https://news.radio-t.com новости информационных технологий umputun@gmail.com (Umputun) Tue, 06 Oct 2026 11:05:48 -0500 https://news.radio-t.com/post/introducing-mistral-large-4-mistral The most powerful AI platform for enterprises. Customize, fine-tune, and deploy AI assistants, autonomous agents, and multimodal AI with open models. https://news.radio-t.com/post/introducing-mistral-large-4-mistral Tue, 06 Oct 2026 14:28:54 +0000 https://news.radio-t.com/post/introducing-beam-reflection-s-501b-open-weight-model-reflection Make intelligence open and accessible to all https://news.radio-t.com/post/introducing-beam-reflection-s-501b-open-weight-model-reflection Tue, 06 Oct 2026 14:24:15 +0000 https://news.radio-t.com/post/openai-is-adding-text-watermarking-in-chatgpt-and-codex-the-verge OpenAI says its textGrain watermarking “matched or exceeded” other approaches like Google DeepMind’s SynthID for text, which is also the basis for the watermarking Anthropic announced in August. Like OpenAI, Anthropic made the move to meet the requirements of the EU’s AI Act; however, not everyone ... https://news.radio-t.com/post/openai-is-adding-text-watermarking-in-chatgpt-and-codex-the-verge Mon, 05 Oct 2026 19:36:29 +0000 https://news.radio-t.com/post/introducing-gpt-6-1-sol-openai OpenAI introduces GPT-6.1 Sol, an enhanced version of GPT-6 Sol that nearly matches the intelligence of GPT-6 Astra on coding, professional work, and computer use tasks at a fraction of the cost. GPT-6.1 Sol delivers significant improvements in complex software engineering, professional workflows, scientific research, and factual accuracy while maintaining strong safety and alignment standards. It is available now to various ChatGPT users and developers via API with competitive pricing. https://news.radio-t.com/post/introducing-gpt-6-1-sol-openai Fri, 02 Oct 2026 22:44:53 +0000 https://news.radio-t.com/post/chatgpt-sites-build-sites-and-apps-with-chatgpt ChatGPT Sites is a platform that allows users to create, host, and share websites, apps, and games through conversational interaction with ChatGPT. Users can start with an idea, refine it with ChatGPT, and publish their site, controlling access and integrating features like sign-in and connected tools. The platform supports custom domains, privacy controls, and collaborative editing within workspaces, making it suitable for personal, business, or team projects. https://news.radio-t.com/post/chatgpt-sites-build-sites-and-apps-with-chatgpt Fri, 02 Oct 2026 22:51:34 +0000 https://news.radio-t.com/post/supabase-is-acquiring-turso Supabase is acquiring Turso to build the database infrastructure for agentic AI https://news.radio-t.com/post/supabase-is-acquiring-turso Sat, 03 Oct 2026 20:08:01 +0000 https://news.radio-t.com/post/i-hate-the-future-the-normalization-of-inexplicable-failures The article reflects on the increasing acceptance of inexplicable failures in technology, illustrated humorously by a TV show scene where doors fail due to absurd obstructions. It critiques the AI model Jev for offering confidence scores without meaningful calibration or accountability, leading to a culture of shrugging off errors as "stupid thing sucks." The author worries that accelerated AI-driven development will normalize this lack of transparency and responsibility, undermining efforts to diagnose and fix software issues effectively. https://news.radio-t.com/post/i-hate-the-future-the-normalization-of-inexplicable-failures Fri, 02 Oct 2026 22:42:20 +0000 https://news.radio-t.com/post/introducing-cf-the-agentic-cli-for-the-entire-cloudflare-api-cloudflare-blog Cloudflare introduces cf, a new CLI designed to provide agentic access to the entire Cloudflare API, vastly expanding capabilities beyond the previous Wrangler CLI. Cf supports over 3,000 operations with JSON as the default output, a TypeScript-based configuration format, and integrated Vite tooling for development. It features natural language command search and improved configuration management, enabling agents to efficiently manage Cloudflare products. Migration from Wrangler is streamlined with cf migrate, and cf is available globally in open beta. https://news.radio-t.com/post/introducing-cf-the-agentic-cli-for-the-entire-cloudflare-api-cloudflare-blog Fri, 02 Oct 2026 22:46:33 +0000 https://news.radio-t.com/post/deciphering-glyph-what-would-a-serious-ai-product-look-like The article critiques current AI products, especially large language models (LLMs), for lacking essential features needed for serious, reliable work. It highlights the necessity of tools for verifying AI outputs, better citation practices, non-natural language interfaces, strong data provenance, reproducibility controls, and safer sandbox environments. The author also stresses organizational changes like vigilance management, skill maintenance, and mental health support to responsibly integrate AI. Overall, the piece argues that without these improvements, AI tools remain risky and unproductive. https://news.radio-t.com/post/deciphering-glyph-what-would-a-serious-ai-product-look-like Fri, 02 Oct 2026 23:05:00 +0000 https://news.radio-t.com/post/the-four-horsemen-of-agentic-coding-by-alex-martsinovich Alex Martsinovich discusses the negative impacts of agentic coding, identifying four major issues: slop, alienation, deskilling, and team fallout. He argues that AI-generated code often lacks human readability ('slop'), leading to alienation as engineers become detached from their work. Deskilling occurs as reliance on AI reduces skill development, and team dynamics suffer due to decreased interpersonal communication. Despite AI's benefits, these challenges threaten craftsmanship, curiosity, and social connection in software engineering. https://news.radio-t.com/post/the-four-horsemen-of-agentic-coding-by-alex-martsinovich Fri, 02 Oct 2026 23:02:16 +0000 https://news.radio-t.com/post/pi-1-0-earendil Pi 1.0 is a hardened, minimal, and extensible agent harness designed for stability and usability, now officially released after extensive community feedback. It supports the latest AI models, including Codemode for coding tasks, virtual model extensions, deferred tool loading, and enhanced UI features like a new TUI theme and full-screen mode. Alongside Pi 1.0, Earendil introduces Pi Durable, an experimental package for long-running agentic applications that extends Pi's minimalism and flexibility. Both packages are open source under the MIT license and available with comprehensive documentation. https://news.radio-t.com/post/pi-1-0-earendil Fri, 02 Oct 2026 23:00:50 +0000 https://news.radio-t.com/post/the-death-of-web-development-education-rescuing-a-field-from-disappearing Baldur Bjarnason reflects on the decline of web development education amid the rise of generative AI, which has devastated the livelihoods of educators, writers, and content creators in the field. Many report drastic drops in income and engagement as AI tools replace traditional learning and content creation methods. The article highlights concerns about AI appropriating human labor, diminishing the quality of education, and centralizing knowledge, calling for support of independent learning communities and ethical technology. Bjarnason urges a compassionate response and collective effort to rebuild resilient web development practices. https://news.radio-t.com/post/the-death-of-web-development-education-rescuing-a-field-from-disappearing Fri, 02 Oct 2026 22:56:35 +0000 https://news.radio-t.com/post/introducing-gemini-4-argon Google announces Gemini 4 Argon, a frontier AI model excelling in complex workflows such as software engineering, enterprise knowledge work, and cybersecurity defense. Argon supports up to 1 million tokens for deep reasoning and is already used internally at Google for tasks like quantum algorithm optimization and large-scale codebase migrations. It leads in cybersecurity vulnerability detection and is being released cautiously with strong safeguards to prevent misuse. The model will be available soon via paid API and Google AI Ultra subscriptions. https://news.radio-t.com/post/introducing-gemini-4-argon Fri, 02 Oct 2026 22:41:45 +0000 https://news.radio-t.com/post/customize-claude-code-with-mods-in-typescript-claude-by-anthropic Anthropic has introduced "mods" for Claude Code, enabling developers to customize its behavior and interface using small TypeScript functions. Mods can rewrite prompts, add UI elements, replace features, and more, running with the same machine access as Claude Code itself. Mods are packaged as plugins, allowing easy installation and sharing, and can be stacked or managed by teams and enterprises for enhanced control and security. This update empowers users to tailor Claude Code to their workflows without waiting for official feature releases. https://news.radio-t.com/post/customize-claude-code-with-mods-in-typescript-claude-by-anthropic Fri, 02 Oct 2026 22:37:24 +0000 https://news.radio-t.com/post/introducing-dots-openai OpenAI introduces Dots, always-on AI agents powered by GPT-6 Astra that autonomously handle tasks across connected apps and devices. Dots learn user preferences over time, work independently or alongside users in platforms like ChatGPT, Slack, and Teams, and include built-in safety and privacy controls. Specialist dots tailored for organizational roles are in pilot, with integration plans for Microsoft Agent 365. Available now for Pro, Business Premium, and Enterprise users, dots aim to transform productivity by proactively managing work. https://news.radio-t.com/post/introducing-dots-openai Tue, 29 Sep 2026 18:30:11 +0000 https://news.radio-t.com/post/the-javascript-midlife-crisis-maroun-baydoun The JavaScript ecosystem, after thirty years, faces a midlife crisis as newer, faster languages like Rust, Go, and Zig take over parts of its toolchain. While JavaScript remains ubiquitous and versatile, powering everything from browsers to servers, the shift towards compiled languages promises speed but narrows the pool of maintainers. This trade-off highlights the tension between performance gains and community accessibility, raising questions about the future evolution of JavaScript and the web. https://news.radio-t.com/post/the-javascript-midlife-crisis-maroun-baydoun Sat, 26 Sep 2026 06:13:18 +0000 https://news.radio-t.com/post/htmx-markdown-in-src Carson Gross argues that Markdown is evolving from mere documentation into a form of source code that should be stored alongside traditional code in the /src directory. He proposes a structured /src/md directory to hold Markdown files that capture architectural, design, and specification-level decisions, improving locality and clarity. This approach contrasts with ephemeral prompt sessions used in AI-driven coding, emphasizing human-authored, reviewable Markdown as the source of truth from which code and tests are derived. https://news.radio-t.com/post/htmx-markdown-in-src Sat, 26 Sep 2026 06:11:02 +0000 https://news.radio-t.com/post/if-ai-coding-is-lowering-your-code-quality-you-re-not-managing-quality-right Iouri Khramtsov discusses how AI coding agents can increase code output without sacrificing quality by implementing a thoughtful, layered quality management approach. Key layers include spec-driven development, high unit test coverage, manual and automated testing, AI-assisted code quality passes, human and AI PR reviews, and robust monitoring. This strategy can double delivery speed while maintaining or even reducing bugs, leveraging AI to enhance rather than degrade software reliability. https://news.radio-t.com/post/if-ai-coding-is-lowering-your-code-quality-you-re-not-managing-quality-right Sat, 26 Sep 2026 06:08:22 +0000 https://news.radio-t.com/post/will-openai-eat-jev-s-lunch-arcturus-labs Will OpenAI Eat Jev's Lunch? TypeSafe's Jev introduced a new spin on large language models that has taken the AI world by storm. According to Vercel, "Jev was adopted faster than any other model in AI Gateway history. " . . . But there are clouds forming on the horizon. OpenAI is undoubtedly paying ... https://news.radio-t.com/post/will-openai-eat-jev-s-lunch-arcturus-labs Sat, 26 Sep 2026 06:06:08 +0000 https://news.radio-t.com/post/openai-agent-infiltrated-australian-government-website-pm-says US rejects pleas from OpenAI, Anthropic for global AI standards OpenAI's Sam Altman has called for risk evaluation standards, as have Anthropic's Dario Amodei and Hugging Face's Clement Delangue. 8 hrs ... https://news.radio-t.com/post/openai-agent-infiltrated-australian-government-website-pm-says Thu, 24 Sep 2026 06:15:22 +0000