Daily digest ยท 30 March 2026

AI Industry
News Digest

xAI

Grok Imagine Drops Its Free Video Generation Tier

If you have been using Grok Imagine to create AI videos without paying anything, that ends from today. xAI has removed the free tier for Imagine video generation entirely, placing all video creation behind paid SuperGrok plans. The entry point is SuperGrok Lite at $10 per month, which provides a limited number of daily creations at 480p resolution and clips of up to six seconds. Users who already hold X Premium or Premium+ subscriptions retain access through those plans, so it is not exclusively a SuperGrok product.

An xAI team member confirmed the change was driven by the growing Imagine user base making the free tier economically unsustainable. For anyone building short-form AI video content on a zero-cost basis, this is a direct disruption to your current workflow.

Sources
EONMSK, 29 Mar
Anthropic

Claude Code Fixes Broken Cowork Dispatch and Ships Broad Reliability Improvements

The 29 March Claude Code release addresses an issue that had been causing messages in Cowork Dispatch to go undelivered, a bug that would have quietly broken automated agentic workflows for anyone using Dispatch to coordinate tasks between devices. Beyond the fix, this is one of the more substantive Claude Code updates in recent weeks. A new X-Claude-Code-Session-Id header is now attached to API requests so proxies can aggregate usage by session without needing to parse the request body, file and skill handling has been made smarter, and VS Code behaviour has been refined.

Long-session users will also notice reduced scroll-to-top resets and less terminal flickering. A rate-limit warning banner showing usage percentage and reset time is now visible directly in the VS Code panel. These changes reflect how actively Anthropic is iterating on its developer tooling now that Claude Code and Cowork are in production use.

Sources
Claude Code Changelog, 29 Mar
Nothing significant today
OpenAI Nothing significant in products or releases in the last 24 hours. For context: OpenAI revamped ChatGPT's shopping experience and began rolling out Walmart's Sparky assistant as an in-app integration earlier this week (announced 24 March, rollout continuing). The Sora video app shutdown, also announced on 24 March, is drawing further analysis coverage today, though neither item was first published within the last 24 hours.
Google DeepMind Nothing significant in products or releases in the last 24 hours. The week's major Google items (Gemini Drop: chat history import, Lyria 3 Pro, Flash Live developer preview) were published on 26 to 27 March, just outside the 24-hour window.
Microsoft Nothing significant in products or releases in the last 24 hours.
Meta AI Nothing significant in products or releases in the last 24 hours.
Mistral Nothing significant in products or releases in the last 24 hours. Voxtral TTS, Mistral's first open-weight text-to-speech model, launched on 26 March, just outside this window.
DeepSeek Nothing significant in products or releases in the last 24 hours. DeepSeek V4 remains unannounced. The 'V4 Lite' label seen on DeepSeek's website earlier in March has not been officially confirmed by the company.
Nvidia Nothing significant in products or releases in the last 24 hours. GTC 2026 announcements (Vera Rubin, NemoClaw, NeMo Agent Toolkit, Nemotron 3 model family) were made on 16 to 18 March.
IBM Nothing significant in products or releases in the last 24 hours.
Snowflake Nothing significant in products or releases in the last 24 hours.
Perplexity Nothing significant in products or releases in the last 24 hours.
Hugging Face Nothing significant in products or releases in the last 24 hours.
ElevenLabs Nothing significant in products or releases in the last 24 hours.

Industry Themes

Two clear patterns emerge from today's news, and a third is lurking in the background.

Theme 1

Free tiers on compute-heavy generative features are disappearing

xAI's decision to put Grok Imagine video generation behind a paywall is the most concrete example today, but it follows OpenAI's Sora shutdown earlier this week, a product that proved too expensive to run at consumer scale. Both moves signal that the era of using frontier video generation for free is largely over. If you are building any workflow that depends on free-tier AI video, it is worth mapping out your actual costs now before more providers follow. The companies that can sustain free access will be those where video is a hook for a larger paid product, not a standalone offering.

Theme 2

Agentic developer tooling is being treated as production infrastructure

The fact that a bug breaking Cowork Dispatch message delivery received a same-day fix in today's Claude Code release tells you something about the state of agentic tooling: developers are now relying on these tools in live workflows, and interruptions are felt immediately. March 2026 has seen an unusually high number of Claude Code releases, each addressing specific friction points in agentic sessions. This is a different kind of product cadence to traditional software updates: the rapid iteration reflects real-time production feedback from developers who are building on top of these tools every day. If you are evaluating Claude Code for serious agentic development work, the support and release velocity right now is a meaningful signal. Note: the xAI and Anthropic stories above surfaced via official channels before mainstream press coverage.

Theme 3

AI is finding its natural role in commerce as a discovery layer, not a transaction processor

OpenAI's pivot away from Instant Checkout and towards retailer-linked in-app experiences (Walmart's Sparky, announced earlier this week) reflects a broader settling of expectations about where AI actually adds value in the shopping journey. The model that is emerging (AI helps you find and compare, the retailer handles the transaction) suits both parties better. Expect most of the major AI assistants to converge on this pattern over the coming months as they expand their commerce integrations beyond the early experiments.

โ† Previous issue Back to Archive Subscribe for tomorrow's issue โ†’

Get tomorrow's digest

Fresh AI intelligence in your inbox every morning, before 8am.

โœ“ You're on the list. First issue lands tomorrow morning.