, and IonQ, as the quantum market is projected to exceed $11 billion by 2030.
Amazon Targets Drug Discovery Bottlenecks
Amazon says its new AI tool can generate hundreds of thousands of drug candidates in weeks, compressing work that typically takes months. Amazon Web Services’ Bio Discovery platform lets scientists run complex drug-design workflows without coding, using foundation models and an AI agent to generate and evaluate molecules. In one collaboration with Memorial Sloan Kettering, the system produced nearly 300,000 antibody candidates and narrowed them to 100,000 for lab testing, with partners like Bayer and the Broad Institute already adopting the tool.
Anthropic Briefed Trump Administration on Mythos Model
Anthropic disclosed it briefed the Trump administration on a powerful AI model it refuses to release publicly due to cybersecurity risks. The company’s co-founder Jack Clark confirmed the outreach at the Semafor World Economy summit, despite an ongoing lawsuit against the Department of Defense over being labeled a supply-chain risk. Reports indicate officials have encouraged banks including JPMorgan, Goldman Sachs, and Citigroup to test Mythos, while Anthropic continues to limit access over concerns tied to misuse.
POWERED BY RECALL 2.0
Your Personal Context Is the Next AI Race
As AI levels the playing field, advantage shifts from model access to what those models are grounded in. Your knowledge.
Recall 2.0 is one place for your knowledge and AI. Save what matters, write what you think, curate an AI that knows what you know.
"Condense my research on LLMs, enrich it with new studies, find the exact point quantization was mentioned in my podcasts."
Talk to your knowledge, the internet, or both in the model you choose.
The future belongs to AI that works from what you know.
Try Recall 2.0 free today or use code FF25 for 25% off the uncapped plan.
REAL OR AI
Think You Can Still Spot AI? Answer ↓
PRODUCT RELEASE
Anthropic Introduces Claude Code “Routines” for Automated Workflows
Anthropic has launched “routines” for Claude Code, a feature that lets users automate development tasks using scheduled jobs, API calls, or GitHub events. Each routine bundles a prompt, repositories, and connectors, then runs autonomously in cloud-hosted sessions—even when a user’s machine is offline.
The system supports triggers like cron-style schedules, HTTP endpoints, and repository activity such as pull requests or pushes. The feature turns Claude into a persistent automation layer for engineering workflows, from code review to incident response. The capability is currently in research preview, with evolving limits and API behavior. → Read the full article here.
REPORT
Stanford Report Finds Widening Gap Between AI Experts and Public
A new Stanford University AI report released April 13, 2026, finds a growing divide between how experts and the public view artificial intelligence, particularly in the U.S. While a majority of AI experts expect positive long-term impacts across areas like healthcare, jobs, and the economy, public sentiment is far more skeptical.
Only 10% of Americans say they are more excited than concerned about AI, with many worried about job loss, rising energy costs, and broader economic disruption. The report also highlights declining optimism among younger users, even as AI adoption remains high. → Read the full article here.
CHIPS
NVIDIA Says AI Turns Months-Long Chip Design Work Into Overnight Task
NVIDIA says it is using artificial intelligence to dramatically accelerate chip design, cutting a task that once took eight engineers 10 months down to a single overnight run on one GPU. The company’s chief scientist, William Dally, described how AI systems—particularly reinforcement learning models like NB-Cell—now handle labor-intensive steps such as standard cell library design.
NVIDIA is also deploying internal large language models trained on proprietary GPU data to assist engineers and boost productivity. Despite these gains, the company says fully autonomous chip design remains far off, with human oversight still required for verification and system-level decisions. → Read the full article here.
ROBOTICS
Google Unveils Gemini Robotics-ER 1.6 To Boost Real-World Robot Reasoning
Google has introduced Gemini Robotics-ER 1.6, an upgraded AI model designed to improve how robots understand and act in physical environments. The system enhances “embodied reasoning,” combining spatial awareness, multi-view perception, and task planning to enable more autonomous robotic behavior. New capabilities include instrument reading—allowing robots to interpret gauges and sensors—and improved success detection to determine when tasks are complete. → Read the full article here.
NEWS
What Else is Happening
Texas man charged after Altman attack: Daniel Moreno-Gama allegedly threw a Molotov cocktail at Sam Altman’s home, citing AI extinction fears; prosecutors filed attempted murder and federal explosives charges.
OpenAI acquires finance startup Hiro: OpenAI bought AI financial planning startup Hiro, founded in 2024 by Ethan Bloch, likely as a ~10-person acqui-hire; the service shuts down April 20 with data deleted May 13.
AI designs ultra-strong rustproof steel: Researchers used machine learning to create a 3D-printable alloy reaching 1,713 MPa strength and 15% ductility, outperforming stainless steel and improving manufacturing efficiency.
Google Chrome adds reusable AI Skills: Google launched “Skills” in Chrome, letting users save and reuse Gemini prompts across sites; rolling out April 14, 2026 to desktop users with English (US) settings.
US Treasury probes Anthropic’s Mythos AI (Paywall): Treasury officials, led by CIO Sam Corcos, seek access to Anthropic’s restricted model to identify security flaws as agencies prepare for risks from advanced AI systems.
REAL OR AI
The Image Is AI. Tune in next week for another Real or AI test.
