🧠 Next-Gen AI Mimics Human Thought and Outsmarts ChatGPT

Hierarchical Reasoning Proves More Efficient Than Giant LLMs

Author

Samantha Vale
August 29, 2025

Typing is a thing of the past.

Typeless turns your raw, unfiltered voice into beautifully polished writing - in real time.

It works like magic, feels like cheating, and allows your thoughts to flow more freely than ever before.

Your voice is your strength. Typeless turns it into a superpower.

Hey there, Tech Trailblazers! 🤖✨

In this issue, we highlight an AI toolkit built for small businesses that automates everyday tasks such as research, content creation, and customer engagement.

It gives lean teams the ability to compete at a higher level without the cost of large tech investments

📰 Upcoming in this issue

  • 🧠 New HRM AI Beats ChatGPT on Tough AGI Tests

  • 🧰 AI Toolkit To Accelerate Small Businesses

  • 🧹 New Method Lets AI Forget Private and Copyrighted Data

🧠 New HRM AI Beats ChatGPT on Tough AGI Tests

Sapient’s hierarchical reasoning model scores 40.3% on ARC-AGI-1 with just 27M parameters and 1,000 samples. Replicators confirm results, but credit training refinements.

Key Takeaways:

  • 📊 Outscores On ARC-AGI: HRM scores 40.3% on ARC-AGI-1 and 5% on ARC-AGI-2, topping O3-mini-high, Claude 3.7, and DeepSeek R1.

  • 🧠 Tiny Model, Sparse Data: With 27 million parameters and 1,000 samples, the model claims efficient learning and faster reasoning than typical LLMs.

  • 🧩 Brain-Inspired Architecture: High-level planning and low-level calculation modules enable single-pass sequential reasoning with iterative refinement across short thinking bursts.

  • ⚠️ Peer Review And Caveats: Results appear on arXiv and ARC-AGI reproductions, which attribute gains partly to a training refinement process rather than architecture alone.

🧰 AI Toolkit To Accelerate Small Businesses

New Scientist rounds up a practical AI toolkit for small businesses. Smart picks, setup tips, and guardrails help owners unlock value fast.

Key Takeaways:

  • 🧪 Start Small: Prioritize one or two pain points, pilot tools, and document wins before expanding companywide.

  • 🔗 Integrate Where You Work: Choose AI that integrates with email, docs, CRM, and chat, so work flows without copy paste.

  • 🔐 Safety And Policy: Protect data, set use policies, and enable multi factor authentication to reduce risk while experimenting.

  • 📏 Upskill And Measure: Teach prompt basics, track time saved, and reinvest gains into marketing, service, and product improvements.

🧹 New Method Lets AI Forget Private and Copyrighted Data

UC Riverside researchers unveil “source-free certified unlearning” to erase targeted data without original datasets. It promises privacy close to full retraining with far less compute.

Key Takeaways:

  • 🧠 Source-Free Unlearning: Technique removes targeted information without access to original training data, preserving model utility.

  • 🧪 Surrogate Data + Noise: Developers use surrogate datasets and calibrated random noise to adjust parameters and erase recoverable traces.

  • 🛡️ Compliance And Rights: Supports GDPR and CCPA obligations and helps media and medical institutions protect private or copyrighted content.

  • 🚀 Path To Larger Models: Research shows privacy near full retraining with less compute, now for simpler models, with plans to extend.

Why It Matters

AI is no longer optional for small businesses. Leaders who adopt simple, ready-to-use toolkits can unlock efficiency, save time, and scale smarter in a competitive market.

Samantha Vale
Editor-in-Chief
Get Nerdy With AI

How was today's edition?

Rate this newsletter.

Login or Subscribe to participate in polls.