Tag
aronprins announces updates to Claude Loop and Codex Loop, adding parallel wave support for faster autonomous coding workflows using Claude Code.
This article introduces the shift from traditional prompt engineering to 'recursive engineering', where developers no longer manually prompt AI but instead let AI automatically execute tasks through loops. It breaks down the core components of recursive systems, the debt brought by automation, and provides two practical methods.
An AI agent using a 30-billion-parameter model leverages headless screenshot loops to autonomously complete a raytraced first-person shooter demo written in C, showcasing advanced autonomous coding and debugging capabilities.
An Anthropic engineer explains how to use Claude Code in a loop to autonomously write code and manage PRs, allowing developers to step away from the keyboard entirely.
Factory announces Factory 2.0, evolving from coding agents to an end-to-end, agent-native software factory that enables organization-wide autonomous software development with model independence and sovereign intelligence.
Autonomous coding is evolving from better prompting to better control systems, with engineers wrapping agents in goals, evaluators, and loops.
Claude Fable 5, the first model in Anthropic's Mythos class, is now generally available in GitHub Copilot. It is designed for long-horizon, autonomous coding and knowledge-work tasks, with improved efficiency over previous Opus-tier models.
Discusses optimal placement of human review in autonomous AI coding agent workflows, considering trade-offs between automation and safety, particularly for risky systems like auth, payments, and database migrations.
Cursor's head of AI engineering shares a 14-minute free resource on how to build autonomous coding agents that ship production code, revealing architectural insights that a $1,000 vibe-coding course won't teach.
gnhf is an open-source tool that lets AI agents automatically loop through code modifications, generate Git commits, handle errors, and record progress while you sleep. It supports multiple AI models (e.g., Claude Code, Codex) and multi-agent parallel work.
A developer introduces Lauren, an open-source tool that makes agent loops dynamically editable mid-run, allowing users to add, refine, or merge tasks without stopping the loop.
Anthropic engineer shared development experience of long-duration AI agents, including multi-role division and independent evaluators, enabling AI to automatically generate complete applications within 3-5 hours, with a 12x improvement in continuous operation capability.
This post provides best practices for writing effective /goal statements for AI coding agents like Claude Code, Codex, and Hermes, emphasizing that goals should describe observable end states that an evaluator can verify.
Explains how Claude Code's /goal command enables autonomous multi-turn task completion with model-verified exit conditions, dramatically reducing the need for manual 'continue' prompts during large refactors or feature implementations.
Xiaomi has open-sourced its MiMo V2.5 Pro model, a 1.02T parameter MoE model designed for autonomous coding tasks. The article details a real-world test showing high efficiency with low API costs due to high cache hit rates.
Devin, the autonomous coding agent from Cognition, has grown from $1M to $445M ARR in months, doubling usage every eight weeks and serving major enterprises including the US Army and Goldman Sachs. Cognition is raising at a $25B valuation and recently acquired Windsurf, with Google having paid $2.4B to license Windsurf's founders.
The author built a benchmark harness to evaluate local LLMs for autonomous Go code generation, focusing on log parser generation for SIEM pipelines, and published results comparing quality vs. speed.
Radiq is a new product intelligence platform designed for the era of autonomous coding.
Reddit user reports Qwen 3.6-27B shows unusually proactive agent behavior, autonomously building, testing and fixing code without prompting.
At I/O 2026, Google released the agent-first desktop platform Antigravity 2.0, integrating Gemini 3.5 Flash, which can autonomously drive 93 sub-agents to build a complete operating system from scratch in 12 hours, greatly compressing the time for complex engineering tasks.