From San Francisco
1 |
The Editorial |
Morning, humans.
Today’s three picks turn on who keeps the wheel.
Anthropic opened Thursday at $111.50 and closed at $97.17, 17% above its $83 offer price and about 13% below the open. The founders’ 50.1% voting stake had a quieter day.
Apple is building a Mac agent while its macOS 27.3 beta gives ChatGPT users 14 setup steps, up from four. Rivals have some clicking ahead.
Canva’s one generative AI toggle arrives after complaints dating to 2023. Finding it takes three menus, and the switch has exceptions, which is quite a service history for an off button.
Stay curious,
Marcus Schuler
2 |
The Big Story |
Anthropic closed its Nov. 19 Nasdaq debut at $97.17 a share, 17% above its IPO price.
ANT finished about 13% below its opening price. More than 846 million shares changed hands Thursday, against 522 million on SpaceX’s opening day.
The seven co-founders hold a combined 50.1% of voting control through a new share class received this fall. The Long-Term Benefit Trust directly elects three of seven board members at the IPO.
Why This Matters:
- Public investors buying into Anthropic have limited power over management under the voting structure established for the IPO.
- Future investment decisions may favor longer objectives over near-term results, a possibility the prospectus explicitly warns could hurt the share price.
3 |
Also Today |
Apple’s macOS 27.3 beta requires ChatGPT users to complete 14 steps to enable third-party computer use, up from four.
Apple job listings describe an autonomous agent that operates apps on a user’s behalf. An introduction is eyed for as soon as the end of 2027. The added permission steps mean more setup for users choosing a rival’s agent.
4 |
The Outside Read |
Dean Ball’s essay “The Silent Consensus,” published on OpenAI’s site, challenges the belief that human control will inevitably pass to AI.
He calls for agents whose permissions are “intelligible and limited” and whose authority users can withdraw. Closing with his baby son, Ball writes: “I do not want him to have a nicer car with a plusher passenger seat. I want him to have the wheel.”
5 |
The One Number |
6 |
Today's Headlines |
- Commerce Department approved more than 879,000 AI chip sales to Chinese companies out of 3.4 million requested from December to May, some to firms the Pentagon lists as “Chinese Military Companies.”
- White House named Kirsten Davies national cyber director Nov. 19, replacing Sean Cairncross, whom two White House officials said chief of staff Stephen Miller pushed out.
- Google added computer use to Gemini for macOS Nov. 19 for U.S. subscribers on AI Plus, Pro and Ultra plans.
- Sona8 turns employee AI interviews into workflow maps agents can use, reporting more than 10,000 interviews across more than 20 organizations.
- Mistral fell 10 points to 75 in Implicator’s Nov. 19 LLM Meter, while Claude scores 92.
7 |
The 5-Minute Skill |
Apple’s macOS 27.3 beta takes ChatGPT users through 14 steps before enabling third-party computer use.
Your raw input:
The prompt:
Why this works: The list makes missing instructions visible before you grant access. Dean Ball asks for agents “whose authority users can withdraw.”
What to use: Use an AI assistant to read the pasted material before enabling computer use in Gemini, ChatGPT, Claude or a plugin.
8 |
What To Watch Next |
Mon 11/23 |
Finance: Zoom reports third-quarter fiscal 2027 results after the market close and holds its investor webinar at 2 p.m. PT. |
Mon 11/23 |
Politics: The European Parliament opens its four-day plenary session in Strasbourg, with the agenda and votes published by Parliament. |
Tue 11/24 |
AI: Big Data Conference Europe and the co-located AI Summit Europe begin with workshops in Vilnius before three conference days. |
Wed 11/25 |
Tech: WeAreDevelopers opens its two-day Bengaluru conference on building and operating software in the age of AI. |
Fri 11/27 |
Fairs: DreamHack Stockholm opens at Stockholmsmassan for three days of gaming, esports, creators and an expo with more than 100 booths. |
9 |
The Rausschmeisser* |
Canva now lets free and paid users disable generative AI with one toggle. The Nov. 19 change leaves background remover and Magic Grab enabled.
Our take: Canva spent three years bolting AI onto its buttons while users complained. October brought Canva Craft, its own model. November brings a setting to make much of that work disappear. The company has found a use for all those features: giving customers something to turn off.
Relief sits under Profile, then Settings, then Permission. Background remover and Magic Grab stay on because, Canva says, they “make everyday tasks faster without generating anything new.” Even the off switch requires Canva to explain which AI counts. Three menus down, the customer gets an off button with a footnote.
FREE · ABOUT FIVE MINUTES
IMPLICATOR