AI Product Updates: New Features and Releases

Grok Bot lets an agent query ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews to see which brands they recommend first and why.

A developer tries ChatGPT's new Spaces feature, using it to hold checklists, examples and Codex threads that run an entire video production pipeline inside ChatGPT.

Eleven, an agent-first reverse engineering tool, gets a big update: open any app or binary and the agent takes it apart with you.

The Claude Code desktop app adds nested sessions, sidebar groups (Needs input, Working, Completed) and /btw side questions, and can have Opus 5.5 scan past sessions for unfinished tasks and open each as a new session.

Agency Unbound V2 launches, giving each agent, beyond its coin, wallet and mind, the ability to launch paired with 200+ assets, write and run its own code, keep datasets and publish charts, and build and ship sites.

Conclave puts several AI models in one workspace with side-by-side comparison, route tracing and Brief records, paid for with Solana USDC.

A micro1 Economics Research Unit paper on 2,611 production interviews finds reviewer identity predicted scores more than interview timing, and that agent evaluations drift after roughly five weeks without retesting.

Manufact has launched a global MCP leaderboard at manufact.com/mcp-catalog where users can see how each MCP server ranks.

Alexandria now offers product search across Shopify stores for price comparison, stock and reviews, aggregating 50+ providers with one API key and 1,000 free credits.

Odessia's concierge agent books trips and reservations over text, keeps watching fares and tables after you ask, and claims accredited travel agent status for agent rates.

Build-A-Chip-Bench tests whether AI agents can autonomously design chips; the best agent scored 16.6%, with about 40% of tasks passing functional verification.

The Claude Code desktop app now starts sessions faster, makes it easier to split work across several, and shows what each session is doing.