AI Daily Digest: October 3, 2026 - Claude Code Mods, Muse on Smart Glasses, DGX Spark 64GB, Tesla Q3 Beat
Claude Code is now hackable with small TypeScript mods, Meta's Muse agent is moving onto smart glasses, and NVIDIA gave its local AI box a new memory tier with a new price. Anthropic's IPO slipped again according to Bloomberg, Tesla beat Q3 delivery estimates, and Microsoft shipped three voice models in a single day. Seven stories, sources at the bottom of each. Anthropic opens up Claude Code…
Anthropic unveiled a feature called mods, which are small TypeScript functions that can alter how Claude Code operates. These mods can modify prompts, manage tool calls, approve permissions, filter tool output secrets, and even customize UI elements like buttons and panels. Mods are installed through plugins and require Claude Code v2.1.287 or later. Anthropic is also integrating its own built-in features into these mods, such as /diff and AGENTS.md loading, among others.
Meta unveiled its AI agent Muse, which is set to become available on smart glasses. Users can initiate Muse using voice commands like "Hey Muse" or a custom name, allowing tasks to run in the background until completion. Muse operates on the MuseSpark model and functions on a private, persistent Linux VM in the cloud, complete with its own browser, filesystem, and terminal.
Users can perform various tasks such as reservations, flight searches, part sourcing, meal logging, workout guidance, and in-store purchases. Payments through Stripe via one-time virtual cards ensure no exposure of actual financial details. Meta assures that Muse data is not utilized for advertising and can be opted out of for future model training.
The launch of Muse on September 8 and its positioning as the center of Meta's consumer AI push were highlighted in Connect 2026.
NVIDIA released a new 64GB unified memory configuration for DGX Spark, starting at $4,999 with a shipping date of October 23. This configuration, sold through OEMs including Acer, ASUS, Dell, Gigabyte, HP, and MSI, retains the GB10 Grace Blackwell superchip, DGX OS, CUDA software stack, and ConnectX-7. The hardware retains most of its previous specifications, with only 8GB of the 64GB reserved for system use, leaving around 56GB for model weights and KV cache.
NVIDIA claims that models up to 100 billion parameters can run locally, with two 64GB units connected over QSFP forming a 128GB memory pool capable of supporting models up to 200 billion parameters. Pricing adjustments and local agent workflows are emphasized as key selling points for this new configuration.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
