Coding

AIAnthropic recommends that AI coding tools like Claude Code test and revise their own output, rather than leaving debugging to developers. The source describes a developer who built a small app with Claude Code and added an AI customer-service bot, but the text provided is only the opening scenario.

AIMiles runs the full reinforcement learning loop on Nvidia Rubin, using SGLang for rollout, Megatron for training, and one container image. On a single 4-GPU tray, Qwen3-30B-A3B's GSM8K reward rises from about 45% to about 95% over 50 rollouts, matching the GB300 curve. The post also reports DeepSeek-V4-Flash end-to-end rollout and training, and Qwen3.5-35B-A3B agentic RL with 64 concurrent mini-SWE-agent sandboxes on SWE-bench Verified, where reward holds near 0.6 and median response length falls about 30%.

AIGitHub makes local sandboxing generally available in Copilot CLI, the Copilot app, and VS Code sessions using Agent Host, limiting agents' access to files, networks, and credentials at no extra cost. The Copilot app now lets users sign in with separate GitHub accounts for the Copilot license and for repositories. Copilot CLI's /model command lists local models from a running Ollama instance alongside cloud models, and VS Code 1.141 adds a side-by-side agent session grid and worktree cleanup.

AIAdo Kukic, an Anthropic account holder, says he builds his own tools to learn, explore, and keep full control over the experience, rather than use Notion, Outline, Auth0, or Traefik. He says Opus 5.5 makes building such apps fun.

AITinker has cut prices up to 70% on long-context prefill and sampling, which now cost the same as short context. The cut lowers the cost of agentic RL rollouts, which spend most of their tokens re-reading growing context, and of evaluating trained models on long inputs. Tinker also added GLM-5.3-Flash and DeepSeek-v4.1-Flash for cost-efficient long-context work.

AIPrime Intellect says it rewrote its Prime Agent coding tool in Rust, using a swarm of more than 2,000 agents over two weeks. The company reports cold start to typing about 13 times faster than the TypeScript version, and memory use over 80% lower after startup. Prime Agent remains open source and adds native Windows support in beta and Homebrew installation.

Why it matters: The post shows how a multi-agent swarm rewrote a coding agent with parity checks, giving a concrete case of agent-driven software engineering with measured results.

AIPrime Intellect says it is turning the workflow behind a recent rewrite into reusable state machines in Prime Agent, letting users run their own agent teams through implementation, review, and verification. The company also says it is accelerating work on capabilities and evals and connecting Prime Agent with cloud agent swarms and hosted training for autonomous research. This release adds native Windows support in beta and Homebrew installation.

AIPrime Intellect says Prime Agent is now open source and that its Rust rewrite is available today. The full blog post covers verification loops, multi-agent orchestration, and performance experiments.

AISherwin Wu, who owns OpenAI's account context here, says he has been tab-accepting about 40-50% of Codex's next-message suggestions after a week of use. OpenAI Devs says composer predictions, which suggest a user's next message from the conversation and their style, are in beta for Pro users.

AIA study by Harvard researchers Fiona Chen and James Stratton, using Jellyfish engineering data from over 700 software firms, finds little evidence that AI coding tools increase software output or reduce employment. The authors report that efficiency gained during coding is absorbed by downstream constraints, mainly longer code review, more pull request revisions, and more reviewer comments.

AIAnthropic releases Claude Code v2.1.296, which adds a code key to the Claude apps gateway's managed.policies[] and an allow_large option to the Read tool for reading large text files in one call. The release also adds autoCompactWindow for subagents and CLAUDE_CODE_WORKFLOW_SUBAGENT_MODEL, and fixes many bugs in hooks, MCP servers, permission checks and self-hosted runners.

AIGitHub adds two Copilot code review admin controls. Organization owners can bill code reviews from members with a Copilot license to the owning organization instead of member quotas, which requires AI Credits paid usage and allows an optional budget. Owners and repository admins can also restrict review requests to users whose Copilot license comes from their organization or enterprise.

AIOpenAI's Codex now suggests the user's next message based on the conversation and how the user talks to it, available in beta for Pro users. Lewis Tunstall of Hugging Face jokes that meat proxies will love the feature.

AIRadixArk says Proximal is using Miles to train coding agents and calls it a flexible, scalable foundation for teams running their own training workloads. Proximal says its training framework is built on Miles, with runs on Modal's on-demand GPU clusters and serverless GPUs for inference. Its sandboxing infrastructure runs on Kubernetes and can handle millions of concurrent rollouts.

AILinus Torvalds says he uses AI to do things he is bad at, such as building a user interface for a guitar pedal project he wrote in C. He says AI is a wonderful tool for beginners, but warns that maintainers are stressed by AI-generated Linux kernel patches and bug reports. Torvalds says AI review tools like Sashiko are now appearing on the Linux Kernel Mailing List, with some subsystem maintainers expecting patches to be reviewed before acceptance.

AIOpenAI's developer account announces composer predictions in Codex, which suggest a user's next message based on the conversation and how the user writes. During the beta, the feature is included at no additional cost for eligible Pro users.

AIAnthropic has granted access to Claude Code Projects for every Pro and Max user on the waitlist. Lydia Hallie asks users for feedback on the feature.

AIAnthropic says a Claude project is an ongoing conversation in which Claude coordinates work and runs each task as its own thread in parallel. Documentation is available at

AIaccording to a post from Kilo. The model has 600B parameters, with 27B active per token, and supports a 1M-token context and vision.

AIOpenAI releases Codex 0.162.1, a bug-fix update for the Codex CLI. It fixes a TUI crash when asynchronous questions contain multiple lines, preserving line breaks and complete hyperlink destinations. It also fixes startup failures caused by mismatches between a running background server's feature settings and CLI defaults.

AIJAN3 CTO Jan Ceuleers says he rebuilt the AQUA wallet from Flutter into React Native, using Claude Code for most of the coding. He started by binding GDK to a React Native app and showing a wallet balance in about half an hour, then ported the Aqua UI component library from Flutter over roughly a year of prior work.

AIAnthropic planted 70 bugs in a 116k-line codebase and tested two approaches over three runs. A single agent found 14, 15 and 27 bugs, while a workflow found 66 bugs in each run.

AIAnthropic says dynamic workflows are powerful but can use many tokens, so users should start with a scoped task and increase complexity gradually. It points to the /claude-api managed-agents-onboard bug-hunter command in Claude Code and to templates at

AIDex Horthy says the share of tasks that can be one-shot without strict process has grown, but alignment, grilling, and planning workflows still matter. He argues that heavy planning on small tasks makes developers feel slower, and predicts tools will add escape hatches so humans or models can decide to ship directly. He adds that as model capabilities improve, the "smart zone" has grown to roughly 200k–400k tokens, and HumanLayer is prototyping research-to-implement and research-to-short-design-to-implement workflows.

AILauren Tan (@poteto) posts a short recommendation to try coding with Grok Bot. Jonathan Wilke (@jonathan_wilke) says he now handles 90% of his coding work through Grok and calls it amazing.