Legacy Code

AfterQuery, an AI model-training startup, has raised a new funding round that values the company at $3.2 billion—just five months after closing a Series A at $300 million. The company has become Y Combinator's fastest-ever company to reach unicorn status. This trajectory suggests significant investor confidence in the commercial viability of AI model optimization and training infrastructure.

MIT Technology Review has published analysis on the business case for modernizing legacy systems in an era of generative AI. The traditional narrative around legacy modernization has framed it as a cost center—a complex, disruptive project with unclear ROI. But with AI systems increasingly requiring structured, clean data and APIs to deliver value, legacy infrastructure has shifted from a "nice to have" modernization to a strategic enabler.

OpenAI has published case studies showing how three companies—Basis, Clay, and Exa Labs—have restructured workflows around AI agents to improve core business functions. Basis used AI to streamline customer onboarding; Clay integrated AI into account management and customer communication; Exa Labs built AI-driven developer integrations. In each case, the companies redesigned processes rather than simply bolting AI onto existing workflows.

Anthropic has rolled out updated versions of its Claude models—Fable 5.1 and Mythos 5.1—with a major pricing improvement. The cost per request for cached reads on Fable has dropped 75%, making it substantially cheaper to use for repetitive tasks that reference the same context repeatedly. Both models show measurable improvements in coding tasks compared to their predecessors.

OpenAI has enabled ChatGPT to securely connect to electronic health record (EHR) systems and other healthcare data sources. This means clinicians can now ask ChatGPT questions about patient histories, medical research, and treatment context, with the model accessing real-time data from trusted healthcare systems rather than relying on its training data alone.

Google has announced a new Android operating system update that deeply integrates Gemini AI into core system functions, including accessibility features, motion sickness mitigation, and user interface customization. Some features bring Android closer to parity with Apple's iPhone capabilities; others leverage Gemini's language understanding to solve accessibility challenges specific to Android's open architecture.

Google has launched Pics, a new creative suite integrated into Google Workspace designed to simplify AI-powered image editing and generation for business users. The tool lets non-designers create and edit professional-quality imagery without requiring Photoshop expertise or external tools. It sits within the Workspace ecosystem, making it accessible to enterprise teams already using Gmail, Docs, and Sheets.

John Deere has introduced an AI assistant called "JD" that is currently being tested with farmers. The tool answers questions about equipment settings, fuel efficiency, crop management best practices, and operational trends by analyzing each farmer's own field, machine, and operational data. Rather than offering generic farming advice, JD customizes recommendations based on historical performance and conditions specific to each operation.

NVIDIA is officially launching DLSS 5 (Deep Learning Super Sampling version 5) on September 3rd, an AI-powered graphics upscaling technology that will be available on RTX 50-series desktop GPUs. The technology uses generative AI to reconstruct lower-resolution game frames into higher-resolution output, effectively allowing games to run at higher visual quality while using less computational power.

OpenAI has announced Astra, its first AI model to reach what the company calls "critical" cybersecurity capability—meaning it can autonomously identify and exploit computer system vulnerabilities. Rather than quietly releasing the model, OpenAI is providing early access to select partners so they can strengthen their defenses before wider availability.