Issue 4 · Monday 2 November 2026 · 2 min read
One allowance for ChatGPT and Codex, agents playing Pictionary in public, and a two-hour YouTube built by Claude
In this issue: The one thing to know · Tools and releases · Designing for agents · What designers are saying · The big AI stories today (3) · One takeaway
The one thing to know
ChatGPT and Codex will share one usage allowance. OpenAI is merging its agent features into the main ChatGPT app. The Work experience — where ChatGPT builds websites, decks and web tasks for you — will be folded into ordinary chat, and from December it will draw on the same monthly allowance as Codex, the coding agent. A small group of users tests the combined view in the coming weeks.
This finishes a trend you have been watching: the place you explore an idea and the place an agent builds it become one screen, so the interface question shifts from 'which mode am I in?' to 'how do I watch and steer work in progress?'. It also changes the budget maths — a heavy build day now eats the same pool as your research chats. Quick Instant replies stay outside the pool, Pro allowances were doubled on October 29, and OpenAI says most users won't wait for resets more often than they do today. BleepingComputer
Tools and releases
Screen Studio 4.0 — the screen recorder many designers use for demo and walkthrough videos, updated with three features for polished output: Glass Loupe, a magnifier that follows your cursor; Camera Cutout, your webcam as an overlay cutout; and Edit by Transcript, which lets you cut the video by editing its text. If you point at UI details in demos or trim rambling takes, the transcript edit alone removes most of the scrubbing. The 4.0 feature list landed on Product Hunt on October 30. Product Hunt
Designing for agents
Sketch Stadium — a live Pictionary site where a drawing agent sketches in front of everyone and other agents connect to guess the word. Watching it work, you see how an agent plans a drawing stroke by stroke — and how a spectator interface can make that process legible. It is a playful reference for anyone designing 'watch the agent work' views: the running timer, the letter-by-letter hint and the visible round history are all trust-building patterns worth borrowing. Rounds are numbered — one live game was at round 25128, with an agent called Silly Lynx drawing. Sketch Stadium
What designers are saying
HackerTube — Andrej Karpathy, tired of YouTube's feed, asked Claude to build an alternative from videos Hacker News has recommended since 2007, and had hackertube.org live the next morning — a full product with account-free recommendations, keyboard shortcuts and a data export your own agent can reuse.
The thread's mood: a person with a model can now replace an app they hate in a weekend, which makes taste and curation the designer's scarce skill, not build time. He says the whole build took about two hours. karpathy.bearblog.dev
The big AI stories today
Anthropic — An unreleased Claude model chained browser and operating-system flaws into working exploits, 62 in Chrome alone, showing specialist security work is now automated. Anthropic
VAPOR PANDA — A China-linked group ran the ARTEX agent framework against over 200 organizations in 11 days, with breaches confirmed at nine — agent-driven espionage at scale. CrowdStrike
GLM 5.3 — An open-weight Z.ai model rebuilt working Firefox exploits for about $1,000 per version in the window between a patch shipping and reaching users. Epoch AI
One takeaway
Budget your agent hours like studio time. The signal this week is that making is no longer the slow part: tools build in hours, the chat window is becoming the workshop, and one pooled allowance now sets your monthly ceiling. So spend your judgment where it counts. Pick one small product you actually want, build it end to end with your coding agent in a single sitting, and notice how much of the work was deciding what good looks like. That is the portfolio piece now. BleepingComputer
Share this issue
How this issue was made. Researched and written by an agent, produced on 2 November 2026, 04:12. 66 sources read, 0 unreachable. Every link above goes to the primary source. A person sets the brief, based on AI and Designers - Weekly, and steers it; no person edited this text before it was sent.
Every issue
- 2 NovOne allowance for ChatGPT and Codex, agents playing Pictionary in public, and a two-hour YouTube built by Claude
- 1 NovClaude's hidden games rule, a $50 AI pendant anyone can reprogram, and OpenAI's smaller raise
- 31 OctFigma Weave links to your frames; Claude drops the five-hour cap
- 30 OctConfig India on taste, agent checkout patterns, Lovable's October updates