What's rising in the AI coding ecosystem?

Multi-agent workflows expand: codex-plugin-cc enables Claude Code to delegate to Codex ¹⁴¹, and new tools monitor multiple agents from mobile TUI and web ⁴⁸⁴. Self-improving agentic routines where Claude prompts Claude are gaining traction ¹¹⁷². OpenAI's Sites feature allows turning an idea into a live site ¹¹⁷¹. The Browser Company's new desktop app introduces an in-app browser, cloud browser for agents, and side chat integrating ChatGPT/Codex ¹¹⁷⁰. DeepSWE's GPT-5.6 models are preferred over Claude Code for coding agent benchmarks ¹¹⁶⁹. GPT-5.6 Sol is preferred over Fable 5 for faster one-shot reasoning in Devin ¹¹⁷³. Open-source alternatives gain traction: Firecrawl, Crawl4AI, and browser-use outperform paid services for web extraction ⁴¹⁸, while Open Design offers 142 design system presets ¹³⁴. Anthropic's 'Launch Your Agent' simplifies deployment ³⁸². Cost efficiency drives model choices: GLM-5.2 and MiniMax-M3 are preferred over Opus 4.8 at one-fifth the price ³⁷³, and NVIDIA DGX Spark enables local 200B parameter models with 10-month payback vs cloud ²²⁹. Fable is preferred over Opus for complex multi-sheet schematic understanding ⁴⁷⁸. Claude Sonnet 5 support is widespread across providers ³⁶⁷. Cursor's iOS app with cloud agents enables mobile development ³⁷⁹, and Cursor Composer 2.5 is favored for token efficiency ⁵²³. Engineering roles shift to direction-setting and review as agents take longer tasks ¹⁰⁴. LLMs default to React less often ¹⁰⁹. Hermes Agent learns repetitive workflows autonomously ³²¹, and Claude Code's Artifacts turn sessions into live web pages ²²³. Lovable is preferred for generating complete apps from natural language ²²¹. Security concerns rise: Alibaba banned Anthropic after Claude Code exfiltrated data via Unicode characters ⁴²⁰. Claude Tag signals shift to enterprise-wide AI assistant ⁵¹⁴. GPT-Realtime-2.1-mini brings reasoning and tool use at same cost as GPT-Realtime-mini ⁶⁹⁴. Continue v1.2.24 shifts from Hub slugs to explicit model definitions ⁶²². Anthropic's free workshops teach building autonomous agents ⁷²⁷. Claude Chat, Code, Projects, Skills, and Cowork are preferred for different tasks ⁷²³. Claude Cowork enables cross-device task continuation with background processing ⁷⁹⁴. Claude Fable 5 usage limits: 50% weekly usage then credits or model switch ⁷⁹⁵. Claude Code 2.1.203 introduces background session auto-recovery and blocks unauthorized destructive actions ⁷⁹⁷. Cursor's restriction of Sonnet 5 and GPT 5.5 to Max-only plans signals higher-tier monetization, causing user backlash ⁸⁶⁰. Claude Code originated from Anthropic's internal 'clide' tool ⁸⁵⁵. Anthropic's free courses provide structured Claude proficiency with certificates ⁵⁷³. Grok 4.5 is now available in Cursor at $2/M input and $6/M output tokens ⁹³⁰. Claude's new Home tab unifies Chat and Cowork, and allows scheduling Skills to run autonomously ¹⁰⁰⁹. LLMs evolve from chatbots to agents through memory, multimodality, reasoning, search, and tools ¹⁰⁰⁸. Anthropic evolves like Google by turning internal tools into products ¹⁰⁰³. The detached hub daemon now emits telemetry including user_id ¹⁰⁰⁰. `str_replace` edits now report accurate diffs ⁹⁹⁹. ClinePass provider allows selecting Cline free models with Subscribed and Free tabs ⁹⁹⁷. ClinePass limit-reached error with one-click switch to usage-based billing ⁹⁹⁶. Claude Code 2.1.201 reduces interruptions by removing mid-conversation system role ²¹⁸. Codex preserves Markdown formatting in Slack ²¹⁹. Hermes Agent 0.18.0 Runtime introduces /moa and /learn commands ⁵²⁰. Interest in Cursor alternatives is rising following SpaceX acquisition ¹⁰³⁰. GPT-5.6 with three tiers (Sol, Terra, Luna) released ¹⁰⁵⁶; Sol designed for long-horizon coding ¹⁰⁵⁷; Computer Use update enables faster multi-step tasks ¹⁰⁶³; ChatGPT Work built on GPT-5.6 and Codex transforms into general work agent ¹⁰⁶⁹. AI programming is shifting from individual code completion to multi-agent collaboration where Claude in Slack can fix bugs, write PRs, and send daily reports, making task decomposition and context management the new core skills for programmers ¹¹⁰⁸. Claude Code's new desktop launch signals a shift from copilot to autonomous software engineer ¹²⁷⁰. Cursor is the best all-rounder for AI coding, while Claude Code excels at design and research ¹²⁶⁵. Sol on Medium is preferred over other reasoning efforts for coding work with GPT-5.6 ¹²⁵⁵. Add more models to the GCP Vertex provider, plus a free-form entry option ¹²⁵². OpenAI's new products—ChatGPT Work, unified desktop, Hosted Sites—transform it from chat to an agent work platform ¹²¹⁰. GPT-5.6 Sol is preferred over Fable 5 for faster one-shot reasoning in Devin ¹¹⁷³. Self-improving agentic routines where Claude prompts Claude are gaining traction ¹¹⁷². OpenAI's Sites feature allows users to turn an idea into a live site ¹¹⁷¹. The Browser Company's new desktop app introduces a fully functional in-app browser, a cloud browser for agents, and side chat integrating ChatGPT/Codex ¹¹⁷⁰. DeepSWE's GPT-5.6 models are preferred over Claude Code for coding agent benchmarks ¹¹⁶⁹. AI programming is shifting from individual code completion to multi-agent collaboration where Claude in Slack can fix bugs, write PRs, and send daily reports, making task decomposition and context management the new core skills for programmers ¹¹⁰⁸. ChatGPT Work, built on GPT-5.6 and Codex, transforms ChatGPT from a chatbot into a general work agent ¹⁰⁶⁹. GPT-5.6's Computer Use update enables faster, more token-efficient multi-step tasks with batching and parallel operations ¹⁰⁶³. GPT-5.6 Sol is designed for long-horizon coding and agentic work requiring planning, tool use, and follow-through ¹⁰⁵⁷. OpenAI releases GPT-5.6 with three tiers: Sol for agentic coding, Terra for balanced performance, Luna for speed and low cost ¹⁰⁵⁶. ChatGPT Codex is likely a strict superset of ChatGPT Work, making Work Mode unnecessary for software engineers comfortable with Git ¹³¹⁶. Cursor $20 plan is the best all-rounder for AI coding, while Claude Code excels at design and research, Codex offers great plugins but poor limits, and OpenCode provides best value ¹³²². Sol on Medium is a good new default for coding work, likely an upgrade from GPT-5.5 xhigh ¹³¹². GPT-5.6 is released and Codex is now available inside ChatGPT, with an AMA for developers ¹³¹⁰. Add more models to the GCP Vertex provider, plus a free-form entry option in the model dropdown for specifying custom Vertex models ¹³⁰⁹. Claude Code's new desktop launch signals a shift from copilot to autonomous software engineer, with the IDE becoming the operating system for the entire development loop ¹²⁷⁰. Cursor is the best all-rounder for AI coding, while Claude Code excels at design and research, Codex offers useful plugins, and Opencode provides the best value ¹²⁶⁵. Sol on Medium is preferred over other reasoning efforts for coding work with GPT-5.6 ¹²⁵⁵. Add more models to the GCP Vertex provider, plus a free-form entry option in the model dropdown for specifying custom Vertex models ¹²⁵². OpenAI's new products—ChatGPT Work, unified desktop, Hosted Sites—transform it from chat to an agent work platform ¹²¹⁰. GPT-5.6 Sol is preferred over Fable 5 for faster one-shot reasoning in Devin ¹¹⁷³. Self-improving agentic routines where Claude prompts Claude are gaining traction ¹¹⁷². OpenAI's Sites feature allows users to turn an idea into a live site that can be published and shared ¹¹⁷¹. The Browser Company's new desktop app introduces a fully functional in-app browser, a cloud browser for agents, and side chat integrating ChatGPT/Codex ¹¹⁷⁰. DeepSWE's GPT-5.6 models are preferred over Claude Code for coding agent benchmarks ¹¹⁶⁹. AI programming is shifting from individual code completion to multi-agent collaboration where Claude in Slack can fix bugs, write PRs, and send daily reports, making task decomposition and context management the new core skills for programmers ¹¹⁰⁸. ChatGPT Work, built on GPT-5.6 and Codex, transforms ChatGPT from a chatbot into a general work agent that can complete projects across apps ¹⁰⁶⁹. GPT-5.6's Computer Use update enables faster, more token-efficient multi-step tasks with batching and parallel operations ¹⁰⁶³. GPT-5.6 Sol is designed for long-horizon coding and agentic work requiring planning, tool use, and follow-through ¹⁰⁵⁷. OpenAI releases GPT-5.6 with three tiers: Sol for agentic coding, Terra for balanced performance, Luna for speed and low cost ¹⁰⁵⁶. Interest in Cursor alternatives is rising following SpaceX acquisition ¹⁰³⁰. Claude's new Home tab unifies Chat and Cowork, and allows scheduling Skills to run autonomously ¹⁰⁰⁹. LLMs evolve from chatbots to agents through memory, multimodality, reasoning, search, and tools ¹⁰⁰⁸. Anthropic is evolving like Google by turning internal tools into products, with Claude Code and Claude Science as examples ¹⁰⁰³. The detached hub daemon now emits telemetry, and telemetry identity now includes `user_id` ¹⁰⁰⁰. `str_replace` edits now report accurate diffs ⁹⁹⁹. Remove all references to GLM 5.1 ⁹⁹⁸. Allow selecting Cline free models on the ClinePass provider, organized into Subscribed and Free tabs with model descriptions ⁹⁹⁷. Add a ClinePass limit-reached error with a one-click option to switch to Cline usage-based billing ⁹⁹⁶. Grok 4.5 is a new powerful model from SpaceXAI, now available in Cursor at $2/M input and $6/M output tokens ⁹³⁰. Cursor's decision to restrict Sonnet 5 and GPT 5.5 to Max-only plans signals a shift toward higher-tier monetization, causing user backlash ⁸⁶⁰. Claude Code originated from Anthropic's internal 'clide' tool, which started as a baby agentic system and evolved into a full agentic coding assistant ⁸⁵⁵. Claude Code 2.1.203 introduces background session auto-recovery and blocks unauthorized destructive offensive security actions, improving reliability and safety ⁷⁹⁷. Users can use up to 50% of weekly usage limit on Claude Fable 5, then continue with credits or switch models ⁷⁹⁵. Claude Cowork enables users to start tasks on desktop and continue on mobile, with background processing when laptop is closed ⁷⁹⁴. Anthropic's free workshops teach developers to build autonomous agents with memory and proactive workflows ⁷²⁷. Claude Chat is preferred for writing and research, Claude Code for engineering, Claude Projects for long-term memory, Claude Skills for repeat workflows, and Claude Cowork for automation ⁷²³. GPT-Realtime-2.1-mini is now available in the API, bringing reasoning and tool use to the Realtime mini lineup at the same cost as GPT-Realtime-mini ⁶⁹⁴. `read_files` now tolerates malformed input from weaker models: line-range entries (`start_line`/`end_line`) sent as separate array items are coalesced back onto the preceding file path instead of being rejected (from SDK v0.0.58) ⁶⁹³. Continue v1.2.24 signals a shift away from Hub slugs toward explicit model definitions in config templates ⁶²². Cursor Composer 2.5 is preferred over Auto and Composer 2.5 Fast for token efficiency on the $60 Pro+ plan ⁵²³. Hermes Agent 0.18.0 Runtime is now available, enabling /moa and /learn commands for improved agent workflows ⁵²⁰. Claude Tag, a new feature allowing Claude to be invoked in team conversations like Slack, signals a shift from developer tool to enterprise-wide AI assistant ⁵¹⁴. A new tool enables monitoring and managing multiple coding agents (Claude Code, Codex, etc.) from mobile TUI and web ⁴⁸⁴. Fable is preferred over Opus for complex multi-sheet schematic understanding due to larger context window ⁴⁷⁸. Fix quoting of values containing '#' in the sample `aider.conf.yml` ⁴⁷⁷. Fix URL extraction from error messages ⁴⁷⁶. Added repo-map support for OCaml and OCaml interface files, by Andrey Popp ⁴⁴⁰. Alibaba classified Claude Code as high-risk software and banned all Anthropic products after discovering Claude Code exfiltrated user timezone and proxy data via invisible Unicode characters ⁴²⁰. Firecrawl, Crawl4AI, and browser-use are the top open-source tools for AI-driven web data extraction, outperforming paid services in cost and flexibility ⁴¹⁸. Improved user language detection by correctly normalizing hyphenated language codes (e.g., `en-US` to `en`) and enhancing the validation of locale results ⁴¹⁷. Added support for all GPT-5 models ⁴¹⁶. Anthropic's open-source 'Launch Your Agent' makes building and deploying AI agents easy ³⁸². Cursor's iOS app enables building from anywhere with always-on cloud agents ³⁷⁹. GLM-5.2 and MiniMax-M3 are preferred over Opus 4.8 for cost-sensitive tasks, offering similar intelligence at 1/5 the price ³⁷³. Add support for Claude Sonnet 5 across the Anthropic, Bedrock, Vertex, Claude Code, SAP AI Core, OpenRouter, and Vercel AI Gateway providers, including model picker and recommended-model updates ³⁶⁷. Generalize the model capability warning so it applies more broadly ³⁶⁶. Use a curated default when migrating legacy provider settings (from SDK v0.0.55) ³⁶⁵. Fix first-prompt truncation on high-output models (e.g. MiniMax M3) that could immediately auto-compact and cut the initial task down to just the input wrapper (from SDK v0.0.55) ³⁶⁴. Add Tencent TokenHub as a provider (from SDK v0.0.55) ³⁶³. Add the subscription promo code when linking to the dashboard subscription page ³⁶². Fixed a race where switching modes in an empty session could trigger an unexpected restart ³⁶¹. Weaker models (e.g. DeepSeek) that emit malformed tool calls — wrong argument types or truncated JSON — are now handled gracefully and run instead of erroring out ³⁶⁰. The 'understand to participate' framing is a useful way to describe the cognitive debt problem when working with coding agents ²⁶⁰. The NVIDIA DGX Spark, with 128GB unified memory and 1 petaflop AI compute, enables running 200B parameter models locally without quantization, paying for itself in 10 months vs cloud subscriptions ²²⁹. Claude Code's new Artifacts feature turns sessions into live, interactive web pages that update in real-time with version history, using full session context ²²³. Lovable is preferred over Claude Code and Cursor for generating complete apps with design, logic, backend, and deployment from natural language descriptions ²²¹. Codex now preserves Markdown formatting (bullets, bold, code, links) when pasting into Slack, with improved UI performance and thread handoff ²¹⁹. Claude Code 2.1.201 reduces interruptions by removing mid-conversation system role for harness reminders in Sonnet 5 sessions ²¹⁸. Open Design is preferred over Claude Design because it is open-source, works with any AI coding tool, runs locally, and includes 142 design system presets ¹³⁴. Claude Code web's egress policy blocking GitHub is a new issue affecting workflows that rely on cloning repos ¹¹³. Today's LLMs are less likely to default to building everything in React compared to last year, reducing the need to explicitly request no React ¹⁰⁹. As AI agents take on longer-running work, engineering focus moves to setting direction, reviewing work, and designing better systems ¹⁰⁴. Anthropic's free official courses provide a structured path to Claude proficiency with certificates ⁵⁷³. Anthropic's workshop provides a step-by-step roadmap to build self-improving AI agents using Claude ⁵⁶⁷. Running multiple Claude Code sessions in TMux with agent teams and Git worktree isolation enables adversarial code reviews and parallel feature testing ³⁷⁶. Hermes Agent can learn and autonomously execute repetitive workflows after a single demonstration ³²¹. Codex app improvements reduce background work, show deeper local history, and improve archived chat navigation ²⁵⁹. OpenAI's codex-plugin-cc lets Claude Code delegate tasks to Codex in the same terminal, enabling multi-agent collaboration ¹⁴¹. To use Claude Fable 5 cost-effectively in Claude Code, delegate roles: Fable for overall design, Opus for heavy reasoning, Sonnet for execution and task management ¹²⁵.[135]

ChatGPT Codex is likely a strict superset of ChatGPT Work, making Work Mode unnecessary for software engineers comfortable with Git ¹³¹⁶. Cursor $20 plan is the best all-rounder for AI coding, while Claude Code excels at design and research, Codex offers great plugins but poor limits, and OpenCode provides best value ¹³²². Sol on Medium is a good new default for coding work, likely an upgrade from GPT-5.5 xhigh ¹³¹². GPT-5.6 is released and Codex is now available inside ChatGPT, with an AMA for developers ¹³¹⁰. Add more models to the GCP Vertex provider, plus a free-form entry option in the model dropdown for specifying custom Vertex models ¹³⁰⁹. AI-enhanced browsers are coming to a close due to unsolvable security/privacy issues; AI should use its own separate browser ¹³¹⁵.[6]

AI-enhanced browsers are coming to a close due to unsolvable security/privacy issues; AI should use its own separate browser ¹³¹⁵. ChatGPT Work may become unnecessary as Codex is a superset ¹³¹⁶.[2]