AI Product Updates: New Features and Releases

A morning roundup covering Meta's Muse Spark Open, Opus 5.5 fast mode leaving the waitlist, GPT-Live-1.1, reduced Codex capacity errors, and more.

Vercel now supports Python and has released an eve Python Agent template with a Next.js frontend and FastAPI backend for building agent apps.

Codex CLI 0.163.0 adds /config search with instant edits and restart flags, a leader key in /keymap defaulting to Ctrl+X, mid-turn follow-ups, and parallel read-only tools.

Vistara released a free Mac beta for Claude Code, Codex and Cursor users, focused on defining what counts as done and verifying results as groundwork for agents hiring and paying each other.

Opus 5.5 fast mode exits beta with no waitlist and up to 2.5x faster output; subscribers pay from extra usage rather than plan limits, and API pricing is double the standard rate.

Beeline is an MIT-licensed team chat where a whole team drives agents from one shared control plane, with GitHub worktrees, 1,500+ app integrations and real-time voice.

Meta's open-weight Muse Spark Open is now on Merge Gateway, aimed at coding and long-running agent work with a 1M-token context and 42B active parameters out of 816B.

Open source project OpenUI releases 1.0, so agents can respond with interactive forms, charts and tables instead of plain text.

Movit has launched a peer-to-peer delivery platform where travelers between cities can carry packages for senders heading the same way, now in beta.

Attio introduced an SQL engine, BI connectors and agent scale, letting agents query customer data, run weighted forecasts and handle up to 100 million records per object.

Photon is in public beta, letting a single agent send messages and make calls across multiple channels, aimed at users outside iPhone, with $100 in credits for builders.

PrismML launched Bonsai 2, open-weight ternary models in 2B, 4B and 9B sizes, with the 2B running at 90 tokens/sec on an iPhone 17 Pro Max and supporting reasoning, vision and tool calling on-device.