Paul J Lipsky's weekly AI news video packs several genuinely useful releases into 25 minutes. The immediate changes are higher Work and Codex allowances on certain ChatGPT Pro plans, texting a Dot for eligible US users, GPT-6.1 Terra in Work and Codex, and editable files in Gemini. The most striking demos, Gemini 4 Carbon, Block's Dessa voice model and Genie 4, have narrower access than the headlines suggest.
Check your plan before changing your workflow. OpenAI's release notes confirm increased Pro $100/$500 Work and Codex limits and a canceled Pro $200 reduction, not a blanket increase for Plus. Dot texting is a US Pro beta. Gemini file editing is the broadest immediately usable change. Carbon, Dessa and Genie 4 are better treated as previews, not software everyone can open today.
Watch Paul's Full Roundup
Creator: Paul J Lipsky. This article draws on his supplied transcript and on product documentation checked 30 October 2026. His screen demonstrations are attributed to him; we did not independently repeat them. OpenArt Director at 05:29 is a sponsored segment, not an OpenAI product or an independent product review.
OpenAI: More Pro Capacity, a Textable Dot and a Faster Worker
Pro allowances are changing, but not for every subscriber
At 00:38, Paul explains the change to ChatGPT Work and Codex usage. OpenAI's 29 October note says Pro $100 and $500 allowances doubled. It canceled the planned reduction for Pro $200 users who already had the higher allowance; some Pro $200 accounts on the lower allowance were also raised to match. This announcement does not double Plus usage. The practical number is the one in your own Usage screen, because plan, model and task can consume allowances differently.
Text your Dot, with an approval step
Paul's 01:47 demo sends his Dot messages, media and location from his phone. OpenAI documents iMessage and RCS support for eligible US Pro beta users. You link and verify the number in ChatGPT; consequential actions still send you back to ChatGPT for approval. The Dot cannot read the rest of your text conversations simply because it has a messaging address. Readers outside the rollout, including those affected by Dots' geographic access rules, should not plan around SMS-style access yet.
Terra is a Work and Codex model, not a ChatGPT chat replacement
At 06:27, Paul tries GPT-6.1 Terra on computer-based work. He finds it quick for straightforward video cuts but less convincing than Sol for the more involved editing he attempted. OpenAI's launch positions Terra for visual reasoning and computer use, makes it available to eligible Plus, Pro, Enterprise and Edu users in Work and Codex, and lists it in the API as gpt-6.1-terra. The standard API price is $1 per million input and $5 per million output tokens. These are vendor claims and prices, not a promise that Terra wins every editing task.
At 24:46, Paul also notes iCloud Mail and Calendar plugins. OpenAI's release notes list them for all ChatGPT plans on web, desktop and iOS. His later discussion of merging Chat and Work is a future product direction, not a change to the interface on the date of this article.
Carbon, Dessa and Genie 4: Impressive, but Not General Releases
Carbon: read the benchmark and the access rule together
Paul describes Gemini 4 Carbon at 08:31 as a frontier-model leap. Google's announcement reports strong benchmark results and an introductory API price, while its rollout starts with limited early access, including selected cybersecurity defenders. A top score on one evaluation does not establish the best model for every workload, and the pricing is not yet an ordinary self-service checkout for most readers. Paul's examples and Google's published numbers are not this site's benchmark tests.
Dessa: a voice demo with important provenance limits
The unsettling voice examples at 09:36 come from Block's Dessa project. Paul reacts to its realistic, conversational speech and discusses possible local execution. Treat the clips as a company demo: the video does not establish a generally available download, public API, or independently measured impersonation resistance. A voice that sounds convincing also raises consent and verification questions; do not use a clip as proof that a caller is the person they sound like.
Genie 4 is a research preview
At 17:36, Paul shows Google's Genie 4 world demos and separately tries the earlier Genie 3 experience in Project Genie. Google describes Genie 4 as research with limited trusted testers; broader Project Genie access and an API are future plans. The older Project Genie is not proof that Paul had hands-on Genie 4 access.
Claude and Grok Bot: Distinguish a Filing, a Waitlist and Shipped Features
Paul tours Claude's iPhone and iPad changes at 12:09; device-specific layouts and account switching are his on-screen observations, not a guarantee that every mobile version has them. The larger company story is real but easy to overstate: Anthropic says it filed an S-1 for a proposed public offering. A filing is not an IPO completion, and its timing and final price were not set by Paul's 13:58 segment. Claude Duet, mentioned at 14:57, remains a US research-preview waitlist. Anthropic's current product page lists Pro, Max and Ultra plans; eligibility is not immediate access.
Grok Bot's official changelog confirms video creation and Vault improvements, including a Payments tab that still requires approval for purchases. Video creation may be gated by account. The Vault guide says a Bot cannot read the secret itself, but Bots on the same account can use saved entries; it is not a separate secret store for each Bot. At 15:23, Paul discusses Bots calling other people as a coming feature. That is different from a user starting a voice chat with a Bot; do not treat outbound calling as a tested feature here.
Gemini: File Editing Is the Practical Release
Paul's 20:02 demo edits a PDF inside Gemini. Google's 29 October announcement says PDF preview and editing work on web and mobile; other supported file formats, including documents, slides, spreadsheets and code, are web-only for now. It is rolling out in all languages where the Gemini app is available. Start with a disposable file, inspect the result and retain the original before relying on an export.
At 21:44, Paul shows sharing a Gemini avatar. At 22:59, he tests a longer Gemini Notebook recording. Those are useful account-level demonstrations, but availability and recording length can differ by plan, app version and platform; we did not find a current official statement establishing the same controls for every user. The seasonal Snake game at 24:58 is a fun extra, not a workflow change.
What Should You Try First?
| If you need... | Try | Check first |
|---|---|---|
| More agent capacity | Review Work and Codex Usage | Confirm your Pro tier and account allowance; the Plus announcement is different. |
| A quick document revision | Edit a duplicate PDF in Gemini | Review every changed section and the export. |
| A phone-based agent handoff | Text your Dot if eligible | US Pro beta, number verification, approval in ChatGPT. |
| Short clips from an agent | Test Grok Bot video | Feature gate, clip length and rights to uploaded media. |
The pattern is simple: use the features with an actual access path today; keep Carbon, Dessa and Genie 4 on a watchlist until their rollout terms are concrete. Paul supplies a useful tour. The product docs decide what your account can presently do.
Video Chapters
| Time | Topic |
|---|---|
| 00:00 | This week in AI |
| 00:38 | ChatGPT usage limits |
| 01:47 | Text your ChatGPT Dot |
| 05:29 | OpenArt Director (sponsored segment) |
| 06:27 | GPT-6.1 Terra |
| 08:31 | Gemini 4 Carbon |
| 09:36 | Dessa voice model |
| 12:09 | Claude app updates |
| 13:58 | Anthropic IPO filing |
| 14:57 | Claude Duet access |
| 15:23 | Grok Bot calls |
| 16:23 | Grok Bot videos |
| 16:33 | Grok Bot Vault |
| 17:36 | Genie 4 worlds |
| 20:02 | Gemini file editing |
| 21:44 | Gemini avatar sharing |
| 22:59 | Gemini Notebook recording |
| 23:39 | ChatGPT Chat/Work merger discussion |
| 24:46 | ChatGPT iCloud plugins |
| 24:58 | ChatGPT Halloween Snake |
| 25:17 | Final thoughts |
Sources and Scope
Source note, 30 October 2026: We used Paul's video and supplied transcript to identify demos and timing, then checked vendor pages for product status. We did not independently test the models, reproduce Paul's accounts, or verify future-release predictions. OpenArt Director is a sponsored segment.
OpenAI: release notes, Dot texting, Terra. Google: Carbon, Genie 4, Gemini file editing. Other primary sources: Block Dessa, Anthropic S-1 announcement, Duet waitlist, Grok Bot changelog and Vault guide. Block's page could not be independently inspected by our web reader, so Dessa details beyond the project link remain attributed to Paul's demonstration.