AI Snake Oil
·
1 month ago
Google claimed its AI agents built an operating system for $916 using Gemini 3.5 Flash and Antigravity 2.0, but the company's disclosure was misleading about key details including that the "single prompt" was thousands of lines long and the system used specialized scaffolding with subagents and anti-cheating measures. The cost reported was $916.92 with a total token budget of 2.6 billion tokens, but Google did not release the lengthy prompt, agent logs, or code, making independent verification impossible. The experiment demonstrates that AI agents can work autonomously on long-horizon tasks but highlights the need for rigorous methodological standards in open-world evaluations rather than relying on vendor-conducted press releases.
Mistral AI
·
1 month ago
Mistral released Mistral Medium 3.5, a 128-billion-parameter model that powers remote coding agents capable of running asynchronously in the cloud while developers work on other tasks. The model scores 77.6% on SWE-Bench Verified for coding performance and costs $1.5 per million input tokens and $7.5 per million output tokens via API. Developers can now launch multi-step coding and productivity tasks from Mistral Vibe's CLI or Le Chat that execute remotely in parallel, with results delivered as pull requests or completed work rather than requiring oversight of each step.
Mistral AI
·
1 month ago
Mistral released Connectors in Studio, enabling developers to access built-in connectors and custom Model Context Protocol (MCP) integrations via API/SDK for use with all model and agent calls. Connectors are now centrally registered and reusable across Mistral apps including LeChat and AI Studio, with programmatic access for creating, modifying, listing, and deleting connectors. Developers can now implement human-in-the-loop approval flows, direct tool calling, and eliminate duplicated integration work across teams by building enterprise AI agents with secure, governed access to systems like CRMs and knowledge bases.
Platformer
·
2 months ago
● 2 sources
Google introduced AI agents that summarize web content directly in email and YouTube, intensifying tensions between publishers and platforms over content aggregation. The feature appears to be part of Google's broader AI integration strategy announced at I/O 2026. Publishers face increased pressure as AI summarization tools reduce traffic incentives for users to visit original sources.
OpenAI Blog
·
2 months ago
Virgin Atlantic used Codex to complete its mobile app redesign and meet a fixed holiday travel deadline. The app achieved near-total unit test coverage and shipped with zero severity-1 defects. The airline was able to accelerate development cycles while maintaining quality standards during a time-sensitive project.
OpenAI Blog
·
2 months ago
OpenAI has been designated a leader in Gartner's 2026 Magic Quadrant for enterprise AI coding agents, with its Codex model recognized for both innovation capabilities and ability to deploy at enterprise scale. The evaluation places OpenAI among a select group of vendors assessed across criteria including technical performance, market presence, and suitability for large organizations. Companies evaluating coding agent solutions can now reference this quadrant positioning when making procurement decisions for their development teams.
