Get tomorrow's brief in your inbox
Today: Adobe put creative AI assistants in Photoshop, Premiere, and Illustrator to automate tedious production tasks. OpenAI added granular spend controls and usage analytics to ChatGPT Enterprise. Federal regulators fast-tracked data center grid connections while AI inference startups raised billions.
Adobe is rolling out AI assistants across its Creative Cloud apps, and this isn't just another generative fill feature. The company added specialized assistants to Photoshop, Premiere, Illustrator, InDesign, and Frame.io that can handle multi-step production workflows through conversational prompts.
For small businesses producing marketing content without dedicated design staff, this changes the cost structure. Premiere's assistant can sort footage into bins, batch rename clips based on what's happening in the video, identify interview questions in recordings, and assemble a working first cut. That's the kind of work that typically takes an editor hours. Photoshop's assistant can swap backgrounds, resize assets for multiple platforms, and organize layers. Illustrator can generate versioned files from spreadsheet data and run pre-flight checks for print issues.
Adobe also expanded Firefly into a unified creative studio with two features that matter for brand consistency: Elements (save characters, locations, and objects for reuse across generations) and Projects (keep assets and context organized across Firefly and Creative Cloud). If you're running seasonal campaigns or product launches, you can maintain visual consistency without manually describing your brand every time.
The Quick Cut feature in Premiere automatically assembles video clips into a polished first draft based on dialogue or visual content. For businesses producing customer testimonials, product explainers, or social videos, that removes the most time-consuming early editing work. You still review and refine, but you're starting from a working cut instead of a blank timeline.
Adobe is also bringing these creative tools to platforms where business owners already work: ChatGPT, Claude, Microsoft 365 Copilot, Gemini, and Slack. The bet is that you'll describe what you need wherever you're already working, and Adobe's tools will execute it. These assistants are in public beta starting today.
Read more | The Verge coverage
OpenAI added spend controls for ChatGPT Enterprise. Admins can now track credit usage across users, products, and models in the Global Admin Console, set default workspace limits, configure limits for specific groups, and create individual overrides for power users. Employees can see their credit usage against their budget and request additional credits with context about what they're working on. This solves the problem where one team's usage blows the budget for everyone else. Available today for ChatGPT Enterprise admins. Read more
GPT-5.5 Instant brought frontier health intelligence to free ChatGPT users. OpenAI says the model performs comparably to their Thinking models on health evaluations, with better recognition of when urgent care is needed, improved handling of uncertainty, and clearer explanations. Physicians rated GPT-5.5 Instant responses as having fewer failure modes than older models and even human physician responses. Factuality issues in health responses dropped 71% in the last two months based on production traffic (billions of messages per week). Free users can access this today. Read more
Simon Willison launched Datasette Apps for building custom HTML tools. This lets you host self-contained HTML+JavaScript applications inside Datasette that can run read-only SQL queries against your data. Apps run in a tightly sandboxed iframe that prevents them from accessing cookies, localStorage, or making HTTP requests to outside hosts. If you're building internal tools or custom dashboards on top of SQLite databases, this is a faster way to prototype than building a full web app. Free and open source. Read more
Amazon is considering selling its Trainium AI chips directly to other companies. AWS AI chief Peter DeSantis told Bloomberg the company is in talks to sell chips for use in data centers. CEO Andy Jassy said in April that if AWS sold chips to third parties, it would be a $50 billion annual run rate business (roughly Intel's size). The catch: AWS's current Trainium capacity sold out almost instantly, and the next-gen Trainium4 (not available for over a year) is already sold out. Selling to others means leaving current customers on waiting lists unless AWS can manufacture a surplus through TSMC, which is currently prioritizing Nvidia. This signals that AWS sees a path to challenging Nvidia's AI chip dominance, but execution will require solving the capacity problem first. Read more
Snap spun off its AI video team into a separate company to cut costs. The new company, Dotmo, will focus on AI models that create interactive gaming experiences. Snap CTO Bobby Murphy will be the lead investor with a significant personal stake but will continue working full-time at Snap. Snap gets a large equity stake in Dotmo in exchange for the talent and technology license. This is Snap's second major spinoff this year (after Specs smart glasses) and follows 1,000 job cuts earlier in 2026. Spin-offs can reduce financial burden while maintaining exposure to upside through equity. If your AI R&D costs are outpacing your core business needs, this model lets you offload expenses while keeping optionality. Read more
Karamo Brown (Queer Eye) launched Kē, a wellness app with his AI digital clone. The app offers personalized fitness plans, nutrition guidance, meditation videos, and a community section. The standout feature is "AI Karamo" powered by Delphi, which lets you talk with a digital version of Karamo for real-time advice in his voice. The AI pulls from interviews, podcast episodes, and other clips to represent him authentically. Brown says it's designed to help people reflect and grow, not replace human connection. Subscription costs $14.99/month after a three-day free trial. Available on iOS and Android. Note: Using the AI feature means sharing conversation data with Delphi, so avoid disclosing sensitive information. Read more
Match Group surveyed 1,000 singles on AI in dating. 47% have a negative view of AI in romantic contexts. 40% say they would refuse to date someone who uses an AI companion app (51% among women ages 18-24). However, 64% said they could see how AI might help them in their dating journey. People are open to AI helping with profile writing and conversation starters, but they don't want to feel like their dating experiences are inauthentic. The key takeaway for anyone building AI dating tools: help with the hard parts, but hands off the human parts. Read more
Federal regulators fast-tracked data center grid connections. FERC told six major grid operators to ensure data centers "are able to connect to the transmission system in a timely and orderly manner." Data centers pay for interconnection costs. Grid operators have 30 days to report spare generating capacity and 60 days to defend or revise electricity rates. The directive also opens the door to alternative transmission technologies (solid-state transformers, superconducting lines). This addresses delays that have forced tech companies to turn to more expensive behind-the-meter power. Electricity demand from data centers is expected to nearly triple through 2035. Wholesale electricity rates are up as much as 267% compared with five years ago. Read more
Baseten is raising $1.5 billion at a $13 billion valuation. The AI inference company closed a $300 million Series E at a $5 billion valuation just five months ago. This latest round represents a 160% valuation increase in less than half a year. It's a split-priced round (some investors at $13 billion, others at $11 billion) co-led by Spark Capital, Sands Capital, Altimeter Capital, and Wellington Management. Baseten routes inference requests to the best-for-task model to control costs while maintaining speed. The "inference gold rush" continues as VCs pour money into the layer between models and applications. Read more
General Intuition is raising $300 million at around $2 billion valuation. The startup builds foundation models that train AI agents how to move through space and time using Medal's dataset of 2 billion videos per year from 10 million monthly active gamers. The raise comes eight months after spinning out with a $134 million seed round. Backers include Jeff Bezos, Eric Schmidt, Khosla Ventures, and General Catalyst. General Intuition builds world models to train agents (the agents are the product), not to sell the models themselves. The company's unique dataset of interactive, first-person gameplay is designed to teach machines spatial-temporal reasoning. New product expected by end of summer or early fall. Read more
Elastic agreed to buy DeductiveAI for up to $85 million. DeductiveAI uses AI to catch and resolve bugs in software (AI site reliability engineering). The startup came out of stealth last November with a $7.5 million seed round led by CRV at a $33 million valuation. This marks a speedy exit driven by the massive influx of AI-written code that needs debugging. Elastic (known for Elasticsearch) will integrate Deductive's tech into its observability platform to automatically monitor performance and resolve system failures in real time. Deductive reached roughly $1 million in ARR but lagged behind competitor Resolve AI, which was last valued at $1.5 billion. Read more
Guardrails Alliance launched as a grassroots super PAC supporting AI legislation. Democratic operatives Shaunna Thomas and Leah Hunt-Hendrix launched the PAC with backing from tech employees and labor unions. It has $5 million today and plans to raise $15 million this cycle (versus Leading the Future's $100 million from tech leaders like OpenAI president Greg Brockman). Guardrails is buying ads to support Alex Bores, a New York congressional candidate who became Leading the Future's first target. Bores is also receiving support from Public First Action, which has backing from Anthropic. This reflects growing division within tech over AI regulation, with grassroots workers opposing big-tech lobbying efforts. Read more
Automate your product video production. Adobe Firefly can now turn product photography into short-form video with lighting, motion, audio, and brand styling. If you're creating social media assets without a video editor, describe your product style and brand, upload photos, and let Firefly generate the video. This works for product launches, seasonal promotions, and platform-specific content. The Quick Cut feature in Premiere can also assemble your footage into a first draft, saving hours of manual editing on customer testimonials or explainer videos.
Set team-specific AI budgets. If you're using ChatGPT Enterprise, use the new spend controls to set default limits for your workspace, configure limits for specific groups (like your marketing team versus your engineering team), and create individual overrides for power users. Employees can request additional credits with context about their work, so you can make informed decisions without increasing limits for everyone. This prevents one team's heavy usage from blocking another team's access.
Adobe's expansion of AI assistants into every major Creative Cloud app signals that multi-step production automation is becoming table stakes for creative software. The pattern to watch: AI that understands app-specific workflows (video timelines, layer management, print pre-flight checks) rather than generic chat interfaces. OpenAI's spend controls and analytics show that enterprises need granular cost management as AI moves from experiments to production workloads. The infrastructure layer (grid connections, inference platforms, AI chips) is seeing massive investment and regulatory support, which will determine who can scale AI applications profitably over the next two years.