Together with
|
BUILDER HUSTLE
AI · TECH · STARTUPS
|
|
|
In This Issue
|
|
💸 Freehand AI Audits Every Invoice and Recovers Overpayments
|
|
🤖 Gemini Robotics 2 Brings Whole-Body Intelligence to Robots
|
|
🎵 Lyria 3.5 Makes Google Flow Music More Expressive and Controllable
|
|
🧠 Kimi K3 Now Runs Locally on a Mac Studio Through Unsloth
|
|
|
💸 FREEHAND AI · INVOICE AUDITING · ENTERPRISE AI
|
|
Freehand AI Raised $75M to Audit Every Invoice, Not Just a Sample.
|
|
|
Freehand AI has raised 75 million dollars after reporting 260 million dollars in recovered overpayments for companies including Meta, Unilever, and Johnson & Johnson. The company says its system has examined more than 19 million invoices, matching charges against contracts, rate cards, purchase orders, bills of lading, and historical records to identify disputed fees. It can then assemble evidence, draft disputes, communicate with suppliers, follow up, and track recovery.
|
|
📌 Full invoice coverage can uncover errors that sample-based manual audits never review.
|
|
⚡ The important leap is not just flagging an issue, but automating the evidence and recovery workflow that follows.
|
|
|
Learn How to Stay Visible in the AI Era
AI is changing how customers discover businesses. If your SEO strategy is built for yesterday's search, your visibility is already slipping. Learn how to optimize your content for today’s AI search results with BELAY’s latest report..
|
🤖 GOOGLE DEEPMIND · GEMINI ROBOTICS 2 · EMBODIED AI
|
|
Gemini Robotics 2 Gives Robots More Dexterity, Whole-Body Control, and Teamwork.
|
|
|
Google DeepMind has launched Gemini Robotics 2, a new generation of models built to help robots reason and act in the physical world. The lineup includes Gemini Robotics ER 2, a vision-language embodied-reasoning model that can communicate with people, understand a scene, plan multi-step tasks lasting several minutes, and coordinate actions with robots. DeepMind also introduced a vision-language-action model for advanced whole-body control and dexterity, plus an on-device version designed for lower-latency local operation.
|
|
📌 Embodied reasoning gives robots a higher-level layer for planning, progress tracking, and adapting when the environment changes.
|
|
⚡ Multi-robot coordination matters for warehouses, manufacturing, and household tasks that cannot be completed effectively by one machine alone.
|
|
|
Read Google DeepMind's Gemini Robotics 2 announcement →
|
|
200 Ways To Make Money With AI
Turns out, AI is good for more than writing emails and generating LinkedIn posts. This guide is packed with 200+ actionable ways to build income with AI.
200 curated AI income ideas for beginners and pros alike
Easy-to-start opportunities you can launch this week
Real world applications for today’s top AI tools
AI powered business models built for today’s economy
Creative ways to turn trends into revenue
AI is changing how people work and how people earn, far beyond the simple “write me an email” prompt. Discover the possibilities and download the free guide to start cashing in today.
|
🎵 GOOGLE DEEPMIND · LYRIA 3.5 · FLOW MUSIC
|
|
Lyria 3.5 Gives Creators More Control Over AI-Generated Songs in Flow Music.
|
|
|
Google has added Lyria 3.5 to Flow Music, upgrading its AI music generator with improved musicality, lyric quality, vocal expression, and creative control. The model is designed to produce richer arrangements and more dynamic vocal performances while following direction more closely. New controls include BPM and song-duration settings, helping creators target a specific tempo and structure. Flow Music also lets users export stems from generated tracks, allowing vocals, drums, bass, and other elements to be edited separately in a production workflow.
|
|
📌 BPM control and stem export turn a generated song into material that can be adapted inside a conventional music-production workflow.
|
|
⚡ Better vocals and arrangement coherence address two of the biggest limitations that have made AI music feel less usable for finished work.
|
|
|
Explore Lyria 3.5 in Flow Music →
|
|
|
🧠 UNSLOTH · KIMI K3 · LOCAL AI
|
|
Unsloth Compresses Kimi K3 So a 2.8T-Parameter Model Can Run Locally.
|
|
|
Unsloth has released dynamic GGUF quantizations for Moonshot AI's Kimi K3, including a 1-bit version that reduces the model from roughly 1.56TB to 594GB. Unsloth says this version retains about 78.9% accuracy by its measurement and can run on a Mac Studio connected to a machine with 128GB of RAM, or on an NVIDIA DGX Station. The project also provides 2-bit, 3-bit, and near-lossless Q8 options for users who have more available memory. Models can be downloaded and managed through Unsloth Studio for local use.
|
|
📌 Quantization compresses model weights so large systems become feasible on local hardware, with a trade-off in output quality.
|
|
⚡ Running locally can improve privacy and remove per-token cloud costs, but 594GB is still a demanding storage and hardware requirement.
|
|
|
See Unsloth's guide to running Kimi K3 locally →
|
|
Stay sharp,
Builder Hustle
