• Everyday AI
  • Posts
  • Ep 802: ChatGPT’s Task Comeback, Claude’s Design upgrade, Codex Copies your workflow and 7 other Fresh AI features you’ll Want to use Today

Ep 802: ChatGPT’s Task Comeback, Claude’s Design upgrade, Codex Copies your workflow and 7 other Fresh AI features you’ll Want to use Today

Claude Code brings Artifacts, OpenAI makes key political hire, GPT-5.6 could drop next week and more

Author

Everyday AI
June 19, 2026

Sup y’all 👋

With our special 800th episode dropping this past Wednesday, we didn’t get into our usual ‘Working Wednesday’ show, where we do a deep hands-on show on a recent AI release.

Here’s some recent AI updates catching my eye:

ChatGPT Tasks — Rolled out to all paid users and allows you to have scheduled agents run at any time.

Claude Design — In its June updates, Claude rolled out some great features like better import/export and a visual editor.

Codex Record and Replay — OpenAI released this banger Codex feature that just watches you work then replicates it.

GitHub Copilot Desktop App — Now generally available, this is Microsoft’s new dev desktop platform.

What should we cover for next week's AI Working Wednesday?

🗳️ Vote to see LIVE results 🗳️

Login or Subscribe to participate in polls.

✌️

Jordan

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: From ChatGPT tasks to Claude Design, 7 new AI features just landed that can save you time at work. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Adobe’s new AI tools to Creative Cloud, Unreal Engine 6 is adding Claude and Gemini integrations, Claude’s big MCP moves for enterprise and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Claude Code brings Artifacts, OpenAI makes key political hire, GPT-5.6 could drop next week and more. Read on for Byte Sized News.

💪 Leverage AI: Here’s the AI updates you mighta missed and how to use them today. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Google Gemini co-lead Noam Shazeer leaves Google for OpenAI, Anthropic says access to Fable 5 could return in days, OpenAI is bringing scheduled tasks back to ChatGPT and more. Check it here!

Ep 802: ChatGPT’s Task Comeback, Claude’s Design upgrade, Codex Copies your workflow and 7 other Fresh AI features you’ll Want to use Today

ChatGPT tasks are back, Jack. ✅

While we were collectively ping-ponging the Anthropic vs. U.S. government saga, the big tech AI players rolled out a TON of fresh AI features that are available today.

↳ Claude Design got a big upgrade

↳ Google Vids got some serious AI sparkle

↳ And there's a new Open Weights model king

Also on the pod today:

• ChatGPT Tasks: Pulse is dead, Tasks are back 🔄 
• Claude Design WYSIWYG editor is a game changer ✏️
• Google Vids avatars can now walk and talk 🎬

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight –  Darkmoon runs the full offensive campaign and delivers validated, evidence-backed findings, MeshPilot is an agentic development environment where you can collaborate directly with autonomous CLI agents, Foglamp helps you Ship AI agents you can actually see

Google Ask Ad Manager — Google is rolling out Ask Ad Manager, a Gemini-powered AI that helps publishers troubleshoot issues, build reports, and navigate Ad Manager with just a prompt.

Adobe Firefly Upgrades — Adobe is giving Firefly a big boost, rolling out AI tools like brand kit generation and video creation to Premiere, Illustrator, and more.

Unreal Engine 6 — Unreal Engine 6 is bringing built-in AI tools like Claude and Gemini to speed up game creation and cut tedious tasks.

Claude Artifacts — Claude Code now lets you turn your coding sessions into live, shareable web pages that update automatically as you work.

DeepSWE — Datacurve updated its popular DeepSWE benchmkark, with Fable 5 coming in 3 percentage points ahead of GPT-5.5.

Perplexity Updates — Perplexity just dropped Brain for Max users, letting its AI actually learn from what worked and what didn’t.

Claude MCP Connectors — Admins can now set up Claude’s MCP connectors for everyone at once with Okta, so users get instant access with zero setup.

AI Toilet — A smart toilet from China drives itself to users, promising easier access and hands-free cleaning.

Anthropic Project Fetch — AI is now speeding through robotics tasks up to 20 times faster than humans, but still fumbles with the trickiest moves.

AI Art Museum — Refik Anadol and Google opened Dataland, the first AI art museum, where data transforms into sensory digital art.

1. OpenAI’s rumored GPT-5.6 launch could arrive Tuesday with Mini and Pro variants ⚡

OpenAI is reportedly preparing to expand its GPT-5 lineup next week with GPT-5.6, potentially launching the standard, Mini, and Pro models around the same time.


The update is said to target Anthropic’s top models with a larger context window, stronger long-range coding skills, faster Codex responses, and possible token price cuts.

2. U.S. Gov calls for approval for Anthropic’s Fable access under export rules ⚖️

According to Bloomberg, the Trump administration has ordered Anthropic to get US approval before foreign nationals can use its Fable 5 and Mythos 5 models, a major and legally murky expansion of export control powers.


The move matters now because it shifts AI oversight from voluntary safety reviews to direct government intervention, raising the possibility that simply using a cloud-based AI model could be treated like receiving sensitive technology.

3. Codex brings Record & Replay workflow capture to Mac 🖥️

Codex just released Record & Replay on macOS, letting users demonstrate a repeatable task once and turn it into a reusable skill for future runs. The feature is currently unavailable in the European Economic Area, the United Kingdom, and Switzerland, and it also requires Computer Use to be available and enabled.

In plain terms, this is Codex moving from “tell me what to do” toward “watch me do it, then handle it next time,” with the biggest payoff for repetitive workflows that have clear steps and outcomes.

4. OpenAI says GPT-5.5 Instant is now much better at health answers 🫀

OpenAI announced that GPT-5.5 Instant has made a major jump in health-related performance, especially in spotting urgent situations, asking better follow-up questions, and explaining medical uncertainty more clearly.

The company says physician-led evaluations found the model performed close to its top Thinking models and was rated higher than physician-written answers in a large comparison of health conversations.

5. Former White House AI adviser Dean Ball joins OpenAI 🏛️

Former White House AI adviser Dean Ball has confirmed he is joining OpenAI to lead a small team focused on frontier AI policy, a timely hire as Washington’s fight over powerful AI systems gets more political and more consequential.


Ball has continued to have the White House’s ear while becoming a leading critic of the Trump administration’s feud with Anthropic, including the Pentagon’s move to label the company a supply-chain risk and new export restrictions on Anthropic’s Fable model.

6. Claude Code adds live, shareable work artifacts 📊

Anthropic announced that Claude Code can now turn an active coding session into a live, interactive artifacts that teammates can open, follow, and revisit in an update mirroring OpenAI’s recent Codex Sites success.

The new beta feature uses the session’s code, connected tools, and conversation context to create pages for things like investigations, PR walkthroughs, dashboards, and release checklists, with updates publishing to the same link.

You don’t have hours each day to try the latest AI features.

That’s why on Fridays, we launched our ‘Friday Features’ series, where we give you the rundown on new AI updates you can use today.

No waitlists. No coming soon.

Just quality-of-life upgrades you can benefit from right now. 

Let’s get into it.

1. ChatGPT Tasks Replace Pulse for All Paid Users ⚡

OpenAI just killed Pulse and replaced it with Scheduled Tasks, now rolling out to all paid plans.

The unlock isn't just scheduling. It's monitoring: tell ChatGPT to watch a competitor's launch page and ping you the moment something meaningful changes, no check time required.

A new dedicated Scheduled page in the sidebar lets you view, pause, edit, and delete every active task in one place.

Works on web, iOS, and Android. Ten active tasks max for now.

Try This

Connect Gmail, Calendar, and Drive to ChatGPT, then set a 6 AM task that triages your inbox before your first meeting. One five-minute setup and you're done.

2. Claude Design's June Update Adds WYSIWYG Editing and Code Sync 🔥

Anthropic just dropped a major Claude Design update and this one WAY changes how it fits into a real workflow.

You can now click any element on the canvas and edit it directly, and your design system imports from GitHub so Claude Design and Claude Code share the same component library.

That round trip from prototype to production with no respeccing in between was the obvious missing piece that kept Claude Design from fitting into serious workflows.

Export now goes straight to PDF, PowerPoint, Canva, Gamma, Lovable, Replit, and more. Available in beta on Claude Pro, Max, Team, and Enterprise plans, included with your existing subscription.

Try This

Prototype your next deck or asset in Claude Design, then export directly to Canva or PowerPoint. If it skips two rounds of revisions, make it your default.

3. Google Vids Gets Directable AI Avatars Powered by Veo 3.1 🎬

Google just updated Google Vids with Gemini 3.1 Flash text-to-speech and Veo 3.1, and the avatar experience is now meaningfully different.

The big unlock is direction: type a prompt telling your avatar to pick up a coffee cup or cross the frame, and it responds in the generated video.

Avatar presets jumped from 23 to 53, spanning photorealistic, 3D cartoon, and graphic novel styles, and per-generation duration limits are gone entirely.

Free for anyone with a US Google account, up to 10 Veo generations a month.

Try This

Take an internal announcement that normally eats a full production day and rebuild it in Google Vids with a directed avatar. If it cuts your time in half, that's your new process.

4. OpenRouter Fusion Now Runs Multiple Models in Parallel 🧪

OpenRouter just fully graduated Fusion, its multi-model blending feature, from public experiment into a fully integrated feature inside the OpenRouter API, available to anyone with an account.

Send one prompt, it fans out to multiple models simultaneously, a judge synthesizes the outputs, and you get one coherent answer.

Budget preset runs Gemini 3 Flash, Kimi K2.6, and DeepSeek V4 Pro in parallel at a fraction of what a single top-tier frontier model call would cost your team per query.

For research, complex analysis, and any query where a wrong output could cost your team real time or real money, this is worth building into your regular workflow. Try it at openrouter.ai/fusion.

Try This

Run your three gnarliest high-stakes prompts through Fusion's quality preset and compare against your usual single-model output. If it catches what your default missed, you've found a smarter workflow for anything that actually matters.

5. Claude Code Artifacts Create Live, Shareable Dashboards 📊

Anthropic just shipped Claude Code Artifacts, turning active session work into live web pages teammates can actually watch in real time.

PR walkthroughs, incident dashboards, and release checklists all update themselves automatically as the Claude Code session keeps working through the problem.

The key word is LIVE. Every open browser tab refreshes the moment Claude Code updates the artifact, so no one needs a Slack summary to stay current.

Team and Enterprise plan only via Claude Code CLI or desktop app.

Try This

After your next sprint retro or incident postmortem, have Claude Code build an artifact capturing what happened and why. Share the link instead of writing the summary.

6. Z.ai Drops GLM 5.2, the Best Open-Weights Model in the World 🏆

Chinese AI lab Z.ai just released GLM 5.2, a 753 billion parameter open-weights model that topped the Artificial Analysis intelligence index this week, landing a score ahead of Gemini 3.1 Pro on their benchmark.

Here's the enterprise angle: MIT license, self-hostable, fine-tunable, and deployable completely air-gapped with no per-token meter and zero dependency on any single vendor.

Context window jumped from 200K to one million tokens. It also outscored Claude Fable 5 on front-end coding benchmarks specifically.

Weights are on Hugging Face, and the API is live at Z.ai now.

Try This

Spin up GLM 5.2 through Z.ai's API and run your most demanding coding task. If it performs at frontier level at a fraction of your current cost, that budget conversation just got a lot easier to start.

7. Codex Record and Replay Turns Any Workflow Into a Reusable Skill 🤖

OpenAI just launched Record and Replay for Codex, and it's exactly what it sounds like: hit record, do the work, and Codex packages the entire sequence into an automation you can trigger any time.

This kills the single hardest part of enterprise automation: having to write out a complex workflow in prompts and words when demonstrating it once is ten times faster and produces a better result.

Skills are editable after the first run. Works across browser, computer use, and connected plugins inside your existing Codex setup, with all your current projects and app connections already in context.

Mac only for now, not yet available in the EU, UK, or Switzerland.

Try This

Pick one workflow your team repeats at least weekly and record it once in Codex today. If it replays clean, you just turned institutional knowledge into a shareable team skill with zero IT involvement.

Reply

or to participate.