AI - POLPROG Learning

AI

Practical know-how on frontend, AI tools and software development.

Practical know-how

Latest articles

  • AI
  • 9 min read

AI Coding Agents in 2026: How They Are Changing Developers’ Work

In 2026, AI coding tools are increasingly moving beyond autocomplete and chat. Codex, Claude Code, GitHub Copilot cloud agent, Google Antigravity, and Gemini CLI can inspect repositories, edit multiple files, run commands and tests, and work on a task for extended periods. That changes the developer’s role: less work is about manually typing every change, while more is about defining goals, constraints, acceptance criteria, and verifying the result.

  • AI
  • 9 min read

GPT-6 Astra is here: benchmarks, pricing, 1.05M context, API changes, and comparison with GPT-5.6 Sol and Claude Fable 5.1

OpenAI officially announced GPT-6 Astra on September 3, 2026. The rollout is staged: the model first reached a limited set of organizations, with OpenAI saying access would expand over the following days to ChatGPT Plus, Pro, Business, and Enterprise, as well as the API, Microsoft Azure, and Amazon Bedrock. Astra provides a 1,050,000-token context window, up to 128,000 output tokens, and Standard API pricing of $10/MTok input and $50/MTok output. This article separates OpenAI-reported results from independent testing and methodology caveats.

  • AI
  • 5 min read

Grok 4.7 has not been officially released: what is actually confirmed as of September 4, 2026

As of September 4, 2026, there is no official Grok 4.7 announcement in SpaceXAI news, the model documentation, or API release notes. The latest confirmed flagship remains Grok 4.6, released on August 12, 2026. Rather than assigning invented specifications, pricing, or benchmarks to Grok 4.7, this article separates confirmed facts from speculation and documents what is officially available today.

  • AI
  • 7 min read

MCP, Model Context Protocol: what it is and how it works in practice

Model Context Protocol, or MCP, is an open standard for connecting AI applications to tools and data sources. Instead of building a separate integration for every model and service, an MCP server describes its capabilities in a standardized form that compatible clients can discover and invoke. By 2026, MCP has moved well beyond its original release: the current specification uses a stateless HTTP core, formal extensions, stronger authorization rules, and mechanisms for long-running work.

  • AI
  • 14 min read

Claude Fable 5.1 is here: benchmarks, pricing, 1M context, and comparison with Fable 5, Opus 5, and GPT-5.6 Sol

On September 1, 2026, Anthropic officially released Claude Fable 5.1 and the restricted-access Claude Mythos 5.1. Fable 5.1 keeps Fable 5’s base input and output pricing, cuts prompt-cache read pricing by 75%, improves agentic and research benchmark results, and introduces meaningful API breaking changes. This article separates confirmed specifications, Anthropic-reported benchmark results, methodology caveats, and practical migration consequences.

  • AI
  • 18 min read

DeepSeek Harness vs Claude Code vs Codex vs Muse Code: the open-source AI agent framework in 2026

Verified analysis of DeepSeek Harness: Everything is a Plugin architecture, Cordis, model providers, subagents, Web UI, security, MIT licence, developer-preview status and comparison with Claude Code, OpenAI Codex and Meta Muse Code.

  • AI
  • 23 min read

GLM-5.3 and GLM-5.3-Flash: benchmarks, cyber capabilities, Ox Alpha, open weights and comparison with Kimi K3, Hy4, Claude and GPT-5.6 Sol

Verified analysis of GLM-5.3 and GLM-5.3-Flash: architecture, 1M context, 753B/40B and 320B/18B, multimodality, Ox Alpha, coding and cyber benchmarks, licenses, API pricing, self-hosting and comparisons with Kimi K3, Hy4, Claude and GPT-5.6 Sol.

  • AI
  • 16 min read

GPT-5.4 leaves Codex on August 31: migration to GPT-5.6 Terra and Luna, pricing, limits and a safe transition checklist

Verified guide to the August 31, 2026 retirement of GPT-5.4 and GPT-5.4 mini from Codex when signed in with ChatGPT. Migration to GPT-5.6 Terra and Luna, model differences, API and Codex pricing, limits, reasoning, Enterprise configuration, automations and deployment checklist.

  • AI
  • 16 min read

Tencent Hy4 preview: 770B parameters, 49B active and a 1M context. Benchmarks, pricing and comparison with Kimi K3, GLM-5.3 and Claude Opus 5

Verified analysis of Tencent Hy4 preview: 770B/49B MoE architecture, 1M context, Gated DSA, IndexCache, MTP, Apache 2.0, benchmarks, internal blind test, API pricing, self-hosting, FP8, vLLM/SGLang and preview limitations.