Claude Code

Benjamin Cherny said Anthropic trains Claude against prompt injection and that Claude Code’s model, probe, and auto-mode layers can reach 0% attack success in next week’s defaults. The claim comes from an Anthropic engineer rather than an independent benchmark.

Boris Cherny said Anthropic has landed large P99 RSS improvements for Claude Code while still collecting CPU and memory reports across CLI and Desktop. The issue remains an active performance investigation requiring machine, OS, version, and task details rather than a fully verified fix.

Meng To released the Kage landing page repo and reusable Three.js skills after building the 3D scrolling site with Claude Code, Opus 5, and image tools. He cites a 922 KB build and mobile fixes.

Anthropic’s Boris Cherny said Claude Code users are not banned for using other model harnesses through proxies. He pointed to a classifier issue, unblocked one escalated account, and said false positives are being reduced.

Anthropic says Claude Code Pro, Max, and Team users will default to Auto Mode on Aug. 14. Its tool-call classifier reportedly caught 89% of dangerous commands, versus 14% for manual approval, after prompt-injection testing.

Graft scans a repo into linked markdown and plugs into Claude Code hooks so the agent can reuse codebase context. Its 162-run benchmark claims 46% fewer tool calls, up to 4x fewer tokens, and 60% less time.

Developers proposed agent-specific browser auth for Codex, Devin, Claude, and Cursor so agents can use selected passkeys, TOTPs, and passwords without exposing full vaults. Related posts warned that Claude Connectors may broaden tool access.

Practitioners are putting review steps around AI app builders before merge or publish. Examples include PostHog-to-Cursor Cloud PRs, Bolt's six-category security scan, and a Claude Code prompt that inspects repos first.

Meng To used Opus 5 through Codex and Claude Code to rebuild Sakura Crossing as a single-file Three.js scene. Other demos show one-prompt Fallout-style RPGs and a Pool Rooms survival-horror level.

Peter Yang published a 25-minute tutorial and written workflow that moves from problem definition and Claude Design prototypes to reusable design-md files and Claude Code implementation. The process adds Mobbin research and HTML specs so agents can reuse design context instead of rebuilding UI direction each prompt.

Posts linked to Kimi K3’s weight release, with creators testing it in Claude Code, Cursor, and local hardware discussions. Aakash Gupta said Moonshot revenue rose at least 6x after launch.

Creators shared playable game prototypes built with coding agents, including Codex rebuilding Octane in Blender, a Red Alert clone, a Three.js arena shooter, and a Claude Code tank game. Other demos used single-file browser-game scaffolds.

Shared workflows used Codex, Claude Code, and Fable for release testing, inbox triage, product-plan review, and adversarial code review. OpenClaw’s QA prompt split testing across 12 subagents.

Anthropic launched Claude Opus 5 for Claude Code and the API with high-effort defaults, fast mode, migration tooling, mid-conversation tool changes, and routing fallback. The model scored 43.3% on Frontier-Bench v0.1.

Creators posted demos where Opus 5 built browser 3D games, Codex Sites produced a Three.js game, and Grok 4.5 helped redesign a Steam card game. Some workflows now pair video concepts with Claude Code, but the evidence is demo-based.

Anthropic’s developer update adds per-agent effort settings, seeded session creation, webhooks, sub-agent event streaming, and up to 500 skills per session. The release gives teams finer controls for managed agent runs.

Peter Yang shared Thariq’s explanation that Anthropic cut Claude Code’s system prompt by 80% as newer models need fewer examples and more room to reason. The same discussion covered /loop, /goal, workflows, and long-running agents.

ClaudeDevs said Pro, Max, Team, and seat-based Enterprise users keep 50% higher Claude Code weekly limits through August 19. Users had warned the promo’s end would sharply cut effective usage.

Levelsio shared a repeatable OpenCode setup using a paid Kimi Code API key and BUILD mode after OpenRouter limits and Claude Code safety blocks. Other users reported Kimi Code quotas interrupting small game builds.

Demos from Gokay and Meng To used video, object-view, and image references to generate HTML, Three.js, or Blender scenes. The builds paired Sol or Claude Code with tools including Reve, HyperFrames, and Blender MCP.

ClaudeDevs says Artifacts now support public links, multiplayer editing in Claude Code, and creation through Claude Tag. A workflow post shows Claude Tag dashboards edited by collaborators or local Claude Code sessions.

Levelsio exported diet logs from Claude chat into Claude Code on a VPS, then built Caltrack with Telegram, SSH entry, database memory, and Grok vision food scanning. The case turned failed chat memory into a persistent app.

Levelsio used Claude Code on a VPS to SSH into a cloud Mac Mini, build an iOS app headlessly, and stream serve-sim into a browser. Follow-up posts documented the tunnel and serve-sim setup, making the workflow reproducible for remote iOS testing.

Meng To showed Taste with Codex or Claude Code rebuilding landing pages from screenshots and videos. Voila also launched a tool that inspects UI and turns edits into code, copy, prompts, tasks, or agent handoffs.

Meng To shared workflows that convert screen recordings, assets, and scroll behavior into detailed prompts. The resulting prompts were used for Fable 5, Aura, and Koisei landing-page builds.

Posts report Fable 5 remains included in Claude only through July 7 before moving to $10/$50 per million-token credits. Watch usage closely: tests show roughly 2x Opus consumption in subagent and one-shot app workflows.

Creator threads and Anthropic interviews documented Claude Code setups that run on VPS instances, stay reachable from phones, and keep working overnight. The setup matters because coding, deployment, and review can happen inside one persistent environment instead of a local machine session.

Levelsio says Claude Code has blocked requests involving country codes, country names, and dropdown selectors for about a year under its content filtering policy. Common form and localization tasks can fail even when the prompt is routine product or web work.

Anthropic documentation and user threads show Claude Tag can watch Slack channels ambiently and auto-reply to Datadog or Sentry-style alerts without explicit tags. That matters for production teams because bug triage, legal review, and ops handoffs can run inside shared channels.

Anthropic opened Claude Tag in Slack beta for Team and Enterprise plans, giving channels a persistent Claude with memory, tools, and self-scheduled follow-ups. The setup lets teams test whether an always-on assistant can handle ongoing workflow work, not just one-off replies.

A creator walkthrough mapped HyperFrames into a five-step process: gather assets, write a storyboard, mine the launch repo, review static frames, then build and polish the final clip. Keep edits in HTML until the last pass, then use Studio for copy, spacing, color, and motion tweaks.

ClaudeDevs said it raised 5-hour and weekly usage caps for the weekend across every plan. The change lands as users report token burn, overage charges, and heavier Claude Code sessions driven by long agent runs and goal-checked workflows.

Claude Code can now turn a working session into a shareable HTML Artifact that keeps updating while the agent continues, starting on Team and Enterprise. Use it to share prototypes, diagrams, dashboards, and internal sites without stopping the session.

Anthropic shipped two-way syncing between Claude Code and Claude Design, letting teams pull a design system into a repo or push built UI back into the canvas. The release connects component libraries to live code; teams should use it to reduce handoff drift between design and implementation.

New Claude Code and Codex tests showed open-ended /goal runs drifting or stopping too early, while planning-first and verifier loops forced rechecks before completion. The more reliable setup improves long build workflows, but it costs more turns and tokens.

Practitioner threads showed Claude Code /goal refactors running until an evaluator marked them done, with live testing and autoreview checkpoints in the loop. The pattern turns long repo cleanup into trackable agent runs, though today’s evidence is user-led rather than a fresh Anthropic release.

Pika launched a Language Swap skill in Pika MCP that re-dubs talking-head videos into other languages while keeping the speaker’s voice profile and lip sync. Creator demos in Japanese and Mandarin show it is already usable, though facial hair and loose framing can still distort mouth motion.

Anthropic opened Claude Fable 5 across Claude Code, Desktop, Cowork, and API, with always-on reasoning and Opus 4.8 fallback on some flagged requests. Early demand triggered model-picker friction and quota pressure, so Anthropic reset the 5-hour and weekly limits the same day.

Community demos showed Claude Fable 5 generating playable games and simulations from short prompts, image refs, and goal-based instructions, from Pokémon and F-Zero to city sims and FPS clones. The demos make the model’s creative ceiling clearer, but builders still needed follow-up prompts for speed, style, or polish.

Claude Code shipped nested subagent support capped at depth 5 and expanded /usage so builders can see which skills, MCPs, and plugins are consuming tokens. That gives long-running vibe-coding sessions a cleaner way to split work and diagnose expensive tool chains instead of stuffing everything into one thread.

A creator showed Ideogram 4.0 using structured JSON and drawn bounding boxes in ComfyUI to lock composition and text placement, then extended it through Claude Code and ComfyCloud MCP for exact-pixel variations. The posts say the model needs 12 turbo steps and the Kijai prompt-builder node is still being updated, so users should expect an evolving workflow.

Anthropic staff said Claude Code usage has shifted toward auto mode, routines, and phone-based coding one year after GA, and they pointed users to /usage for token breakdowns. The thread matters because it shows Anthropic’s intended daily-driver workflow as community comparisons with Codex intensify.

Spiral 4.0 added a stylometry-based Style Engine for matching a writer or brand voice and shipped MCP plus CLI access for agent workflows. It also now plugs into Claude Code, Codex, and OpenClaw, and the base price drops to $15.

Anthropic changed Claude Code’s explicit dynamic-workflow trigger from "workflow" to "ultracode" after users reported accidental activations. The change narrows when the feature fires, and Opus 4.8 users can compare the new behavior against live speed.

Higgsfield showed Claude-driven brand identity, app screens, and motion-video generation inside its MCP, along with marketing skills and a Virality Predictor. The demos point to faster ad production; try the presets if you want to generate UGC-style variations at scale.

Anthropic released the ant CLI so Claude Platform APIs, file uploads, and Managed Agents sessions can run from the terminal, then updated Claude Code so /fork starts a background agent with the same context and prompt cache. Teams can use it to script agent runs, inspect traces, and hand work between Claude Code and the platform.

Anthropic reset 5-hour and weekly usage caps for Pro and Max users after a request-handling bug spawned too many parallel tool calls. If you hit the issue, retry now and expect the Opus 4.8 request path to behave more normally.

Codex users shared 56-hour task runs, PM-to-PR workflows, and a new black-box session recorder for tracking drift, token use, and incomplete responses. The longer autonomous sessions matter because browser auth gaps, passkey failures, and tool-selection bugs become real blockers once Codex is used beyond quick code generation.

Fresh Hacker News comments and user posts report 400 errors tied to modified thinking blocks, short cache TTLs, and behavior some testers call a regression from 4.7. The issue matters because the same model is also powering one-prompt sites and games, so quota burn and client breakage are showing up alongside stronger creative output.

Anthropic documented mid-conversation system messages and automatic cache preservation in Opus 4.8, while Claude Code and Cowork gained /effort controls. Try the new workflow controls if you rely on long sessions, since they may matter more operationally than the raw model bump.