Daily ·
Gemini 4 Carbon tops MathArena, Tesla Cybercab pre-orders open
Gemini 4 Carbon takes the MathArena math crown at 15x the cost, Anthropic resets Claude Opus 5.5 prompting, Tesla opens $299-a-month Cybercab pre-orders, and SpaceX catches Starship's ship for the first time.
Ten signal stories from the last twenty-four hours, from a math benchmark that changed hands to the first catch of a Starship upper stage.
1. Gemini 4 Carbon tops MathArena — at fifteen times the cost
Google's Gemini 4 Carbon is the first non-OpenAI model in a while to lead MathArena, knocking GPT-6.1 Sol off the top spot, with Anthropic's Claude Fable 5.2 second. The catch is price: at the reasoning settings used in testing, Carbon costs roughly 15 times as much as GPT-6.1 Sol, and Fable 5.2 about 40 times. Carbon's most common failure mode is exceeding the time limit — in one case it launched a script that ran for more than eleven hours without checking its progress. The Decoder · Nov 14
2. Anthropic: drop "think hard" for Claude Opus 5.5
Anthropic has made Claude Opus 5.5 the default model on Pro, Max, Ultra and Team plans, and published a guide arguing that habits carried over from older models now hold it back. Top of the list: cut phrases like "think hard" or "think carefully", because the model always reasons on its own and scales depth to the task and the selected effort level. Anthropic's tests found responses without those phrases started faster with no measurable quality drop. It also recommends starting at Medium effort — the Opus 5.5 default — which in internal tests matched the previous Opus 5 on High. The Decoder · Nov 14
3. Tesla opens Cybercab pre-orders for personal use
Tesla told "select customers" they can now pre-order the autonomous Cybercab for personal use. The order page revealed a $29,990 price tag plus $299 a month to operate the vehicle, with estimated delivery in December 2026 — but buyers "must accept delivery" in one of a handful of Texas cities, and the car only operates inside certain geofenced areas. Elon Musk said the Cybercab will be available nationwide as regulatory approval arrives, and that the monthly fee is "too high": it should be closer to the ~$115 a month Tesla charges for Full Self-Driving and connectivity on other models. TechCrunch · Nov 14
4. SpaceX catches a Starship upper stage for the first time
Starship's spacecraft returned from a 22-hour voyage in orbit and descended into the grasp of the launch tower's mechanical arms at Starbase, Texas — the first catch of the ship rather than the booster. SpaceX called it a step essential to the decade-long effort to build a fully reusable rocket. CNA · Nov 14
5. Helen Toner: we still can't trust AI CEOs
Helen Toner, executive director of Georgetown's Center for Security and Emerging Technology and one of the four OpenAI board members who voted to fire Sam Altman, argued on The Verge's podcast that the AI industry cannot be relied on to police itself. Her case: the companies have every incentive to downplay risk, so oversight has to come from outside the profit motive. The Verge · Nov 14
6. Amazon widens Alexa for Shopping to fifteen markets
Alexa for Shopping is now fully available in 13 countries, with a beta in two more, and the text-Alexa feature has opened to all users. Amazon launched the shopping assistant in May; the expansion turns the assistant into a checkout surface across fifteen markets. PYMNTS · Nov 14
7. The AI-fueled ransomware boom
Researchers tracking ransomware say claimed attacks posted to leak sites hit nearly 3,000 over the summer — a 20 to 40 percent increase since spring by some estimates. Fully automated ransomware hasn't arrived yet, but analysts argue that relatively unsophisticated crews are using AI tools to run attacks well above their weight. Wired · Nov 14
8. AT&T: nearly 500 agentic workflows in production
AT&T says it has close to 500 agentic workflows running in production and plans to triple that number next year. The company processes roughly 45 billion AI tokens a day, and its agentic platform handles more than a million transactions a month, spanning customer provisioning and sales prospecting — work that increasingly runs without a human in the loop. SiliconANGLE · Nov 14
9. The AI graveyard is filling up
TechCrunch is keeping a running list of AI projects and startups that shut down or missed expectations. Character.ai filed for bankruptcy last month, five years after launching; privacy-focused assistant Ollie announced its shutdown weeks ago. S&P Global Market Intelligence estimates about 42 percent of AI initiatives are ultimately abandoned by their corporate parents. TechCrunch · Nov 14
10. China's Singles' Day is selling compute and avatars
During China's November 11 shopping festival, a beauty livestreamer sold AI memberships, and former world champion swimmer Ning Zetao appeared on a shopping livestream alongside an AI double of himself. The festival that once sold discounted goods is now a channel for AI subscriptions and digital personas. PYMNTS · Nov 14
Try This Next
- Test the Opus 5.5 prompting reset — strip "think hard" from your system prompt, run at Medium effort, and diff the output against your old High-effort runs. The Decoder
- Read up on the MathArena result — check what reasoning settings were used before comparing Carbon's price per task to Sol's. The Decoder
- Audit your agent spend — AT&T's 500 workflows and 45 billion daily tokens are a useful benchmark for what production-scale agentic usage looks like. SiliconANGLE
More Dailies
Content on Anagnorisis is summarized, paraphrased, and editorialized from publicly available sources for length and clarity. Original sources are linked where available. All trademarks belong to their respective owners.
