The biggest AI tools news of the past 48 hours is that coding agents are turning into collaborative team workspaces: Anthropic just brought live, shareable Artifacts to Claude Code, while Google pushed Veo 3 to general availability and OpenAI shipped a batch of quality-of-life upgrades to ChatGPT. Below is everything worth knowing, item by item, with sources so you can dig deeper.
Claude Code Turns AI Coding Sessions Into Live Web Pages
Anthropic this week extended its Artifacts feature into Claude Code, giving engineering teams a way to turn an AI coding session into a live web page that colleagues can open, explore, and watch update in real time.
The practical upshot is that coding-agent output is no longer trapped in one developer's terminal. When Claude Code produces a new iteration, the shared page refreshes in place and even holds the viewer's scroll position, so teammates see changes the moment they're published. Every revision is kept in version history, which means a team can roll back to an earlier version if a change goes sideways. Access defaults to private, and the feature is available in beta for Claude Team and Enterprise organizations through the Claude Code CLI and desktop app.
This is a meaningful shift in how agentic coding output gets reviewed. Instead of pasting screenshots into Slack, a team can watch the same artifact evolve live and version it like any other shared document.
Source: DevOps.com
Claude Code Adds Auto Mode on Pro and Better Usage Visibility
The same Claude Code update broadens its agentic auto mode and layers on useful diagnostic tooling. The highlights:
- Auto mode now runs on Pro accounts and supports Sonnet 4.6 alongside Opus, replacing permission prompts with background safety checks.
/usagebreaks down what's driving your plan limits by skill, subagent, plugin, and MCP server./code-reviewreports correctness bugs in your code.- Background sessions now appear in
/resumeand stay alive when pinned.
Together these push Claude Code further toward autonomous, longer-running workflows while finally giving developers clear answers to the question "what is actually eating my plan limits?" The /usage breakdown in particular should help teams budget more predictably as agents take on bigger jobs.
Source: Claude Code release notes
Google Pushes Veo 3 to General Availability — at a Lower Price
Google's Gemini API changelog logged several developer-facing launches, led by Veo 3 and Veo 3 Fast reaching general availability. The GA release comes with lower pricing plus new options for aspect ratios, resolution, and seeding.
That combination matters because production-grade AI video has been expensive and hard to control. Lower pricing plus more granular output controls makes Veo 3 considerably more practical for developers building real products rather than demos.
Google also shipped two cost-conscious infrastructure updates:
- Support for the Embeddings model in the Batch API, which helps cut costs on large-scale embedding workloads.
- The Batch API added to the OpenAI compatibility library, making it easier to run batch queries with existing tooling.
Source: Gemini API changelog
ChatGPT Improves File Handling and Big Pastes
OpenAI shipped a couple of small-but-welcome composer upgrades. You can now upload up to 20 files at once (up from 10), which makes it easier to analyze document sets, compare multiple files, or feed broader context into a single message.
Separately, for ChatGPT Free and Go users, long pastes are now handled differently: paste more than 10,000 characters into the composer and ChatGPT automatically converts the content into an attachment instead of dumping it into the text field. The result is a cleaner composer and less risk of a giant paste silently consuming your entire context window.
Neither is a headline-grabbing feature, but both reduce everyday friction for people doing real document work.
Source: ChatGPT release notes
ChatGPT App Polish and a GPT-5.3 Instant Tweak
OpenAI also rolled out a broader app-experience update across web, iOS, and Android. The notable additions:
- Pronunciation help in 60+ languages
- World Cup updates
- Better chat organization and sharing
- Stronger app permissions
- Faster photo uploads on iOS
- New Record & Replay for Codex on macOS (the standout for developers)
On the model side, OpenAI is rolling out an update to GPT-5.3 Instant that improves follow-up tone and reduces "teaser-style" phrasing in responses — the kind of vague "let me know if you'd like more!" filler that pads answers without adding value.
Source: ReleaseBot — OpenAI/ChatGPT
The Model Release Cadence Stays Relentless
Independent trackers logged fresh model activity over the past 24–48 hours. Price Per Token's changelog of new AI and LLM releases from OpenAI, Anthropic, Google, Meta, Mistral, and others was last updated June 23, 2026 — a reminder that the pace of releases remains effectively daily.
On benchmarks, the LLM Stats Leaderboard currently shows Claude Mythos Preview leading on GPQA Diamond at 94.6%, the most discriminating reasoning benchmark at the frontier. For teams choosing a model for hard reasoning tasks, a new frontier preview topping the hardest benchmark is a useful signal worth tracking.
| Update | Who | Why it matters |
|---|---|---|
| Live Artifacts in Claude Code | Anthropic | Collaborative, versioned agent output |
Auto mode on Pro + /usage | Anthropic | More autonomy, clearer limits |
| Veo 3 GA, cheaper | Production AI video gets accessible | |
| 20-file uploads, smart pastes | OpenAI | Less everyday friction |
| Claude Mythos Preview tops GPQA | LLM Stats | New reasoning frontier |
Sources: Price Per Token · LLM Stats
The Takeaway
The through-line this week is maturity: coding agents are becoming team-shareable and self-documenting, AI video is moving from novelty to affordable production, and consumer apps are sanding down rough edges. If you build with these tools, now's a good time to revisit your Claude Code workflow and check whether Veo 3's new pricing changes your video roadmap.
Ready to go deeper?
Browse all AI toolsFrequently Asked Questions
What are live Artifacts in Claude Code?
They turn a Claude Code session into a live web page that teammates can open and watch update in real time. The page refreshes in place when Claude produces a new iteration, keeps full version history for rollbacks, and defaults to private. It's in beta for Claude Team and Enterprise organizations.
Is Google's Veo 3 now available to all developers?
Yes. Google moved Veo 3 and Veo 3 Fast to general availability via the Gemini API, with lower pricing and new options for aspect ratios, resolution, and seeding, making production-grade AI video generation more accessible.
What changed with ChatGPT file uploads and pasting?
ChatGPT now lets you upload up to 20 files at once, up from 10. For Free and Go users, pasting more than 10,000 characters automatically converts the text into an attachment instead of inserting it into the composer, keeping the input cleaner and preserving context window space.