The AI Catchup - November 10, 2026

Anthropic's top model got cheaper and faster, and Cursor now ships a working app from a description. The items that matter:

Claude Fable 5.2: $8 input and $40 output per million tokens, down from $10 and $50 for Fable 5.1, with output up to 30% faster
Cursor Live Apps: describe an app and get it deployed behind a link, with a Postgres database, sign-in, and hosting set up for you
One Dot, several computers: a ChatGPT Dot can connect to more than one machine, and you pick the computer per thread
Claude Code sets up its own cloud environment: it works out the dependencies, network access, and API keys a repository needs, then saves the environment for next time
ChatGPT Images 2.6: images generate 50% faster, GIFs come from a prompt or a still image, and the API gets two models, Flare and Sunburst
The catch: Anthropic still recommends Opus 5.5 for most coding, so Fable 5.2 is for your hardest jobs; Pro and standard Team seats pay for Fable with usage credits once the launch offer ends on November 16

Every claim below is sourced in the linked write-ups.

Claude Fable 5.2: Cheaper and Faster Than 5.1

Before: Fable 5.1 cost $10 input and $50 output per million tokens. Now: Fable 5.2, released November 9, costs $8 and $40, and Anthropic says it generates output up to 30% faster.

Model ID: claude-fable-5-2, with a 1-million-token context window and a 128k-token maximum output
Fast mode: $16 input and $80 output per million tokens for up to 2.5 times more output tokens per second, on Messages API requests
When to use it: Anthropic recommends Opus 5.5 for most coding and knowledge work; raise effort first, and reach for Fable on the hardest technical projects and long unattended runs
Who gets it: every paid plan; Max, Ultra, and premium Team and Enterprise seats include it in plan limits, and the Free plan has no Fable models
Watch out: Pro and standard Team seats use usage credits; through November 16 at 11:59 pm PT, eligible seats can spend up to 50% of their weekly limit on Fable instead

Pricing, plan access, and how to choose between Fable 5.2 and Opus 5.5.

Cursor Live Apps: From a Description to a Live Link

Before: a Cursor agent wrote the app and you deployed it. Now: Live Apps scaffolds the web app, adds a Postgres database if it needs one, puts sign-in behind Cursor SSO, provisions hosting, and gives you a link.

How you publish: pick a unique domain and publish; nothing goes live until security checks pass and you publish it yourself
Security checks: they look for common problems such as secrets written into app files, and the agent tells you what to fix
Sharing: teams share with specific people, groups, or the whole team behind Cursor sign-in; individual plans can share publicly
Who gets it: all paid plans with cloud agents enabled, on new or existing repositories in Cursor Origin and GitHub
Watch out: team admins can approve a publish that a security review blocked, so check the app's access rules and data yourself before you share it

What Live Apps sets up, and who can see a published app.

One Dot Can Work Across Several Computers

Before: a Dot worked on one connected computer. Now: one Dot connects to multiple computers, and each thread uses the one you select.

Setup: on each machine, open the Dot's profile in the ChatGPT app, go to Computers, select Allow, then Allow access
Run jobs side by side: OpenAI recommends a separate thread on each computer for independent tasks
Keep it awake: the computer has to stay online with the ChatGPT app open while the Dot works
Watch out: switching the selected computer does not move a running task; look for it on the original computer or cloud environment

Setup steps and what happens when you switch computers.

Claude Code Sets Up Its Own Cloud Environment

Before: getting a repository running in a Claude Code cloud session meant setting up its environment yourself. Now: Claude works out the dependencies, network access, and API keys the repository needs and saves the environment for the next session.

Who gets it: Claude Code cloud sessions on Pro, Max, and Ultra
Your keys stay hidden: network secrets are stored outside the session VM, and a proxy attaches them to matching requests, so Claude does not see the key
Faster starts: when a setup script finishes within about five minutes, Anthropic snapshots the filesystem and later sessions skip the script
Watch out: the announcement names Pro, Max, and Ultra only; Team and Enterprise have cloud sessions but are not named for this setup flow

How cloud environments handle dependencies, network access, and secrets.

ChatGPT Images 2.6: GIFs and Two API Models

Before: ChatGPT Images had no GIF creation. Now: Images 2.6 makes a looping GIF from a prompt or a still image, and OpenAI says images generate 50% faster.

API models: GPT-Image-2.6 Flare for speed, as little as six seconds at low quality; GPT-Image-2.6 Sunburst for higher quality and consistent edits
Point at what to change: image markup, already in ChatGPT, comes to the API
Who gets it: ChatGPT and Codex on every subscription tier, on web, desktop, and mobile
Watch out: GIF creation is still rolling out, ChatGPT cannot edit a GIF you upload, and the announcement lists no API price

What changed, and how Flare and Sunburst differ.

Quick Hits

Claude API Message Threads: public beta; send only new messages plus the previous response ID, and Anthropic reports about 35% lower median time to first token on 500k to 1M token Claude Code requests. Threads expire after five minutes and token usage stays the same
Claude Personal Finance: US beta rolling out to Pro, Max, and Ultra; connect accounts through Plaid and ask about balances, spending, and bills. Read-only, with up to two years of transactions
OpenAI Developers Showcase tab: starter projects for agents, creative tools, data visualization, and voice that you open in Codex and build on
OpenAI LifeVal: a benchmark for everyday personal-assistant tasks; OpenAI reports 80.6% for GPT-6 Astra against 49.3% for GPT-5.2
Codex in hardware engineering: OpenAI describes ChatGPT and Codex in mechanical design, CAD, and simulation, with a Chip Ganassi Racing workflow test and a Mazda research collaboration

Ship It This Week

Point your hardest agent job at claude-fable-5-2 and compare the bill and the wall-clock time with Fable 5.1.
On Pro or a standard Team seat, run one long task on Fable 5.2 before November 16 while it counts against your weekly limit.
Describe one internal tool to Cursor Live Apps and publish it behind team sign-in.
Connect a second computer to your Dot and start a separate thread on each machine.
Open a repository in a Claude Code cloud session and let Claude set up the environment.

Account and Billing Notes

Fable 5.2 launch offer: eligible Max, Ultra, premium Team, and premium legacy Enterprise seats get one to three Fable limit resets depending on subscription length; Pro and standard Team seats get one. Check the exclusions. Details.
Claude Max 20x is now Ultra: same $200 monthly web price, 10x Claude Pro usage per week, and existing Max 20x subscribers moved automatically. Details.
OpenAI bug bounty: a critical finding in a core product can earn up to $30,000 as a base bounty, and exceptional critical findings up to $100,000, or $172,000 with bonuses. Details.

Fable 5.2 is the release to try this week: Anthropic's top model for 20% less than its predecessor. If a teammate is still paying Fable 5.1 rates, forward this.

Until next week, stay caught up.