AI CODING
220 items
219 posts, 1 tool
Step 5 Preview is StepFun's 600B MoE with 27B active and a 1M window. What it scores, what a session costs, and what to check before the October 15 weights.
Sign in with ChatGPT lets coding agents spend your Plus or Pro plan allowance instead of an API key. How the OAuth flow works and the per-app cap to set first.
Antigravity vs Cursor: Antigravity is free with a weekly limit; Cursor starts at $20/month with the widest model menu. Pricing, limits and a decision guide.
CC Switch is a free open-source app that switches API providers for Claude Code and Codex in one click and syncs MCP servers and skills. How to use it safely.
Superpowers for Claude Code: how to install the MIT-licensed skills plugin, what its brainstorm, plan, TDD and review skills do, and when it is worth it.
Codex usage limits and pricing per ChatGPT plan: what Plus, Pro $200 and Pro $500 buy, how resets work, and what to do about 'Selected model is at capacity'.
NVIDIA OpenShell runs autonomous agents in policy-enforced sandboxes: an open-source runtime that moves agent security from prompt rules to infrastructure.
CodeMidas turns existing repository code into executable coding-agent RL tasks, tests, and verifiers, instead of relying on issues, commits, or hand labels.
Handshake's DeepSWE audit found frontier coding agents reasoning about imagined graders. Stop rewarding patches that pass tests but drift from the user's spec.
Agent Retrieval Bench isolates the part of coding-agent work most evals hide: did the agent find the right repository files before it started editing?
Claude Code plugin evals score a plugin against a no-plugin baseline, write JSON and HTML reports and gate CI, making plugins testable agent infrastructure.
Codex CLI 0.154.0 adds experimental worktrees, inline answers and approval hardening. Why durable agent workspace control matters more than the model picker.
Diffs vs whole files for code editing agents: a paper finds full-file generation beating diff edits on Flutter/Dart tasks. The lesson: route by task locality.
Coding agents need better human loops: a position paper argues agent research over-optimizes solo autonomy while developers need to steer and verify agents.
CodeNib treats repository context as compiled infrastructure, so coding agents stop rediscovering the same repo through grep and reads. What it builds.
Don't Paste the AI, Vomit and NoBuzz compared: AI slop tools for a shared link, a local rewrite or a second-model Claude Code skill, and when to skip them.
arrayref 0.3.10 was compromised to pull in a proc-macro1 typosquat whose build script ran a remote binary. What happened and why coding agents walk into it.
Slack Code gives coding agents their own Slack channels with diffs, previews, and an audit log. When it beats Claude Code, Cursor, or OpenCode, and when not.
A new August 2026 paper argues that coding agents improve faster when they compare attempts across tasks and lineages, not just retry one failed trajectory.
GitHub Copilot for JetBrains adds memory across chat sessions, Ollama as a BYOK provider, and enterprise settings for MCP access. What each feature does.

Get Smarter About AI Dev
New tutorials, open-source projects, and deep dives on coding agents - delivered weekly.