Memory Breakthrough Turns Experiences Into Agent Skill 🧠

From Trials to Tactics Agents Reuse What Works

Author

Samantha Vale
October 09, 2025

In partnership with

Your Boss Will Think You’re an Ecom Genius

If you’re optimizing for growth, you need ecomm tactics that actually work. Not mushy strategies.

Go-to-Millions is the ecommerce growth newsletter from Ari Murray, packed with tactical insights, smart creative, and marketing that drives revenue.

Every issue is built for operators: clear, punchy, and grounded in what’s working, from product strategy to paid media to conversion lifts.

Subscribe for free and get your next growth unlock delivered weekly.

Hey there, Tech Trailblazers! 🤖✨

In this issue, researchers unveil ReasoningBank, a new memory framework that enables AI agents to learn from both successes and failures, applying those lessons to future tasks. It pairs structured memory with test-time scaling, allowing agents to compare multiple solution paths and refine stable strategies.

Across benchmarks such as WebArena and SWE-Bench-Verified, ReasoningBank outperformed both memory-free agents and traditional memory systems, improving success rates by up to 8.3 percentage points while reducing the number of interaction steps. The result is more adaptive, efficient agents that will continuously enhance in real workflows.

📰 Upcoming in this issue

  • 🧠 New Memory Framework Helps AI Agents Handle Real-World Unpredictability

  • 🫀 AI Tools Bring New Hope for Cardiovascular Medicine

  • 🤖 From Automation to Agency: Why Agentic AI is a New Era for HR Tech

🧠 New Memory Framework Helps AI Agents Handle Real-World Unpredictability

Researchers introduce ReasoningBank, a system that converts an agent’s experiences into reusable strategies. Combined with memory-aware test-time scaling, it delivers higher accuracy and efficiency across coding and web tasks.

Key Takeaways:

  • 🧩 ReasoningBank Basics: Distills strategies from both successes and failures, storing structured memories that agents can retrieve during inference.

  • 📈 MaTTS Boost: Memory-aware test-time scaling compares parallel or sequential trajectories to enhance accuracy, consistency, and efficiency.

  • 🧪 Benchmark Gains: Outperforms memory-free and workflow memory models on WebArena and SWE-Bench-Verified while cutting interaction steps.

  • 🏢 Enterprise Impact: Reduces trial-and-error costs and powers adaptive agents for customer support, software development, and data analysis.

🫀 AI Tools Bring New Hope for Cardiovascular Medicine

Researchers believe that AI, omics, and systems biology can enable precision heart treatments that were once considered impossible. A growing call for global collaboration could save millions of lives.

Key Takeaways:

  • 🔎 Precision Medicine Push: AI, omics, and systems biology pinpoint targets and personalize therapies for specific heart disease pathways.

  • 🧬 RNA Therapies Rise: RNA-based drugs regulate multiple genes, offering faster development and stronger cholesterol control than traditional options.

  • 💊 One-Size Limits: Standard treatments like statins ignore biological variation, creating significant outcome gaps.

  • 🌍 Global Collaboration Needed: Experts call for open science, funding, and leadership to expand access and reduce the projected 26 million annual heart-related deaths by 2030.

🤖 From Automation to Agency: Why Agentic AI is a New Era for HR Tech

UNLEASH spotlights Everest Group’s Sharath Hari on the next phase of enterprise AI — agents that act, adapt, and collaborate. He argues HR leaders are well-positioned to guide adoption and turn pilots into measurable value.

Key Takeaways:

  • 🧩 What Agentic Means: Agents coordinate tasks, adapt in real time, and work across systems, moving beyond simple automation.

  • 📉 Maturity Reality: Most organizations are still in early assistant or quasi-agent phases, with fully autonomous agents emerging slowly.

  • 🧭 HR’s Role: HR-driven pilots improve efficiency, compliance, and employee experience, paving the way for broader adoption.

  • 🤝 Getting Ready: Success depends on leadership, quality data, secure infrastructure, and pragmatic scaling with the right partners.

Quick poll 💬

After reading, what will your team apply first?

Login or Subscribe to participate in polls.

Why It Matters

Enterprises can build agents that learn from experience, reduce trial-and-error costs, and handle messy real-world tasks more reliably. Memory, combined with scalable inference, points toward agents that manage larger parts of software, support, and data processes with less oversight.

Until our next issue,

Samantha Vale
Editor-in-Chief
Get Nerdy With AI

P.S. Interested in sponsoring a future issue? Just reply to this email and I’ll send packages!

How was today's edition?

Rate this newsletter.

Login or Subscribe to participate in polls.