Skip to content
MustHave.ai
MustHave Hoodie Seal
THE MUSTHAVE NEWSROOM

AI news

What’s new. What matters. What do you think?

Top stories

By member votes

Anthropic Says Claude Leads 26% of Its AI R&D: What the Metric Actually Counts

Anthropic AI R&D metrics say Claude leads 26% of measured work and supports more than 90%. The denominators and verification limits change the story.

Figma Weave Can Turn Any Design Frame Into a Brand-Controlled AI Workflow

Figma Weave can use an existing design frame as a workflow node, preserve selected structure and generate controlled campaign or localized variants.

GPT-5.5 Retirement Is October 14: The ChatGPT and Codex Migration Checklist

The GPT-5.5 retirement reaches ChatGPT and Codex on October 14. Audit saved model choices, agents, scheduled work and scripts before the cutoff.

ChatGPT for Word Is Available on Every Plan, but It Can Use Your Codex Allowance

ChatGPT for Word is now available on every ChatGPT plan. Learn what it can edit, what context stays separate and when it shares Codex usage.

cPanel Meridian AI Features: A Complete Guide to the Assistant, Sitejet, Node.js and MCP

cPanel Meridian AI features now connect account guidance, Sitejet website generation, Node.js app hosting and WebPros MCP. This guide explains setup, limits and security.

Google CC Agent Expands to Families With Shared Memory and Separate Permissions

The Google CC agent now supports groups of up to six with shared memory and per-person permissions. Here is the household data and action boundary.

Raindrop Raises a Series A for AI Agent Simulations That Replay Production Before Deployment

Raindrop Simulations replay agent traffic and tests on every pull request before production. The company also announced a Series A and $50 million raised.

Anthropic Releases 36 Biomolecular Optimization Kits After Reporting Up to 4x Speedups

Anthropic biomolecular kits package 36 model optimizations with reported speedups. Here is how to reproduce results and handle the security assumptions.

Figure Helix 2.5 Reaches 56% in 30 Unseen Homes, but Zero-Shot Needs Context

Figure Helix 2.5 reached a company-reported 56% success rate across 30 unseen homes. We audit the protocol, baseline and meaning of zero-shot.

Anthropic Life Sciences Verification Program Opens High-Risk Claude Access Under Review

The Anthropic Verification Program creates standard and high-risk life-science access paths. Here are the eligibility, privacy and renewal boundaries.

Google Antigravity Agent 09-2026 Replaces the May Preview: Migrate Before October 5

Antigravity Agent 09-2026 replaces Google's May preview and changes file-tool names and edit schemas. Use this deadline-first migration checklist.

OpenAI Astra for Law Adds a Daily Legal Search Index, With Early Access Limits

OpenAI Astra Law combines GPT-6 Astra with a daily U.S. legal search index. Here is what it covers, who can use it and what lawyers still must verify.