Fresh stories
Agent teams compare Claude, GPT, Kimi, and MiniMax routing against $2k monthly API bills
A 24/7 agent-team writeup routed planning to Claude, implementation to Kimi and MiniMax, and review to GPT, while other sources quantified Codex, Opus 4.7, and Claude Code cost edges. The setup can cut spend and provider dependence, but it also requires tighter specs, verification loops, and more harness maintenance.
Agent builders ship repo-aware MCP layers, local-first ADEs, and exact-edit tools
New releases included Repowise for repo graphs and bug-risk scoring, a policy-aware MCP proxy, the ÄKÄ local-first ADE, datasette-agent-edit 0.1a0, and a Claude workspace picker. The batch adds repo structure, safer execution, and reusable edit primitives around agent development.
VS Code updates `git.addAICoAuthor`: false Copilot trailers force revert
VS Code briefly changed the `git.addAICoAuthor` default from `off` to `all`, then reverted after Copilot trailers appeared on non-AI commits and even when AI features were disabled. The incident matters because commit-history attribution is durable, so consent and runtime checks have to stay aligned with config defaults.
Agent teams compare Claude, GPT, Kimi, and MiniMax routing against $2k monthly API bills
A 24/7 agent-team writeup routed planning to Claude, implementation to Kimi and MiniMax, and review to GPT, while other sources quantified Codex, Opus 4.7, and Claude Code cost edges. The setup can cut spend and provider dependence, but it also requires tighter specs, verification loops, and more harness maintenance.
Coding-agent teams introduce hooks, smoke tests, and stop conditions after production misfires
Practitioners shared model-tiered Claude Code setups, proof-of-done smoke tests, and stop-condition checklists after reports of agents deleting live data and leaving generated code harder to debug. The harness now carries verification, least privilege, and kill switches instead of the prompt.
Agent builders ship repo-aware MCP layers, local-first ADEs, and exact-edit tools
New releases included Repowise for repo graphs and bug-risk scoring, a policy-aware MCP proxy, the ÄKÄ local-first ADE, datasette-agent-edit 0.1a0, and a Claude workspace picker. The batch adds repo structure, safer execution, and reusable edit primitives around agent development.
GitHub Copilot adds Raycast tips to 1.5M PRs before GitHub disables the feature
A PR-description tip feature inserted Copilot and Raycast promo text into more than 1.5 million pull requests, then GitHub turned it off after backlash. The injected copy remains in repository history because PR descriptions cannot be automatically rolled back.
Claude Code leak reports ANTI_DISTILLATION_CC, fake_tools, and a 3,167-line print.ts loop
OpenAI launches gpt-image-2 API with 1024×1024 pricing from $0.006 to $0.211
VS Code updates `git.addAICoAuthor`: false Copilot trailers force revert
Meta AI support lets attackers swap Instagram recovery emails and bypass 2FA
Top storiesthis week
Claude Code, Codex, and Cursor users compare AGENTS.md and CLAUDE.md memory files
Threads across Cursor and Claude communities converged on the same session-amnesia toolkit: AGENTS.md or CLAUDE.md files, scoped plans, scaffold prompts, and tighter workflow boundaries. Teams should duplicate project context across tools for now, since Claude Code, Codex, and Cursor still read different memory surfaces.
Micropython-wasm, TakoVM, and FSM runtimes add WASM and gVisor guards for agents
New open-source runtime projects moved agent control away from unconstrained loops toward WebAssembly sandboxes, gVisor-backed filesystems, deterministic FSMs, and YAML state engines. Teams should use these guards to bound file access, orchestration, and history growth before a model can mutate full system state.
Google releases Gemma 4 QAT: E2B drops to ~1GB and Ollama, SGLang, vLLM add support
Google published Gemma 4 QAT checkpoints and mobile-focused quant formats, cutting Gemma 4 E2B to roughly 1GB of memory. Ollama, SGLang, and vLLM added day-one support, making local deployment more practical on phones, laptops, and low-VRAM GPUs.
OpenAI fixes mistaken ChatGPT suspensions and restores subscriptions and credits
OpenAI said an issue incorrectly suspended some ChatGPT accounts, then began restoring access, subscriptions, and credits. Users who were locked out should check account status and verify service access before resuming work.
Vercel opens Skills API with 600,000 skills for agents and platforms
Vercel made the skills.sh API generally available, exposing more than 600,000 skills as a registry-style service for agents and platforms. The launch gives teams a discoverable capability layer for reuse across agent surfaces.