Good morning, AI enthusiasts. Accenture's CEO Julie Sweet delivered a stark message this week cutting 11,000 employees who couldn't be retrained for AI roles fast enough, while growing the firm's AI talent base to 77,000 and training 550,000 workers in generative AI capabilities.
Is this the blueprint other consulting firms and tech companies will follow in 2025?
In today's recap:
Accenture cuts 11K jobs, trains 550K in AI
New optimizer stabilizes neural network training
MIT tool slashes medical imaging work by 75%
8 new AI tools and more
AI & JOB
Accenture cuts 11K jobs in AI pivot
Source: Getty Images
Recaply: Consulting giant Accenture is reshaping its workforce with a clear message: "adapt to AI or leave." The company cut over 11,000 jobs in three months while investing heavily in training workers for an AI-driven future.
Key details:
Accenture reduced its global workforce from 791,000 to 779,000 employees between May and August 2025.
CEO Julie Sweet says the company is "exiting" staff who cannot be retrained for AI roles on a compressed timeline.
The firm launched an $865 million restructuring program expected to save over $1 billion.
Despite cuts, Accenture trained 550,000 workers in generative AI and grew its AI talent from 40,000 to 77,000 professionals.
The company plans to increase headcount next year in key markets while continuing to prioritize AI skills.
Why it matters: This move signals a broader trend across the tech sector. Companies are not simply downsizing but rotating talent to match AI demands. The message to workers is stark: continuous learning is no longer optional. As AI reshapes job requirements, professionals must stay ahead or risk being left behind.
THINKING MACHINES
AI gets smarter training method
Source: Thinking Machines
Recaply: Thinking Machines just dropped their second research piece on something called Modular Manifolds. Think of it as giving AI models a personal trainer that keeps them in peak shape during learning. Instead of letting neural networks get sloppy during training, this method keeps their internal components properly sized and balanced.
Key details:
The method solves the problem of AI weights growing too large or small during training.
It introduces a "manifold Muon" optimizer that constrains matrices to specific shapes.
It prevents numerical issues that can break AI model training.
It makes training more predictable and reliable across different network sizes.
The work was built through collaboration with researchers Jianlin Su and Franz Louis Cesista.
Why it matters: This could make AI training more stable and efficient. When neural networks stay healthier during learning, they perform better and require less babysitting from engineers. It's another step toward making AI development more automated and reliable.
AI RESEARCH
New AI cuts medical image work by 75%
Source: MIT
Recaply: Researchers at MIT just solved a major time drain in medical research. Their new AI system called MultiverSeg lets scientists segment medical images with simple clicks and scribbles instead of hours of manual work. Before this tool, researchers analyzing brain scans or X-rays had to manually outline regions of interest in every single image. Now they can train the AI once and watch it handle entire datasets automatically.
Key details:
Scientists can now segment entire image datasets without repeating work for each image.
The AI learns from previous images, eventually needing zero user input after initial training.
By the ninth image, users only need two clicks for accurate results that match specialist-trained models.
No machine learning expertise or special computational resources are required to get started.
MultiverSeg reduces manual segmentation time from hours to minutes per image across various medical imaging types.
Why it matters: This breakthrough could transform how quickly medical studies happen. When researchers spend less time on tedious image work, they can focus more energy on finding treatments and cures that help patients. Studies that were previously too time-consuming to conduct become feasible, potentially unlocking medical discoveries that could save lives.
NEWS
Beyond the big stories in AI today
Factory AI released Droids publicly, a software development agent ranking #1 on Terminal-Bench with 58.75% accuracy, outperforming Claude Code and Codex CLI.
Anthropic plans to triple its international workforce and expand its applied AI team fivefold in 2025, serving 300,000+ enterprise customers with 80% usage outside the U.S.
xAI announced a government partnership providing Grok AI models to federal agencies for $0.42 per organization through March 2027, including dedicated engineering support.
Tencent released HunyuanImage-3.0, an 80-billion-parameter multimodal image generation model that handles complex prompts and long text rendering for commercial use.
Inclusion AI released Ring-flash-linear-2.0 and Ring-mini-linear-2.0, reasoning models with hybrid architecture achieving 40B dense model performance while activating only 6.1B parameters.
YouTube introduced YouTube Labs featuring AI music hosts that share stories, trivia, and commentary about music within the YouTube Music app for select U.S. users.
Warp introduced Warp CLI functionality enabling developers to run AI agents directly from terminal interfaces for coding assistance and command automation.
Apple is building an internal ChatGPT-like app code-named "Veritas" to test revamped Siri capabilities, targeting a March 2026 launch with enhanced personal data search.
TOOLS
Trending AI Tools
🎨 Turbologo - Generate professional logos instantly with AI
✍️ Sudowrite - AI writing partner specifically designed for fiction
🎥 Jogg AI - Generate avatar-based videos from URLs or existing content instantly
🎬 Short AI - Turn long videos into viral short clips with captions
💻 Astra - Fast, customizable WordPress theme with pre-built sites
🚀 BrandWell - All-in-one AI engine for writing, detection, and growth
📚 LearnWorlds - Create, market, and sell online courses with AI tools
🎤 PodBean - Podcast hosting platform with monetization features
