1. Neuro-Symbolic AI Slashes Energy Use by 100x While Boosting Accuracy
Researchers at Tufts University have unveiled a neuro-symbolic approach that combines neural networks with human-like symbolic reasoning, cutting AI energy consumption by up to 100x. Tested on robotic visual-language-action models, the system achieved a 95% success rate on the Tower of Hanoi puzzle versus just 34% for standard approaches, using only 1% of the training energy and 5% during operation. The work will be presented at the International Conference of Robotics and Automation in Vienna this May.
2. Utah Expands AI Prescription Pilot to Psychiatric Medications
Utah has approved a second AI prescribing pilot, allowing Legion's AI system to autonomously renew 15 previously prescribed lower-risk psychiatric medications. The system cannot write new prescriptions, change doses, or handle controlled substances, antipsychotics, or lithium. Utah requires human escalation for safety flags and physician review of the first 1,250 requests before wider expansion. This follows the state's January launch of the first-ever AI prescription renewal program for chronic conditions.
3. Stripe's "Minions" Ship 1,300 Pull Requests Per Week With Zero Human Code
Stripe engineers have deployed autonomous coding agents called "Minions" that generate over 1,300 merged pull requests per week containing zero human-written code. Tasks originate from Slack messages, bug reports, or feature requests and flow through "blueprints" — orchestration flows that alternate between deterministic code nodes and open-ended agent loops. Each minion gets at most two CI rounds before terminating at a pull request, all within Stripe's hundreds of millions of lines of Ruby code.
4. White House Unveils National AI Policy Framework, Seeks Federal Preemption
On March 20, the Trump administration released the National Policy Framework for Artificial Intelligence — legislative recommendations urging Congress to establish a unified federal approach and preempt state AI laws deemed overly burdensome. The framework covers child safety, consumer protection, data center energy costs, national security, and IP. Congressional Democrats immediately countered with the GUARDRAILS Act, which would block efforts to impose a moratorium on state-level AI regulation.
5. OpenAI Deploys GPT-5.3-Codex-Spark on Cerebras Wafer-Scale Chips
OpenAI launched GPT-5.3-Codex-Spark, its first production AI model running on Cerebras Systems' wafer-scale chips instead of Nvidia GPUs. The streamlined coding model runs at over 1,000 tokens per second, enabling near-instant feedback for real-time software development. The move follows OpenAI's $10 billion deal with Cerebras in January, marking a strategic diversification away from Nvidia dependency. The model is rolling out as a research preview to ChatGPT Pro subscribers.
6. Google Research Teaches LLMs to Reason Like Bayesians
Google Research published a method for training large language models to approximate Bayesian reasoning by mimicking the predictions of an optimal Bayesian model. Without specific training, LLMs default to simple heuristics; the new approach significantly improves performance and generalizes to unseen tasks. Bayesian-tuned versions of Gemma-2-9B and Llama-3-8B achieved an 80% agreement rate with normative Bayesian strategies, pointing toward more principled probabilistic reasoning in future models.
7. AI Industry Pours $100M+ Into 2026 Midterm Elections
With federal AI regulation looming, the AI industry is spending heavily on the 2026 midterms. Innovation Council Action, a political group tied to two Trump advisors, announced plans to spend at least $100 million. Interest groups funded by AI industry leaders are split on how the government should oversee AI, and the spending is already shaping political ads and campaign narratives ahead of November. The stakes are high as both the White House framework and state laws like the Colorado AI Act compete for regulatory dominance.
8. AI Virtual Try-On Tech Goes Mainstream in Retail
AI-powered virtual try-on technology is reshaping online retail, with Google integrating the feature directly into product search results across its platforms starting April 30. A wave of startups is tackling retail's "silent killer" — high return rates — by letting shoppers visualize fit and style before purchasing. The technology combines computer vision and generative AI to overlay garments on customer photos, helping retailers cut return-related losses that cost the industry billions annually.
// KEY TAKEAWAYS
This week's AI landscape reveals a field rapidly moving from research demos to real-world deployment. The neuro-symbolic energy breakthrough and Google's Bayesian training method show fundamental research still driving efficiency gains, while Stripe's 1,300 weekly autonomous PRs and Utah's expanding AI prescription pilots demonstrate production-scale AI already reshaping industries. Meanwhile, the collision between the White House's federal preemption push and the $100M flowing into midterm elections signals that AI governance is becoming the defining policy battle of 2026.