OpenAI’s Terra Puts Office Drudgery on Sale
Apparently, even AI has to justify its hourly rate.
OpenAI launched GPT-6.1 Terra on Monday, targeting the mindless invoices, forms, and app-hopping that consume the workday. Standard API pricing is $1 per million input tokens and $5 per million output tokens, which OpenAI describes as a 50% cut versus GPT-5.6 Terra and GPT-6.1 Sol.
OpenAI says Terra finished professional tasks over three times faster than GPT-6.1 Sol and GPT-6 Astra on Agents’ Last Exam. The pricier models still get its recommendation for the hardest coding jobs. Bargain shopping has limits, so don’t fire your senior devs just yet.
Independent testing adds useful nuance: Artificial Analysis found Terra guessed on 38% of questions it couldn’t answer, down from its predecessor’s 88%. Less shameless bluffing is progress, but your accounting software still deserves supervision.
Terra is available in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users, plus OpenAI’s API. It’s also reaching GitHub Copilot through a gradual rollout and is generally available on Amazon Bedrock and Microsoft Foundry.
Separately, Oracle’s expanded OpenAI partnership aims to connect ChatGPT with enterprise applications and databases. The common ambition? Put AI where the paperwork lives.
Cheaper models could make repetitive automation easier to justify, but the useful metric is what a correctly completed workflow costs after human review. If employees spend their saved time fixing the agent’s sloppy work, congratulations on automating the creation of another job.
Siri Finally Remembers the Conversation
Your assistant is developing a memory, and your update queue is developing urgency.
Apple released iOS 27.2 on Monday, finally giving Siri AI the ability to remember what you literally just said. Your digital assistant can now mine contextual details from your voicemails, calendar invites, and message attachments, while pulling somewhat better answers from your browsing history. Fewer infuriating rounds of “we already discussed this” are a welcome upgrade.
Of course, a smarter AI means its digital tentacles reach deeper into your life, making app permissions well worth revisiting. A newly renamed Excluded Apps setting lets you deliberately block the assistant from snooping in specific applications—because a helpful assistant doesn’t need an open invitation into the darkest, weirdest corners of your phone.
The Siri AI beta also now supports French, Japanese, Korean, Portuguese, and Spanish. Naturally, it requires an iPhone 15 Pro, Pro Max, or an iPhone 16 series model to run, and it remains glaringly unavailable on iPhones in the EU and mainland China.
Even if you refuse to talk to your phone, the update quietly patches 163 security flaws across iOS and iPadOS. These include an ironic Apple Intelligence vulnerability that might let remote attackers view your private information, plus a Hide My Email glitch that does the exact opposite by exposing your actual address. While Apple hasn’t reported active exploitation of these specific vulnerabilities yet, you should probably install the patch immediately.
Siri can remember your conversations; please don’t forget the security update.
Microsoft Offers 60,000 Workers an Exit
In a bold new interpretation of “AI-driven efficiency,” Microsoft offered voluntary buyouts to roughly 60,000 employees on Monday, making about 27% of its global workforce eligible for an exit package. That is just the invitation list, not a confirmed departure count. Participating teams include Copilot, Windows, Xbox, and finance.
The official exit offer covers selected employees up to the vice president level in the US, Canada, UK, and India. Participants have until early December to accept and would depart on Dec. 31, with terms resembling standard severance.
This follows a previous April retirement push for roughly 8,750 US employees based on age and tenure. More than 30% accepted. The new program casts a much wider net, proving you don’t need to be of retirement age to be offered the door.
The truly awkward detail: Copilot workers can qualify, while some data-center and cybersecurity staff are excluded. As Microsoft funds its massive AI infrastructure buildout, even working directly on AI doesn’t guarantee immunity from the cost-cutting menu.
Meanwhile, in Hollywood, Skydance is tossing out buyouts as it merges Paramount and Warner Bros. Discovery, alongside a mandatory five-day office return and looming layoffs. Different industry restructuring, similarly uncomfortable inbox. For workers, a “voluntary” offer still means weighing severance against the grim uncertainty of staying.
The next chapter of digital transformation comes with mandatory departure paperwork.
Amazon bans “tokenmaxxing” AI usage
Amazon issued internal guidelines discouraging software engineers from “tokenmaxxing,” the practice of using AI tools excessively without regard for output quality or cost. The company called metrics like token usage and code generation “deeply flawed” measures of AI adoption in a memo posted publicly after internal circulation in September.
- Amazon shut down an internal leaderboard in May after discovering employees were gaming the system by maximizing token usage to climb rankings, with no correlation to actual productivity gains.
- The memo emphasized that engineers should not use AI when it doesn’t improve their work, stating “if your work does not benefit from current AI tooling, that is a finding, not a failing.”
- Instead of tracking token volume, Amazon now measures “feature velocity”—successful production launches and meaningful customer-facing changes—as the real indicator of whether AI tools deliver ROI.
Amazon’s pivot away from token-based metrics signals a broader shift in how enterprises should evaluate AI productivity tools. Companies racing to adopt AI are discovering that usage volume doesn’t equal value, and rising costs without clear returns force a more disciplined approach. For business leaders, this is a reminder that blind AI adoption creates waste—strategic deployment tied to measurable outcomes wins.
Google DeepMind releases Genie 4, its most advanced world simulation model
Google DeepMind shipped Genie 4, a real-time world simulation model that generates photorealistic environments with native audio—including ambient sounds and interactive dialogue—while running continuously as users navigate and modify scenes through text prompts. The model integrates capabilities from the Gemini family: Gemini Omni grounds simulations in real-world appearance and physics, while other Gemini models provide reasoning to maintain coherence across multi-step events.
This marks Google’s first real-time world model to generate audio natively and simulate complex human behavior with interactive dialogue. Early enterprise partners are testing Genie 4 for scenario generation before building physical systems, potentially compressing months of pre-production work into rapid iteration cycles.
Nvidia and AMD Buy Into Meta
Nvidia and AMD plan to buy $4 billion of Meta stock, taking a slice of the up-to-$95 billion share sale funding Meta’s data center buildout. The equity comes bundled with expanded chip partnerships, so Meta’s biggest suppliers are now also helping finance the hardware they sell it.
Google Unveils Genie 4 World Model
Google introduced Genie 4, its next world model for generating interactive environments with AI. The update brings more realistic worlds and lets users change a scene while exploring it, keeping Google at the front of the world-model race that robotics, gaming and simulation startups are now betting on.
ChatGPT Plugs Into iCloud Mail
OpenAI added iCloud Mail and Calendar plugins to ChatGPT, letting Apple users connect their email and schedules the same way Gmail users already can. It pushes ChatGPT deeper into Apple’s ecosystem and turns it into a more practical personal assistant for the iPhone crowd.
Anthropic Starts Discovering Its Own Drugs
Anthropic says its scientists have spent months in deep research partnerships with six organizations developing new medicines, and that it has started drug discovery programs of its own. Its in-house work targets rare childhood muscular dystrophy, while partner projects span cancers and food allergies, a step from model vendor toward a biotech pipeline.
Anduril Builds American Laptop
Anduril and Qualcomm unveiled BlueChip, a rugged laptop whose first version uses about 90% US or allied components, including a Snapdragon chip built in Arizona. Palmer Luckey wants 100% domestic parts, even ones made only in China today, giving government buyers a supply chain they can audit.
AI Companies Race to Raise Cash
AI companies are raising money at a record pace. The data center buildout is so expensive that even Silicon Valley’s richest giants are burning cash, and they are now competing with one another to sell stock and debt to fund it, a sign of how capital-hungry the AI race has become.
OpenAI Sued Over Fatal Medical Advice
OpenAI faces a wrongful-death lawsuit alleging ChatGPT guided a woman with epilepsy to stop taking anti-seizure medication and told her to hide the changes from her doctors. Her family claims the advice contributed to her death. The allegations raise serious questions about AI medical guidance and chatbot liability.
Crypto Slump Drags Down Robinhood
Robinhood shares fell about 4% after third-quarter results. Revenue met expectations, helped by its push into options, prediction markets and other financial services, but a weak crypto market weighed on profits, showing how exposed the Menlo Park broker still is to digital-asset trading cycles.
Musk Accuses Ukraine of Starlink Violations
Elon Musk accused Ukraine’s military of violating Starlink’s terms by using its commercial network for offensive drone strikes, claiming SpaceX had closed the loophole. Ukrainian officials deny the allegation, saying such missions rely on the military-controlled Starshield system. The dispute highlights Ukraine’s dependence on privately controlled satellite infrastructure.
Draft Rules Threaten China Biotech Deals
Draft COINS regulations from the US Treasury Department have set off an intense debate over China biotech deals. The proposal mostly spares plain licensing deals but targets equity stakes, joint ventures and NewCo-style spinouts, which could narrow a dealmaking pipeline many US drugmakers now rely on for new drug candidates.
Korea Busts Illegal Tesla FSD Unlocks
Seoul police booked 80 people for using aftermarket devices to switch on Tesla’s Full Self-Driving in cars not approved to run it in South Korea, and seized 21 devices. The case shows how hard demand for FSD is pushing against the regulators that gate where it can be used.
Renewables Set to Overtake Gas Capacity
A SUN DAY Campaign analysis of EIA data projects renewables rising from 37.7% to 40.4% of US generating capacity within a year, overtaking natural gas. The projection includes estimated growth in small-scale solar, which EIA’s utility-scale forecast excludes. The figures measure installed capacity, not actual electricity generation.
Anthropic Discounts Claude for Governments
Anthropic expanded its partnership with the US General Services Administration. Under a new OneGov agreement, federal, state, local, tribal and territorial governments can buy Claude at 40% off, widening Anthropic’s push into the public sector as it competes with OpenAI and Google for government contracts.
Xbox COO Exits After Months
Xbox COO Helen Chiang has suddenly left Microsoft after 21 years and only months in the new role. The departure adds to leadership churn at Microsoft’s gaming unit as the company keeps cutting costs and reshaping teams around AI.
ANYbotics Ships Explosion-Proof Robot Dog
ANYbotics has started delivering ANYmal X, a legged inspection robot built to meet the strictest safety standards of the oil, gas and chemical industries. It can patrol explosive atmospheres autonomously, one of the clearest paying jobs for legged robots outside the lab.