Blog

Evals are the single most important (and most misunderstood) component of any AI deployment, and they deserve a deeper look. Continue Reading →

Anthropic released Claude Opus 4.7 on Wednesday with impressive numbers: 10.9 percentage points higher on SWE-bench Pro (the gold-standard coding test), 3x more production tasks resolved on Rakuten’s benchmark, 98.5% on visual acuity up from 54.5%, and state-of-the-art scores on finance evaluations. For devs, this is a genuine step forward. For consumers, the story is a bit different. Continue Reading →

Adobe just launched Firefly AI Assistant, a conversational agent that orchestrates tasks across Photoshop, Premiere Pro, Lightroom, and Illustrator using natural language commands. Instead of switching between applications and navigating menus, you can tell the assistant to resize images for social media, color-grade footage to match brand guidelines, or generate logo variations. It coordinates the work across whatever Adobe tools the task requires. Continue Reading →

Google just launched Skills in Chrome, which converts your AI prompts into reusable one-click tools. Instead of retyping "make this recipe vegan" across multiple food sites, you save it as a Skill and run it instantly on any page. You can also share these custom workflows or grab pre-built ones from Google's Skills library. Continue Reading →

Microsoft is exploring ways to integrate OpenClaw-style autonomous features into its Copilot AI assistant, according to The Information. Omar Shahine, a corporate VP at Microsoft, confirmed the company is “exploring the potential of technologies like OpenClaw in an enterprise context.” OpenClaw (356k GitHub stars) gained massive popularity earlier this year despite raising serious security concerns. Continue Reading →

As we start to deploy enterprise-grade AI platforms at scale, leadership teams are coming face to face with two formidable challenges. The first is Technical Debt, Ward Cunningham's enduring metaphor for describing outdated technical infrastructure. The second is Cultural Debt, the sum of every unresolved habit, unexamined process, and unspoken assumption your organization carries forward because, well, this is the (your company name goes here) way. Continue Reading →

Anthropic launched Claude Managed Agents in public beta at eight cents per agent runtime hour plus model usage fees. Developers get sandboxed containers, authentication, checkpointing, error recovery, session persistence, and end-to-end execution tracing: every piece of infrastructure that separates a demo from a production deployment, available as a set of composable APIs. Continue Reading →

Meta just launched Muse Spark, its first major AI model since its $14.3 billion investment in Scale AI nine months ago, which pulled Scale AI's founder Alexandr Wang into Meta. This is the inaugural release from Meta Superintelligence Labs, Wang's new unit tasked with rebuilding Meta's AI capabilities from scratch. Continue Reading →

Anthropic announced that it built a model so capable at finding and exploiting software vulnerabilities that it will not release the model to the public. Continue Reading →

Google quietly launched "AI Edge Eloquent," an offline-first dictation app for iOS that cleans up your speech and removes filler words. The app uses local Gemini models to transform rambling speech into polished text, with options for "Key points," "Formal," "Short," and "Long" formatting. It can import keywords from Gmail and work completely offline once the models download. Continue Reading →