Claude Code Artifacts Released - a new way to share your work

#1 📝 OpenAI News

New usage analytics and updated spend controls for enterprises - June 18, 2026 — OpenAI added credit usage analytics to the Global Admin Console that combine ChatGPT and Codex credit consumption into a single view, enabling admins to track usage and credit trends over time, identify top users and emerging patterns, break down spend by user/product/model, and pull the same data via the unified Cost API. They also updated spend controls so admins can set a default workspace limit, configure group limits and individual overrides, and allow employees to view their credit usage and request additional credits with context; these features are available now to ChatGPT Enterprise admins and users.

#2 📝 Anthropic Engineering

How we contain Claude across products - Anthropic has deployed Claude across claude.ai, Claude Code, and Claude Cowork while containing blast radius via environment controls (sandboxes, VMs, filesystem/egress limits), model-layer controls (system prompts, classifiers, probes, training), and restricting external-content/tool access, noting Claude Mythos Preview was judged too risky to ship in April 2026. Telemetry showed users approved ~93% of permission prompts, Claude Code auto mode blocks roughly 83% of overeager behaviors before execution, and Claude Opus 4.7 resists prompt-injection with about 0.1% success on single attempts and ~5–6% after 100 adaptive attempts.

#3 📝 OpenAI News

Improving health intelligence in ChatGPT - GPT-5.5 Instant (released May 2026) is available to all free ChatGPT users and, on aggregate health evaluations including HealthBench and HealthBench Professional, performs comparably to OpenAI’s frontier Thinking models. OpenAI reports more than 230 million weekly health users, physician-led reviews of 3,500 responses rated GPT-5.5 Instant higher than physician-written responses across accuracy, communication and completeness, and production monitoring showed a 71% reduction in health responses with at least one flagged factuality issue over the past two months.

#4 📝 Claude Code Blog

Steering Claude Code: CLAUDE.md files, skills, hooks, rules, subagents and more - Announces new steering controls for Claude Code, including CLAUDE.md files and mechanisms like skills, hooks, rules, and subagents to guide agent behavior. These additions make it easier to customize and control code-generation workflows and agent orchestration.

#5 📝 Claude Code Blog

Claude Code now supports artifacts - Announces support for artifacts in Claude Code, enabling storage and management of generated outputs and related files within Claude Code workflows. This feature helps teams preserve, share, and reuse code outputs and assets created by agents.

Also covered by: @Claude, @Claude, @Boris Cherny, @Thariq

#6 𝕏

OpenAI released research on methods to train AI models to maintain safe, beneficial behavior across unfamiliar, high-stakes tasks and under sustained pressure.

#7 𝕏

Anthropic tested Claude programming a robodog in Project Fetch Phase 2 using Opus 4.7, achieving ~20× speed over last year’s best human+Opus 4.1 team, though the robodog still failed to fetch a beach ball.

#8 𝕏

Google DeepMind launched the AI Control Roadmap, a structured framework for building and managing advanced AI at Google by proactively anticipating and mitigating unintended behaviors.

#9 𝕏

Jeff Dean highlights a new IEEE Micro paper tracing Google’s TPU supercomputers from v2 to Ironwood over five generations—detailing shifts like air-to-water cooling, 2D-to-3D torus interconnects, and a ~30× boost in TFLOPS/Watt as workloads pivot to transformers.

#10 in

Peter Yang built a concise SKILL.md template for Claude Code/Codex that specifies the AI advisor’s role, context files, advice style, and learning‐update rules. He then uses personal context docs (plan.md, learnings.md, eval.

#11 𝕏

Philipp Schmid shows how to build a real-time translation app using Google’s Gemini Live API for streaming transcription/translation, LiveKit for audio routing, and deploy it on Cloud Run—with full example code on GitHub.

#12 𝕏

Cursor launched `/automate`, a skill that converts plain-language task descriptions into fully configured automations. It automatically sets up triggers, instructions, and tool integrations.

#13 𝕏

Cognition’s Devin tool reasons across your entire codebase to uncover deep business-logic flaws—like an unauthenticated password-reset endpoint—by tracing request flows through auth layers that pattern-matching scanners miss.

#14 𝕏

There's An AI For That launched a Hyperagent integration with Google’s Gemini Omni API that lets agents ingest raw video and automatically output enhanced footage, slashing the complexity, cost, and speed barriers of traditional video workflows.

#15 𝕏

There’s An AI For That introduced Guild’s new Insights Dashboard, giving teams a real-time, unified view of agent usage and costs.

#16 𝕏

xAI added its Grok models to Databricks Agent Bricks, letting enterprises bring SpaceXAI’s latest AI to their data for powering capable AI agents.

#17 𝕏

LlamaIndex 🦙 launched LiteParse v2.1, an entirely LLM-free markdown parser that delivers the fastest output and outperforms all other model-free competitors across three benchmark datasets.

#18 𝕏

Sebastian Raschka highlights GLM-5.2, the latest open-weight model built on GLM-5/5.1’s MLA and DSA architectures. It adds an IndexShare mechanism to reuse sparse-attention indices every four layers, slashing 1M-token inference costs.

#19 𝕏

Aravind Srinivas launched Brain, a self-improving context graph that aggregates all your sessions, connectors, and files and refreshes itself overnight. It feeds continuous state into every task on Computer and is now available to all Perplexity Max subscribers.

#20 ▶️

wtf is Loop Engineer & how to setup for real

AI Jason

#21 ▶️

How I Turned Codex Into My AI Life Coach in 13 Minutes (5-Step Tutorial)

Peter Yang

#22 📝 Ampcode Chronicle

A Faster Librarian - Librarian now runs on GPT-5.5 (no reasoning) with websocket mode and an updated system prompt, making it ~2.9–3x faster and 43% cheaper while maintaining similar quality (F1 0.47 → 0.48). In internal tests latency dropped from 237s to 81s (websocket ≈1.3x, model ≈2.2x), average cost fell from $1.21 to $0.69, it issues ~8 parallel tool calls per turn (up from ~3) and finishes searches in ~5 turns instead of ~15 (example: 2m/$1.08 → 40s/$0.47).

#23 𝕏

Santiago Checkmarx surveyed 2,350 engineers, finding nearly half of production code is AI-generated and that companies relying more on AI ship vulnerabilities at 3.4× the rate of those using it less.

#24 𝕏

Cursor launched an emoji trigger in Slack for Cursor Automations—react to any message with an emoji to instantly kick off a run.

#25 𝕏

Philipp Schmid built a realtime translation app with Gemini Live Translate, Next.js, LiveKit, and Cloud Run that streams host audio via WebRTC to LiveKit and pipes PCM frames for on-the-fly translation.